lanying's blog

DeepSeek-V4-Pro Production Release – Performance Closing in on Fable 5 at 1/60th the Price

Thu, 13 Aug 2026 00:00:00 GMT

Overview

Late on August 12, 2026, DeepSeek quietly upgraded the V4 Pro model from preview to production. The update arrived without a formal launch event or release notes – the only indication was an update to the model version identifier in the API pricing page, now showing DeepSeek-V4-Pro-0813. The model name deepseek-v4-pro remains unchanged, so existing users can start using the new version without any code changes.

Earlier that same day, Elon Musk's xAI also officially released Grok 4.6, which scored 1753 Elo on the GDPVal-AA v2 leaderboard, surpassing Fable 5. The simultaneous release of two high‑value flagship models drew widespread industry attention.

This article analyses the V4 Pro production upgrade across several dimensions: performance gains, benchmark results, cost‑effectiveness, and selection guidance between Flash and Pro. It should be noted that as of the time of writing (August 13, 2026), DeepSeek has not yet published a complete changelog for V4-Pro-0813. Benchmark data in this article comes primarily from community sources and unofficial channels – official figures should be awaited for confirmation.

V4 Series Parameter Recap

To better understand where this upgrade fits, here's a quick recap of the V4 series specifications:

Specification	V4-Pro Preview	V4-Pro Production (0813)	V4-Flash Production (0731)
Total Params	1.6T	1.6T (unchanged)	284B
Active Params	49B	49B (unchanged)	13B
Architecture	MoE	MoE	MoE
Context	1M tokens	1M tokens	1M tokens
Max Output	384K tokens	384K tokens	384K tokens
Release Date	2026‑04‑24	2026‑08‑13	2026‑07‑31

Key takeaway: The model architecture has not changed. 1.6T total parameters, 49B active parameters, MoE architecture, 1M context window, and 384K max output length – all these hardware‑level specs have remained unchanged since the April 24 preview release. Similar to the Flash production release on July 31, the Pro production performance gains come primarily from extensive post‑training tuning.

In other words: the Preview version was the April baseline, Flash production was the July‑tuned lightweight version, and this Pro production release is the August‑tuned flagship version. Flash production had previously surpassed Pro Preview on Agent benchmarks; with Pro now receiving the same post‑training treatment, the performance gap has been corrected.

Ten Benchmark Comparisons: Preview vs Production

The following table shows cross‑verified benchmark comparisons. Note that Preview numbers are official scores from the April release, while Production numbers are from leaked community sources for the 0813 version – no independent third‑party reproduction results are available yet.

Benchmark	V4-Pro Preview	V4-Pro Production (0813)	Improvement
Terminal Bench 2.1	72.1	87.9	+21.9%
NL2Repo	38.5	61.5	+59.7%
Cybergym	52.7	83.3	+58.1%
DeepSWE	12.8	62.7	+389.8%
Toolathlon Verified	55.9	74.1	+32.6%
Agent Last Exam	16.5	25.7	+55.8%
Automation Bench (Public)	12.8	31.8	+148.4%
DSBench-FullStack	41.8	71.1	+70.1%
DSBench-Hard	31.1	67.2	+116.1%
HLE (w/ tools)	48.2	60.0	+24.5%

Key highlights:

DeepSWE jumped nearly 5× (12.8 → 62.7). This is DeepSeek's in‑house software engineering benchmark, evaluating the model's ability to modify code and fix bugs in real‑world engineering scenarios. The Preview's 12.8 was low; the Production's 62.7 surpasses Anthropic's previous‑generation flagship Opus 4.8 at 58.0 – a qualitative leap.
Terminal Bench 2.1 hit 87.9. This widely cited Agent benchmark evaluates the model's ability to execute terminal commands, modify configurations, and fix issues. The Production score of 87.9 is just 0.1 points behind the top‑ranked Fable 5 at 88.0.
Cybergym surpassed Fable 5: 83.3 vs 83.1. This benchmark tests cybersecurity attack/defence capabilities – surpassing the closed‑source leader in this domain is a significant milestone.
Automation Bench and DSBench series doubled: automation workflows and full‑stack development capabilities both saw multiplicative gains – a generational leap.

Overall, the performance gap between Preview and Production is substantial.

Horizontal Comparison: Against Flash Production and Competitors

The following table shows leaked cross‑benchmark comparisons for the 0813 release against mainstream flagship models including Fable 5, Opus 4.8, GLM‑5.2, and Kimi K3:

Benchmark	V4 Pro Production	V4 Flash Production	GLM‑5.2	Kimi‑K3	Opus‑4.8	Fable 5
HLE (w/ tools)	60.0	—	54.7	56.0	57.9	63.0
Terminal Bench 2.1	87.9	82.7	81.0	88.3	85.0	88.0
NL2Repo	61.5	54.2	48.9	—	69.7	—
Cybergym	83.3	76.7	—	80.0	78.3	83.1
DeepSWE	62.7	54.4	46.2	67.5	58.0	70.0
Toolathlon‑Verified	74.1	70.3	59.9	76.5	76.2	77.9
Agent Last Exam	25.7	25.2	23.8	27.6	25.7	—
AutomationBench	31.8	25.1	12.9	30.8	27.2	29.1
DSBench‑FullStack	71.1	68.7	61.8	73.7	71.6	77.2
DSBench‑Hard	67.2	59.6	54.5	63.0	71.7	68.3

The data shows a clear profile for Pro Production – it may not top every single benchmark, but it sits firmly in the top tier across all of them. It trails Kimi K3 by just 0.4 points on Terminal Bench (88.3 vs 87.9) but leads Opus‑4.8 (85.0) by 3 points; it takes first place on Cybergym at 83.3; both DeepSWE and NL2Repo are among the top; DSBench‑Hard at 67.2 is second only to Opus‑4.8's 71.7.

"No obvious weak spots" is a core competitive advantage for Agent models. Agent tasks aren't single‑turn Q&A – they involve multi‑step long‑chain execution: terminal operations, code modification, tool calls, data parsing, environment setup, and more. A deficiency in any single capability can cause the entire chain to fail. The Preview's weak performance on long‑code tasks like DeepSWE (12.8, indicating a high failure rate) was a significant limitation; with that gap now addressed, the model's real‑world Agent usability has improved dramatically.

It's also worth noting that the performance gap between Flash Production and Pro Production has returned to "Pro slightly ahead" – about 5–10 points across benchmarks. Despite having a fraction of Pro's parameters (284B vs 1.6T total, 13B vs 49B active), Flash remains exceptional value.

The Technical Path Behind the Performance Gains

Consistent with the Flash production upgrade, Pro's performance gains come primarily from post‑training optimisation.

DeepSeek hasn't published a technical report for Pro Production yet, but based on Flash's changelog, it's reasonable to assume the same upgrade strategy was applied – the model architecture, parameter count, and active parameter count were all left untouched, while the post‑training pipeline was thoroughly reworked. To use an analogy: it's the same engine, but the transmission, ECU calibration, and suspension have been retuned – delivering better performance without hardware changes.

This suggests that under the MoE + post‑training technical roadmap, there's still significant room to push the performance ceiling. Especially for long‑chain Agent scenarios, the model architecture itself is already capable – the bottleneck lies more in large‑scale post‑training tailored to code toolchains, terminal operations, and multi‑step reasoning. Training data quality and tuning strategies may have a greater impact on final performance than simply adding more parameters.

For context, here's a quick recap of V4's core architecture:

MoE (Mixture of Experts): The 1.6T parameters are split into thousands of "expert sub‑models"; each forward pass activates only 64 experts (totalling 49B parameters) – meaning the model has the knowledge capacity of 1.6T but the inference compute cost of a 49B model.
Hybrid attention (CSA + DSA + mHC): Reduces computational overhead in long‑context scenarios – at 1M tokens, per‑token inference cost drops to about 10% of V3.2, and KV Cache usage is significantly lower.
Engram memory technology: Improves long‑context precision recall – better at retrieving details from large codebases and long documents.

New Capabilities: Multimodal Support, Responses API, and Codex Integration

The Preview was text‑only; Production natively supports image reasoning for the first time. According to 36Kr's hands‑on testing, the DeepThink engine can analyse screenshots, interpret images, and process mixed text‑image documents within a single reasoning flow. For example, submitting an error screenshot allows the model to directly read the code, error messages, and stack trace from the image and propose a fix – eliminating the need for separate OCR preprocessing in development workflows.

On the API side, Pro Production shares the same interface capabilities as Flash Production:

Native Responses API (OpenAI‑compatible) support – Codex tools (Codex CLI, ChatGPT desktop, VS Code Codex plugin) can be integrated by simply changing the base_url, with the same automated setup script as Flash
Also supports Anthropic API format – Agent tools like Claude Code and OpenCode can switch backends by changing just one line of code
Thinking mode toggle (enabled by default), JSON Output, Tool Calls, conversational prefix completion (Beta), and FIM completion (Beta, non‑thinking mode only) are all supported

In short, if you already have code calling Flash, migrating to Pro is as simple as changing the model parameter from deepseek-v4-flash to deepseek-v4-pro.

Pricing & Concurrency: 6 CNY/MT Output, Three Times Flash

The following pricing is from the official pricing page as of August 13, 2026:

Pricing Item	V4 Pro Production (Off‑Peak)	V4 Flash Production (Off‑Peak)	Ratio
Input (cache hit)	0.025 CNY	0.02 CNY	1.25x
Input (cache miss)	3 CNY	1 CNY	3x
Output	6 CNY	2 CNY	3x
Concurrency Limit	500	2500	1/5

Pro is priced at about three times Flash – input cache miss rises from 1 CNY to 3 CNY, output from 2 CNY to 6 CNY, with cache‑hit pricing much closer. The concurrency difference is even more pronounced: Flash at 2500 vs Pro at 500 – consistent with their positioning (Flash for high‑volume requests, Pro for high‑value long tasks).

Peak‑hour pricing doubles (the time‑of‑use mechanism applies to Pro as well): 9:00‑12:00 and 14:00‑18:00 Beijing time (7 hours per weekday) – all pricing items double, meaning Pro output during peak hours is 12 CNY per million tokens.

International competitor pricing comparison (converted to USD):

Model	Input (USD / MT)	Output (USD / MT)
DeepSeek V4 Pro	0.435	0.87
Claude Opus 4.8	5.00	25.00
GPT‑5.5	5.00	30.00
Fable 5	10.00	50.00

V4 Pro's output price is about 1/28 of Opus 4.8 and 1/57 of Fable 5 – roughly on par with Grok 4.6's output price ($6). Combined with its Terminal Bench 2.1 score just 0.1 points behind Fable 5, the claim of "performance closing in on Fable 5 at one‑sixtieth the price" is well grounded.

Two additional points:

DeepSeek has already posted a price increase notice on the pricing page: "We plan to raise DeepSeek API pricing in the near future, with a significant increase expected – please plan your usage accordingly." The current pricing window may be limited.
The SWE‑bench Verified 80.6% figure has not been included in the main analysis, as it has only appeared in encyclopaedia entries and third‑party articles without official confirmation – treat it as unofficial until formally published.

Flash vs Pro – Selection Guide

Use Case	Recommendation	Reasoning
Daily chat, light coding, content generation, Q&A	Flash	Flash Production already surpasses Pro Preview in performance, at 1/3 the cost
Large codebase refactoring, multi‑step terminal ops, complex Agent tasks	Pro	Clear advantages on long‑chain benchmarks like DeepSWE and Terminal Bench
Data analysis / data science (DSBench‑type tasks)	Leans Pro	DSBench‑Hard/FullStack show a performance gap; better completion rates on complex data tasks
High‑concurrency, low‑latency API services	Flash	5× the concurrency limit (2500 vs 500) at 1/3 the cost
Ultra‑long context (entire books / large codebases)	Either; Pro if budget allows	Both support 1M context; Pro has better recall and reasoning stability in long‑chain tasks
Security / red‑team scenarios	Pro	Cybergym 83.3 vs 76.7 – a clear edge
Budget‑sensitive / individual developers	Flash primarily; switch to Pro for complex tasks	Flash handles most tasks well; significantly better cost efficiency

Bottom line: For 95% of daily tasks, Flash Production is more than sufficient. Pro's 3× price doesn't translate to a 3× experience improvement in most scenarios. Pro's strengths are concentrated in long‑chain Agent tasks, high‑difficulty coding, and complex data science – and for many users, the "small difference" they perceive is simply because their tasks aren't demanding enough to reveal the 5‑point gap.

Additionally, according to 36Kr's hands‑on testing, there are two practical considerations:

In large‑scale code generation, Thinking mode can consume 80%+ of output tokens on reasoning, compressing the actual content. In BFS/DFS/A* visualisation tests, Thinking mode truncated output to about half, while disabling it allowed full execution in 54 seconds.
Third‑party intelligence scores show divergence: Artificial Analysis gives V4‑Pro‑0813 an intelligence index of 53, only 1 point above Flash‑0731's 52 – reinforcing the view that "the gap isn't noticeable in everyday scenarios."

Also worth watching is the DeepSeek Harness inference framework, currently in closed beta. The Harness team opened beta signups on August 1, with early testers reporting strong results. The framework's impact on model performance may be larger than expected – whether the benchmark scores above are boosted by Harness optimisation is currently unconfirmed. Once Harness is publicly released, Pro's practical performance may have further room to improve.

Concurrency, API Migration, and Practical Tips

Concurrency Limits

500 concurrent requests is plenty for most applications, but it's important to understand the positioning difference from Flash's 2500. Pro is designed for high‑value long tasks – using it for bulk text processing, data cleaning, etc., may hit concurrency bottlenecks, and peak‑hour costs double, making it less economical. Batch processing is best scheduled during off‑peak hours (nights/weekends) to halve costs; for non‑latency‑sensitive tasks, Flash reduces cost by about two‑thirds and eliminates concurrency concerns.

API Migration Guide

If you're already using DeepSeek API, switching to Pro is as simple as changing the model name:

from openai import OpenAI

client = OpenAI(api_key="<your_api_key>", base_url="https://api.deepseek.com")

# Original Flash call
response = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role":"user","content":"Write a quicksort implementation"}],
    temperature=1.0,
)

# Switch to Pro – all other parameters stay the same
response = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role":"user","content":"Write a quicksort implementation"}],
    temperature=1.0,
)

For Anthropic‑format integration, simply change the base_url to https://api.deepseek.com/anthropic.

Codex Integration

Pro also supports Codex integration with the same setup as Flash – use the automated PowerShell script:

irm https://cdn.deepseek.com/api-docs/codex-deepseek-setup-en.ps1 | iex

The script backs up existing configs, writes to models.json, modifies config.toml, and validates syntax – aborting if anything goes wrong.

Cost Optimisation Tips

Schedule batch tasks during off‑peak hours (12:00‑14:00, 18:00‑09:00, weekends/holidays) to halve costs.
Place fixed system prompts and knowledge‑base prefixes at the very start of the messages array to maximise cache hit rates. Pro's cache‑hit input pricing is 0.025 CNY per million tokens – about 120× cheaper than cache‑miss at 3 CNY.
Route tasks by complexity – simple intent classification, categorisation, and summarisation can use Flash at 1/3 the output cost, saving significantly over time.

Pricing Trends and Industry Context

The "significant API price increase" notice on the pricing page is worth paying attention to. DeepSeek posted a related announcement on August 6, and the Pro production release didn't coincide with a price adjustment – the length of this low‑price window is uncertain. DeepSeek has recently raised substantial funding at a high valuation, but compute costs remain a core constraint – V4 Flash alone has surpassed 4.66 trillion tokens in weekly calls on OpenRouter, topping the global charts for several consecutive weeks. The pressure behind such massive usage at consistently low prices is real.

Time‑of‑use pricing is essentially resource scheduling via price signals: when compute is scarce during weekday daytime hours, prices rise, shifting non‑urgent batch workloads to off‑peak periods to ensure stability for finance, development, and real‑time Agent scenarios. For developers, rather than worrying about price fluctuations, it's better to build tiered routing and caching optimisation into your architecture – even if prices rise later, well‑prepared users can still keep costs under control.

Summary

V4 Pro Production represents a substantive generational leap, not a minor iteration: DeepSWE improved nearly 5×, Terminal Bench hit 87.9 (just 0.1 points behind Fable 5), Cybergym surpassed Fable 5, and multimodal and Codex support have been added. Combined with the 6 CNY per million tokens output price, it's one of the most cost‑effective flagship‑class Agent models available today – far cheaper than comparable overseas models.

But Pro isn't the best choice for every scenario: For everyday tasks, Flash Production already outperforms Pro Preview, at one‑third the cost and five times the concurrency. If Flash already meets your needs, don't pay extra just for benchmark numbers. Only when you encounter long‑chain Agent tasks where Flash frequently fails should you consider switching to Pro.

A final note: this article was written on August 13, 2026. Some data comes from community sources and third‑party reports – DeepSeek has not yet published a complete changelog. Treat benchmark numbers as references, and validate performance against your own use cases.

With Pro Production now live, the V4 series product matrix is largely complete – Flash offers extreme cost‑effectiveness, Pro delivers flagship performance, with output pricing ranging from 2 to 6 CNY per million tokens, further lowering the global AI API pricing bar. For developers, this trend is broadly positive.

References

Information compiled from the following public sources. These are from the internet and not guaranteed to be fully accurate – for reference only. Prices, policies, and benchmark methodologies are subject to change – always refer to DeepSeek's official API documentation and latest announcements.

EdgeOne vs ESA – How Strong Is EdgeOne, What's the Difference, and Which One Should You Choose?

Wed, 12 Aug 2026 00:00:00 GMT

Product Positioning and Background

Edge security acceleration platforms are the next evolution of traditional CDN products. They integrate static acceleration, dynamic full-site acceleration, WAF (Web Application Firewall), DDoS mitigation, bot risk management, edge computing, and more – all unified under a single site-based management model. This contrasts with traditional setups where CDN and security products are separate, requiring multi-console operations. Both Tencent Cloud EdgeOne and Alibaba Cloud ESA belong to this category, positioning themselves as competitors to Cloudflare.

Tencent Cloud EdgeOne (EO)

EdgeOne was officially launched in 2023, evolving from Tencent's SCDN product. Its core feature is site-based management – once you add a site, all subdomains are automatically covered without needing to configure each domain individually or repeatedly maintain SSL certificates. It is Tencent Cloud's next-generation flagship CDN product, officially positioned as a Cloudflare competitor.

Alibaba Cloud ESA

ESA stands for Edge Security Acceleration, rebranded and rebuilt from the original DCDN (Dynamic Content Delivery Network) product in 2024. In the DCDN era, static and dynamic acceleration capabilities were independent; ESA consolidates them while strengthening native security modules. The API version 2024-09-10 indicates a major architectural overhaul.

Both vendors claim 3200+ global nodes, with 2300+ nodes in mainland China. These figures should be taken as a reference – actual node quality and routing logic should be verified through real-world testing. Both have top-tier domestic node coverage in China, significantly outperforming many overseas vendors that claim "global nodes" but lack actual domestic resources.

Free Tier Comparison

The table below summarises the core specifications of each product's free tier:

Feature	Tencent Cloud EdgeOne Free	Alibaba Cloud ESA Free
Price	0 CNY/month	0 CNY/month
Traffic	Unlimited	Unlimited
Requests	Unlimited	Unlimited
Validity	Permanent (as long as site isn't deleted)	Permanent (supports 0‑CNY renewal)
How to Get	Limited beta – via redemption codes or official campaigns	Direct 0‑CNY subscription from the console
Max Per Account	Up to 5 (via codes + campaigns)	1 per UID (additional via campaigns)
Sites	1	1
Subdomains	200 subdomains per site	1 wildcard domain only; no individual subdomain support
Rule Engine	20 rules	Only basic rules available
SLA	None	None
Support	Community only	No ticket support
Mainland China Nodes	Requires domain ICP filing	Requires domain ICP filing
Rate Limiting	Single‑thread throttling on large files; ~500 KB/s for files >10 MB (observed)	Peak bandwidth and per‑request rate limits apply (undisclosed thresholds)

Common ground: zero cost, unlimited traffic/requests, permanent availability, no SLA guarantees; both require ICP filing for mainland China node access. Note that "unlimited traffic" doesn't mean unlimited throughput or no business‑use constraints – implicit restrictions exist.

Getting the Free Tier

Alibaba Cloud ESA is straightforward – after account verification, you can directly order the free tier from the console. One site per account by default. Partner campaigns (e.g., with aaPanel) may offer additional site quotas for multi‑site scenarios.

Tencent Cloud EdgeOne is still in limited beta and not directly purchasable. There are three channels:

Community sharing campaigns – complete speed tests or edge function trials, share results on social media; each task unlocks one free plan, up to 4 per account.
Redemption codes – distributed via KOL partnerships, cloud vendor member events, or developer cooperation programs. Codes are valid for 7 days, one use per account.
Only the main account (not sub‑accounts or collaborators) is eligible; the account must be verified.

EdgeOne also partners with third‑party communities to distribute free plans – e.g., the aaPanel collaboration: docs.bt.cn/notices/free-edgeone-plan. The barrier to entry has been progressively lowering.

Site and Domain Quota Differences

EdgeOne Free supports 200 subdomains per site – ideal if your main domain has many sub‑projects or micro‑services, all manageable under a single site.

ESA Free only supports a single wildcard domain. While it covers subdomains, you cannot configure per‑subdomain caching or access control rules. Combined with limited rule engine quotas, fine‑grained management is constrained.

EdgeOne also has an automatic downgrade mechanism: if a paid plan expires, the site isn't shut down – it automatically downgrades to the free tier, losing SLA but keeping the service running. Great for users who worry about forgetting to renew.

Rate Limiting Details

Both products advertise unlimited traffic but impose restrictions on large‑file use cases.

EdgeOne Free explicitly states it's not suitable for video or large‑file distribution. In real‑world tests, files over 10 MB are throttled to ~500 KB/s on a single thread – a pattern confirmed by multiple independent tests.

ESA Free doesn't publish a fixed throttling threshold; documentation only mentions peak bandwidth and per‑request rate limits exist. Community tests show no single‑file throttling like EdgeOne; actual download performance depends more on origin bandwidth. However, the free tier is intended for personal projects and development/testing – using it for high‑traffic production may trigger risk‑control throttling.

Real‑World Performance and Node Coverage

Actual performance depends on file size, geographic location, and origin conditions. There's no absolute winner. Summary by scenario:

Small files (<10 MB): EdgeOne performs better. EdgeOne provides L2 regional caching – if an edge node misses, it pulls from a regional cache node, improving hit rates across multiple edge nodes. ESA Free only has single‑layer edge caching, making cold content more likely to trigger origin pulls and causing greater performance variance. ESA's three‑tier caching is only available in paid plans.
Large files (>10 MB): ESA is better. EdgeOne Free's single‑thread throttling significantly impacts large‑file transfers. ESA has bandwidth caps but no per‑file throttling – provided your origin has sufficient outbound bandwidth.
Node coverage: ESA has slightly more domestic nodes; EdgeOne has broader overseas coverage with Anycast routing, mainly via Hong Kong, Japan, and Singapore nodes – better international access. ESA Free does not allocate Hong Kong direct‑connect nodes, making it nearly unusable for un‑filed domains within China.
Cache preheating is unavailable in both free tiers – it's a paid‑only feature.

Security Capabilities Comparison

Both free tiers include platform‑level DDoS mitigation, basic WAF, basic CC protection, and automatic SSL certificate issuance/renewal – sufficient for personal site security needs.

EdgeOne Free offers more practical security features for site owners:

AI crawler management – intercept, challenge, or allow GPTBot, ClaudeBot, etc., while maintaining proper search engine indexing.
Pre‑configured vulnerability protection rules to block common Web vulnerability scans.
Multi‑dimensional access control – IP, geographic region, Referer, User‑Agent, and ASN‑based policies.
Adaptive CC frequency control to mitigate malicious request floods.

ESA Free leans more basic – WAF and DDoS are available, but advanced capabilities like bot management and managed protection rules require upgrading to the Standard plan. ESA's advantage lies in origin protection – paid plans support encrypted tunnels between edge nodes and the origin to hide the origin IP.

Edge Ecosystem and Additional Capabilities

Many users view these solely as CDNs, but both are complete edge platforms with additional capabilities.

Static Site Hosting

EdgeOne Pages integrates with GitHub for auto‑build and deployment from source code, with automatic domain binding and certificate issuance. It also supports Pages Functions (edge functions) and KV storage. Static blogs and documentation sites can be deployed with dynamic logic (comments, analytics, on‑site search) via edge functions – no server needed.

ESA also offers Pages static hosting with one‑click blog/demo deployment and edge function support. However, the ecosystem and documentation are less mature than EdgeOne Pages.

Edge Functions and Edge Containers

Both offer edge functions – JavaScript code executed at edge nodes, with auto‑scaling and zero maintenance. EdgeOne Free provides 3 million function invocations per month; exceeding that disables function rules. ESA Free also offers basic edge function capabilities.

Additionally, ESA supports edge containers – you can run container images directly, suitable for complex/heavy workloads in enterprise scenarios. EdgeOne currently focuses on edge functions as the primary solution.

Layer‑4 Proxy (L4)

Layer‑4 proxy forwards TCP/UDP traffic, enabling low‑latency access and DDoS protection for non‑HTTP services – games, real‑time audio/video, etc. This is primarily for enterprise paying customers and rarely used by personal sites.

DNS Hosting and Certificates

Both support DNS hosting. With NS‑mode access, resolution, acceleration, and certificates are all managed in one console, with DNS‑level DDoS protection. ESA supports CNAME flattening; enterprise versions support wildcard certificates.

Tip: If you're already using partitioned DNS resolution with another provider, stick with CNAME‑mode access rather than switching NS to avoid configuration conflicts.

Logging and Other Features

EdgeOne Free supports basic community querying; full offline logs and real‑time logs are paid add‑ons. ESA Free doesn't support offline log export. Features like staged environment management and CAPTCHA sliders require ESA Standard or above.

Onboarding and Common Pitfalls

Both support NS‑mode (change domain DNS servers) and CNAME‑mode (add CNAME records with your existing DNS provider) – choose based on your existing DNS setup.

Mainland China nodes require domain ICP filing. Un‑filed domains can only use "Global (excluding mainland China)" acceleration regions. EdgeOne's international site claims free access without filing; the China site requires it. ESA Free is effectively unusable without filing.

Frequently encountered issues:

ESA Free has a review process after site submission – in some cases up to 2 business days. Plan ahead if your site is time‑sensitive.
EdgeOne may temporarily scale back edge nodes under large DDoS attacks – it auto‑recovers, but access quality may degrade during peak attacks.
Both free tiers are intended for testing and personal non‑critical use – not recommended for production workloads. Community cases show ESA Free triggering risk‑control throttling under high traffic.
Product policies, campaigns, and quotas evolve rapidly – always check the latest official documentation before subscribing.

Paid Plan Pricing Reference

Free tiers are only suitable for testing and personal small sites. As your business scales, you'll need to upgrade. Price comparison:

Tier	Tencent Cloud EdgeOne	Alibaba Cloud ESA
Entry	Personal: 29.9 CNY/mo (50GB / 3M requests)	Basic: 9.9 CNY/mo (50GB)
Mid	Basic: 399 CNY/mo (500GB / 20M requests)	Standard: 375 CNY/mo (500GB)
High	Standard: 3800 CNY/mo (3TB / 50M requests)	Advanced: 3600 CNY/mo (3000GB)

For comparable tiers, ESA is cheaper. Blocked traffic, HTTPS, and WAF requests are not counted against quotas – you won't get surprise bills during attacks. EdgeOne's paid advantages lie in feature completeness, intelligent acceleration strategies, overseas link quality, and the Pages ecosystem – ideal for users willing to pay for capabilities.

Selection Recommendations

Use Case	Recommendation	Notes
No ICP filing, want free access from China	EdgeOne International free tier	Hong Kong/Japan/Singapore nodes handle China access via Anycast; ESA Free has no Hong Kong direct‑connect, poor usability without filing
Filed domain, personal blog / docs site with low traffic	EdgeOne preferred	Higher cache hit rates for small files, 200 subdomains per site, rich security extras
Image hosting, download sites, large‑file distribution	ESA	No single‑file throttling unlike EdgeOne
Need to manage many independent sites	EdgeOne	Up to 5 free plans per account
Need zero‑cost static hosting + edge capabilities	EdgeOne Pages	GitHub integration, auto‑build, KV, edge functions
Formal paid production workload	ESA	Lower cost at same tier, no bill inflation under attacks

Important Notes

Free tiers are officially positioned for personal projects and development/testing – no SLA guarantees. Do not use them for core production workloads – risk of throttling or suspension.
You can use both products simultaneously – split domains between them, or use smart DNS to route traffic across both to complement each other.

Summary

For personal non‑critical workloads, EdgeOne Free offers stronger overall capability. Though the application process is more involved, it provides generous subdomain quotas, excellent small‑file acceleration, good overseas access, and built‑in Pages static hosting.

For large‑file distribution, or if you plan to purchase a paid plan later, ESA has clear cost advantages – simple to activate and budget‑friendly.

Whichever you choose, domain ICP filing is mandatory for mainland China node access.

Note: Both products iterate quickly. Quotas, restrictions, and campaign rules in this article may change – always refer to official documentation for current information.

References

Information compiled from public sources for reference only; not guaranteed to be fully accurate. Always check official documentation for the latest rules.

DeepSeek-V4-Flash Production API Public Beta

Fri, 31 Jul 2026 00:00:00 GMT

Introduction

As of today (July 31, 2026), DeepSeek has officially launched the public beta of the DeepSeek-V4-Flash production API. It's been about three months since the V4 preview release on April 24 – and this update boils down to one key phrase: a massive leap in Agent capabilities.

DeepSeek has directly published results from nine benchmarks, with multiple metrics significantly surpassing the previous V4-Pro-Preview. You could even say it's "punching above its weight" – reminiscent of the original V4 release that turned the hierarchy on its head.

A Quick Recap of the V4 Series

On April 24, 2026, DeepSeek dropped the V4 series unannounced, with two versions: Pro and Flash.

Comparison	V4-Pro	V4-Flash
Total Parameters	1.6T	284B
Active Parameters	49B	13B
Architecture	MoE	MoE
Context	1M tokens	1M tokens
Max Output	384K tokens	384K tokens

Pro is designed for maximum performance, going head‑to‑head with closed‑source flagship models. Flash, on the other hand, is much smaller in parameter count, trading size for lower latency and cost – all about cost‑effectiveness. What's interesting is that this production Flash now pushes Agent capabilities even beyond the preview Pro – despite having roughly an order of magnitude fewer parameters.

Nine Benchmark Results

Here are the full results from the official changelog:

Benchmark	Score
Terminal Bench 2.1	82.7
NL2Repo	54.2
Cybergym	76.7
DeepSWE	54.4
Toolathlon Verified	70.3
Agent Last Exam	25.2
Automation Bench (Public)	25.1
DSBench-FullStack (internal)	68.7
DSBench-Hard (internal)	59.6

A few things worth highlighting:

Terminal Bench 2.1 scored 82.7 – DeepSeek previously reported V4-Pro-Preview at 67.9 on Terminal Bench 2.0. While 2.0 and 2.1 aren't directly comparable, it's still impressive for a lightweight model with only 13B active parameters.
Cybergym at 76.7 – a cybersecurity attack/defence simulation test, showing solid performance in security‑adversarial scenarios.
DSBench-FullStack and DSBench-Hard are internal test sets for full‑stack development and hard Coding Agent tasks. Publishing internal benchmarks signals DeepSeek's intention to position Flash as the go‑to foundation for developers and Agent applications.

The official disclosure also noted that for the public‑benchmark Code Agent tasks, they used the upcoming DeepSeek Harness minimal mode as the testing framework, with max sampling, top_p=0.95, and temperature=1.0. Worth noting that this is the first time DeepSeek has officially named its own Harness testing framework – suggesting they're not just building stronger models, but also aiming to establish a standardised Agent evaluation ecosystem.

Same Model Architecture, Stronger Performance

The official statement is clear: The model architecture and size of DeepSeek-V4-Flash-0731 remain identical to V4-Flash‑preview – only the post‑training has been redone.

The base model hasn't changed – what's changed is the training methodology. This sends a clear signal: in the Agent capability race, there's still enormous room for improvement through post‑training strategies. Even with the same architecture, finer post‑training tuning can produce qualitative leaps in benchmark performance – which is more instructive for the industry than simply stacking more parameters.

It's also worth mentioning that the V4 series has made aggressive optimisations for long‑context efficiency. In 1M‑token scenarios, Flash's per‑token inference cost is only about 10% of V3.2, and KV Cache usage has dropped significantly – long context is no longer a "luxury for the wealthy." This is one of the reasons Flash can keep its pricing so low.

Native Responses API Support with Codex Adaptation

The production V4-Flash natively supports OpenAI's Responses API format and is specifically adapted for Codex.

All Codex clients – Codex CLI, ChatGPT desktop, and the VS Code Codex plugin – share the same configuration file. One configuration applies to all of them. DeepSeek also provides a one‑line PowerShell auto‑setup script:

irm https://cdn.deepseek.com/api-docs/codex-deepseek-setup-en.ps1 | iex

The script automatically backs up existing configs, writes to the models.json model registry (declaring context windows, inference tiers, tool call formats, etc.), modifies config.toml, and validates syntax before writing back – aborting if anything goes wrong. Pretty reliable.

One thing to note: currently only deepseek-v4-flash supports Codex integration; V4-Pro is expected to gain support in early August.

Pricing

Flash continues its value‑for‑money approach. Official pricing (per million tokens, off‑peak):

Item	Off‑Peak Price	Peak Price
Input (cache hit)	0.02 CNY	0.04 CNY
Input (cache miss)	1 CNY	2 CNY
Output	2 CNY	4 CNY
Concurrency Limit	2500	-

Compared to V4-Pro (off‑peak input cache hit 0.025 CNY, cache miss 3 CNY, output 6 CNY), Flash's cost advantage is clear – especially with the generous 2500 concurrency limit for high‑throughput scenarios.

Peak/Off‑Peak Pricing

The V4 series also introduces an interesting mechanism: time‑of‑use pricing, essentially bringing electricity market logic to compute:

Peak hours: Weekdays, Beijing time 9:00‑12:00 and 14:00‑18:00 (7 hours per day)
Off‑peak hours: The remaining 17 hours, plus weekends and public holidays

All pricing doubles during peak hours – think of it as a "meter" on your tokens. Since the GPT‑3 era, large‑model API pricing has been flat‑rate; DeepSeek is the first in the industry to try this.

Two important details:

This only applies to developer API calls – C‑端 users on the App/Web interface see no price change
Off‑peak pricing matches previous rates – it's not a "discount" after a price hike. The peak premium buys you priority compute access

For developers, this opens up cost‑saving strategies: data cleaning, document summarisation, offline batch processing – tasks that don't need real‑time responses – can be shifted to nights or weekends to cut costs in half. For real‑time Agents, online customer service, and other peak‑sensitive scenarios, you'll need to optimise prompt length and caching to keep costs down. And with cache‑hit input pricing as low as 0.02 CNY per million tokens, caching your system prompts and knowledge‑base prefixes can further reduce spend.

Upgrade Scope

It's important to note that this upgrade only covers the V4-Flash API:

Updated: DeepSeek-V4-Flash API
Not updated: DeepSeek-V4-Pro API
Not updated: App / Web model endpoints
Coming soon: DeepSeek-V4-Pro production release (official says "as soon as possible" – rumoured for early August)

So if you're using the DeepSeek App or web interface, you won't notice any difference today. Also, the old deepseek-chat and deepseek-reasoner model names were deprecated on July 24 – they now map to deepseek-v4-flash in non‑thinking and thinking modes, respectively.

More Technical Insights into the Maimai DX Cabinet – V2

Thu, 30 Jul 2026 00:00:00 GMT

Original post

Changes in Newer Batch Cabinets

If you look closely, you'll notice some hardware changes in newer batches. These include the removal of the player camera (which was useless for the Chinese server anyway), and an upgrade of the base board from HX2 to HX2.1.

Removing the player camera is actually a good thing – the USB hub power supply issue that caused a chance of 3100 errors on boot becomes less likely. Instead of leaving it there as a useless decoration, it's better to just remove it. The downside is that the Chinese server will most likely never enable this feature again – and you also lose a spot to put stuff on :)

The upgrade from ALLS HX2 to HX2.1 should bring some performance improvements, though exactly how much is unclear. It could also just be supply chain management at work :)

Here's the comparison between HX2 and HX2.1:

Component	HX2	HX2.1
Motherboard	Gigabyte MCH31AM-S2 H310	Gigabyte MCH31AM-S2 H310
CPU	Intel Core i3-8100	Intel Core i3-8100
RAM	2x4GB	2x4GB
GPU	NVIDIA GeForce GTX1050 Ti	NVIDIA GeForce GTX1650 Ti
SSD	TDK GBDisk GS1 120GB	TDK GBDisk GS1 120GB
HDD	WD5000LUCT 500GB	WD5000LUCT 500GB

As you can see, apart from the GPU, there's not much change – mainly because the GTX 1050 Ti was discontinued.

I only know for sure that the Chinese server cabinets got this upgrade. Not sure if the Japanese or international versions have been updated – I haven't found any info on that.

There's a cabinet near me that's already been upgraded to HX2.1, but honestly, I didn't notice any significant difference in my personal experience. Maybe the HX2 just wasn't cutting it anymore and that's why they upgraded 🙂

Buttons, Screen, and Touch Panel

1. Screen

In my previous article, I mentioned why Japanese cabinets don't need as much force to make the screen flash white. But after some testing and playing on mysterious machines at mysterious locations, I've realised there might be more factors at play.

Both international and Japanese cabinets give me a feeling that they're not as solidly built as the Chinese ones – they feel softer. And when I actually looked at them, the gap between the touch panel and the screen is actually pretty similar to the Chinese cabinets. Besides, the touch panel is still a piece of glass – no matter how much it deforms, it's unlikely to hit the screen directly. So my suspicion is that the metal frame holding the touch panel might have different thickness or build quality compared to the Chinese version. After all, the Chinese cabinets are built locally based on specs from Japan – and hey, "Made in China" quality is something else 🙂

Apart from the metal frame, it could also be the wooden board that holds the frame and screen together – the materials might differ. International and Japanese cabinets also use different screens from the Chinese ones, so screen quality could be a factor too. When you hit the buttons or the screen, the force transmits through the metal frame to the wooden backboard. If that board flexes more, the screen might shift back and forth. And since the screen isn't directly bolted to the backplate – just held by a few screws and brackets – if the screen is more prone to deformation, the LCD panel could shift and hit the backlight, causing that white flash. The Chinese cabinets might use sturdier wood and screens, so the same force doesn't cause the flash. Again – Made in China 🙂

Newer cabinets use BOE screens, model DV430FHM-NN1, with a resolution of 1920x1080, 16:9 aspect ratio, and 60Hz refresh rate.

2. Buttons and Touch Panel

2.1 Buttons

Honestly, I personally prefer the feel of the original Japanese buttons over the Chinese ones. I like the softer feel of the Japanese buttons, and the click sound is crisp and satisfying – the Chinese buttons sound dull and feel less responsive.

Once you've tried the Japanese buttons, it's hard to go back – the sound is just so pleasant. But there are ways to optimise Chinese buttons to get closer to that Japanese feel.

For the feel: The most direct method is to replace the spring leaves. Just search for the Japanese copper spring leaves – you can find them on various channels, or if you have the means, buy directly from SEGA in Japan. They're also available on second‑hand markets or Taobao.

For the sound: The dull sound of Chinese buttons is mainly due to the EVA foam padding used as a dampener – it absorbs a lot of the impact and compresses significantly, creating a sticky, mushy feel. My solution for my own controller and my friends' cabs is to buy official Wahlap buttons, remove the EVA foam pad, and replace it with a rubber pad (10mm wide, 1mm thick). This swaps out the EVA foam for rubber. However, rubber does get slightly sticky on the surface – especially in hot weather when it feels like it's "sweating" – so I recommend covering the rubber surface with a layer of tape, like PET tape.

Of course, an even more direct solution is to just buy SEGA buttons outright. You can get them through various mysterious channels or directly from SEGA Japan (though they're pretty hard to get in mainland China).

(Wahlap official parts are pricey – spring leaves are 30 CNY each, optical sensors are 30 CNY each, button casings are 80 CNY each, decorative rings are 80 CNY each, and the mounting frame and base plate are also 80 CNY each. The touch sensor glass panel? A whopping 7,500 CNY.)

2.2 Touch Panel

A very common issue on older cabinets is the touch panel's metal frame rusting. After just one play session, white gloves turn yellow. Rather than spending a fortune on a replacement metal frame, a more common fix is to apply anti‑rust tape along the edges of the frame.

Here's a solution shared by 菜子头Nanoko – check it out.

The solution I use is a bit more hardcore. A while back, I was asked by an arcade in a certain region to service their Maimai and Chunithm machines, and I used this approach: use an angle grinder to remove the rust, then repaint with anti‑rust metal paint. The overall result was pretty good – though unfortunately, I didn't take any photos at the time so I can't show you. One important note: there are ABS plastic spacers on the inside that isolate the metal frame from the touch glass. Don't over‑grind those areas, or the spacers won't sit flush and the glass won't be properly secured.

Build Your Own QQ Chatbot with Napcat & AstrBot!

Wed, 08 Jul 2026 00:00:00 GMT

WARNING!!!

Using a personal QQ account to run a bot is technically against Tencent's Terms of Service and may result in your account being flagged or restricted. Do not use this for any illegal purposes! If your account gets flagged, we take no responsibility!

Introduction

AI large language models have been evolving rapidly, and QQ bots are becoming increasingly common. I've wanted to build my own QQ bot for a while, but most solutions are quite involved – setting up environments, configuring protocols, and often spending a lot of time without even getting it to run properly.

Then I came across the AstrBot + NapCat combination – it's surprisingly easy to set up. AstrBot handles the "brain" (integrating AI models and managing various features), while NapCat handles sending and receiving QQ messages. The two communicate via the OneBot protocol, and the entire setup can be deployed with a single Docker command.

What are AstrBot and NapCat?

In simple terms, AstrBot is a chatbot framework that supports various AI large language models (DeepSeek, OpenAI, Gemini, etc.), complete with a plugin system and a web management panel – it's very feature-rich.

NapCat is a protocol adapter for QQ. It simulates QQ client login and message handling, acting as a bridge that connects QQ to AstrBot.

The flow looks something like this:

You send a QQ message → NapCat receives it → AstrBot processes it → Calls the AI model → AstrBot generates a reply → NapCat sends it back to QQ

Prerequisites

Before we start, you'll need the following:

A Linux server (at least 2 vCPU and 2 GB RAM recommended, Ubuntu 20.04+ or CentOS 7+)
A QQ account (to be used as the bot – an older, regularly active account is preferred, as new accounts are more likely to get flagged)
An LLM API Key (for DeepSeek, you can sign up at platform.deepseek.com – they give free credits upon registration)
Docker and Docker Compose installed

Deployment

1. Create the Project Directory

SSH into your server and create a directory for the configuration files:

mkdir -p /opt/astrbot && cd /opt/astrbot

2. Write docker-compose.yml

Create a docker-compose.yml file in this directory with the following content:

version: "3.8"

services:
  astrbot:
    image: soulter/astrbot:latest
    container_name: astrbot
    ports:
      - "6185:6185"
    volumes:
      - ./data:/AstrBot/data
      - ./astra_db:/AstrBot/astra_db
    restart: always
    networks:
      - astrbot-net

  napcat:
    image: mlikiowa/napcat-docker:latest
    container_name: napcat
    ports:
      - "6099:6099"
      - "6195:6195"
      - "6199:6199"
    environment:
      - NAPCAT_UID=0
      - NAPCAT_GID=0
      - WS_ENABLE=true
      - WS_URL=ws://astrbot:6199/ws
    volumes:
      - ./napcat/config:/usr/src/app/napcat/config
      - ./napcat/QQ:/root/.config/QQ
    restart: always
    networks:
      - astrbot-net

networks:
  astrbot-net:
    driver: bridge

A few important notes about this configuration:

WS_ENABLE=true enables NapCat's WebSocket client mode
WS_URL=ws://astrbot:6199/ws is the key – it tells NapCat to connect to AstrBot's WebSocket service. Make sure to include /ws at the end, otherwise the connection will fail
Both containers communicate via the custom astrbot-net network, so you don't need to worry about IP addresses

If pulling Docker images is particularly slow for you, you can add a domestic mirror prefix to the image name, such as m.daocloud.io/docker.io/soulter/astrbot:latest.

3. Start the Containers

Run the following command in the directory containing docker-compose.yml:

docker compose up -d

Wait about a minute, then check the status with docker compose ps. Both containers should show Up – that means they're running successfully.

Configuring NapCat (Logging into QQ)

1. Get the WebUI Token

Run the following command to view NapCat's startup logs:

docker logs napcat

Look for a line containing something like NapCat WebUI Login Token: xxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx and note down that Token.

2. Log into the WebUI

In your browser, visit http://your_server_ip:6099 and enter the Token you just retrieved to log in.

Then follow the NapCat WebUI prompts to scan the QR code with your phone's QQ app to log in to your bot's QQ account.

Configuring AstrBot

1. Access the Admin Dashboard

Open your browser and go to http://your_server_ip:6185. The default username and password are both astrbot. You'll be prompted to change the password on first login – after that, log back in.

2. Connect an AI Model

In the left sidebar, navigate to "Model Providers" → "Chat", then click "Add Provider":

Provider: Select DeepSeek/OpenAI API Compatible
API Address: https://api.deepseek.com/v1
API Key: Paste the Key you obtained from DeepSeek
Model: Enter deepseek-chat

3. Create a Bot Instance

In the left sidebar, go to "Bots" → "Create Bot":

Platform Type: Select QQ Personal Account (OneBot v11)
WebSocket Address: 0.0.0.0:6199
Reverse WebSocket Token: Leave blank
Enabled: Toggle on

After saving, restart AstrBot and watch the logs. If you see a message like [Info] Connected to OneBot v11 client, that means AstrBot and NapCat have successfully established a connection.

Verify It Works

Once all steps above are completed, send a message to your bot from another QQ account to test it. Under normal circumstances, the bot will call the DeepSeek API to generate a response.

You can also customise the bot's response style by configuring the System Prompt (persona) in the AstrBot admin dashboard.

Common Issues

1. Docker image pull is too slow?

Use a domestic mirror by prefixing the image name with m.daocloud.io/docker.io/, for example:

image: m.daocloud.io/docker.io/soulter/astrbot:latest

2. WebSocket connection fails?

Check these points:

Does WS_URL end with /ws?
Are both containers on the same Docker network?
Is port 6199 open in your cloud server's security group?

3. QQ account got flagged/restricted?

Tencent's anti‑bot mechanisms are beyond our control. We recommend using an older, regularly active account, avoid frequently restarting containers (which causes repeated logins), and refrain from sending sensitive content.

4. How to update?

In the project directory, run:

docker compose pull
docker compose down
docker compose up -d

5. How to back up?

Simply archive the ./data, ./napcat, and ./astra_db directories. To restore, extract the backup files back into place and restart the containers.

Summary

With AstrBot + NapCat + Docker, you can go from zero to a fully functional QQ bot in about 10 minutes. And AstrBot offers much more than what's covered here – it also has a plugin system, MCP service extensions, multi‑platform support, and more. Feel free to explore on your own!

Using UnityExplorer to Retrieve Three Encrypted Values from a Certain 8‑Button Rhythm Game's Chinese Server

Mon, 06 Jul 2026 00:00:00 GMT

WARNING!!!

This article does not include any software programs. It is for reference and educational purposes only. Do not use for illegal activities!

Due to its nature, this content may carry copyright/DMCA risks. If you believe this infringes upon the copyright of any software, please contact me and I will remove it.

Main Content

1. Download and Install UnityExplorer

https://github.com/sinai-dev/UnityExplorer/releases

We need the UnityExplorer.MelonLoader.Mono.zip version.

After downloading, you'll get a zip file. Extract it and you'll see two folders: Mods and UserLibs. Copy these into the game's installation directory.

Once copied, launch the game.

2. Retrieve AesKey, AesIV, and ObfuscateParam

After launching the game, press F7 in‑game to open UnityExplorer. Click the leftmost tab in the top menu – Object Explorer.

Select Object Search, then under Searching for, choose Class.

Next, enter Net.CipherAES in the Class filter field, then click the Search button.

In the search results, click on Net.CipherAES. In the Inspector window that appears, look for CipherAES.AesKey and CipherAES.AesIV – these are the AesKey and AesIV values we need.

The same process applies to ObfuscateParam. Simply change the Class filter to Net.Packet.Packet and click Search.

In the Inspector window, Packet.ObfuscateParam is the value we're after.

How to Clean a Maimai DX Arcade Cabinet

Thu, 02 Jul 2026 00:00:00 GMT

Introduction

The most important thing before cleaning a cabinet is to get permission from the arcade. Never disassemble the machine without authorization! We also do not encourage non‑professionals to perform this themselves.

If possible, having two or more people is highly recommended – the components are quite heavy, and the disassembly process can be tricky. Two people make things much easier.

This article focuses mainly on how to remove the touch panel assembly for screen cleaning. Button maintenance will not be covered here.

If you prefer a video guide, we recommend this one ↓

Some photos in this article reference that video since I didn't capture all angles myself.

Preparation

Keys – for the front service panel and rear inspection hatch. Ask the arcade tech or staff.
Screwdrivers – a Phillips screwdriver and a hex key set. Electric ones are better if available. It's also helpful to have a shorter screwdriver for tight spaces.
Wrench – if the cabinet is against the wall and needs to be moved out, you'll need a wrench to loosen the leveling feet. An adjustable wrench is best.
Small brush – for dusting. Arcades usually have vacuum cleaners on hand, which are very convenient.
Two cloths – one dry, one damp.
Power off – It's best to shut down the machine before disassembly. Use the key to open the central service panel, where you'll find the power switch, or locate the switch at the bottom behind the 1P side.

Disassembly Steps

1. Remove the Light Box (Top Marquee)

First, remove the front light box. This is relatively simple – a total of seven hex screws per side (7 on each P side), as shown below:

Important: Keep the removed screws in a container so they don't get lost.

Once the screws are out, you can lift the entire light box off. It's a bit heavy – two people are recommended. If you need to set it on the floor, place something soft underneath to avoid scratches.

After removal, use a vacuum to suck out the accumulated dust – this area usually holds the most. Then wipe it clean with a damp cloth.

For the transparent acrylic part of the top frame, if available, use some acrylic cleaner. If not, a regular cleaner with a cloth will work fine.

2. Remove the Touch Panel Assembly

This step requires you to get behind the machine, so if it's up against a wall, you'll need to pull it out. If you can't move it easily, use a wrench to loosen the leveling feet. Since it's so heavy, I had to ask the arcade tech for help on this one 😅

Once the machine is pulled out, open the rear upper inspection hatch – the 1P side uses hex screws, while the 2P side requires the key.

Inside, you'll need to unplug five cables. When unplugging, gently wiggle them side‑to‑side. Always hold the connector, not the wires – pulling by the wires can damage them. See image below:

After all cables are disconnected, remove the six screws – these are Phillips, so switch to your cross‑head screwdriver.

(Since I didn't have a complete photo, here's a diagram from the manual instead~)

With all screws removed, go to the front of the cabinet. Lift the entire touch panel assembly upward, then pull it backward to detach it. It's quite heavy, so two people are strongly recommended.

Once removed, you can clean the inside of the touch panel assembly. We recommend gently wiping with a dry cloth – do not use alcohol or any corrosive liquids, and avoid using a damp cloth.

You can also clean the display panel underneath, as well as any other dusty areas inside the cabinet – a vacuum and a cloth work well here.

3. Reassembly

Reassembly is simply the reverse of disassembly. Keep these points in mind:

Make sure the touch panel assembly is fully seated against the frame and securely fastened.
Prefer manual screwdrivers when re‑installing. If using an electric screwdriver, set the torque to a low setting – high torque can crack the wooden frame or acrylic panels.
It's recommended to test the machine before re‑installing the front light box – ensure everything works, then put it back.

4. Testing

Once powered on and in game, open the central service panel and press the "test" button.

Button 1 = Up
Button 4 = Down
Button 5 = Confirm

Navigate to Touch Sensor Test and press Button 5. Use Button 1 or 4 to select either 1P or 2P, then press Button 5 to confirm.

Wait for the self‑test to complete, then test the touch sensors. If the arcade has the official touch test hammer, it's recommended for more accurate results. If not, wearing gloves also works.

Of course, you can also just play a round to test things out!

Summary

Cleaning isn't actually that difficult – but the end result is a fresh, clean machine that feels great to play on. And just like that, you've got yourself a sparkling new Dilabra (Maimai DX)!

</div>

Building a Simple API Proxy with Cloudflare Workers

Thu, 25 Jun 2026 00:00:00 GMT

Introduction

A couple of days ago, while integrating GitHub OAuth login into my website, I ran into a problem – mainland China servers can't directly access the GitHub API. I'm sure many developers have encountered this issue, as GitHub's API is often unstable or times out when accessed from within China.

There are several solutions: buying an overseas server to set up a proxy, using a VPN, using a third‑party proxy service, etc. Considering cost and stability, I ultimately chose to use Cloudflare Workers to build a simple API proxy.

Why Cloudflare Workers?

Generous free tier – 100,000 free requests per month, which is more than enough for personal projects
Global CDN acceleration – Cloudflare has nodes worldwide, providing fast access
Simple deployment – Just a few lines of code, deployable directly from the Dashboard
No server management – Zero infrastructure to maintain

As for the other options, here's why I didn't choose them:

Buying an overseas server – Too expensive; I wanted a near‑zero‑cost solution
Using a VPN – Essentially circumventing the firewall, which is illegal; using a proxy avoids this issue
Third‑party proxy services – Stability and security are questionable compared to a self‑hosted solution

Implementation

The Cloudflare Workers code is very simple, written in JavaScript/TypeScript. The core idea is to forward requests to the GitHub API and return the response to the client.

export default {
  async fetch(request, env, ctx) {
    // Handle OPTIONS preflight requests
    if (request.method === "OPTIONS") {
      return new Response(null, {
        headers: {
          "Access-Control-Allow-Origin": "*",
          "Access-Control-Allow-Methods": "GET,POST,OPTIONS",
          "Access-Control-Allow-Headers": "Content-Type"
        }
      });
    }

    const reqUrl = new URL(request.url);
    let targetUrl;
    
    // /login path proxies to github.com, all others to api.github.com
    if (reqUrl.pathname.startsWith("/login")) {
      targetUrl = `https://github.com${reqUrl.pathname}${reqUrl.search}`;
    } else {
      targetUrl = `https://api.github.com${reqUrl.pathname}${reqUrl.search}`;
    }

    // Forward the request
    const forwardReq = new Request(targetUrl, {
      method: request.method,
      headers: request.headers,
      body: request.body
    });

    const resp = await fetch(forwardReq);
    const newResp = new Response(resp.body, resp);
    newResp.headers.set("Access-Control-Allow-Origin", "*");
    
    return newResp;
  }
};

Deployment Steps

Sign up for Cloudflare – Visit Cloudflare and create a free account
Create a Worker – In the Dashboard, expand "Compute" in the left navigation, click "Workers & Pages", then click "Create Application" in the top right. Choose "Start from Hello World", give it a name (e.g., "github-api-proxy"), and click "Deploy"
Paste the code – Navigate to your project page, click "Edit Code" in the top right, and paste the code above into the editor
Set up a route (optional) – If you have your own domain, you can set up custom routes in Workers & Pages → Routes. I strongly recommend binding your own domain
Deploy – Click "Deploy" and your proxy is live

After deployment, you'll have your own GitHub API proxy, accessible at something like: https://your-worker.your-subdomain.workers.dev/

Usage Example

Where you previously called the GitHub API directly:

https://api.github.com/user

Now change it to:

https://your-worker.your-subdomain.workers.dev/user

Important Notes

Rate limiting – The GitHub API has rate limits. Since your proxy still uses a single IP to make requests, be mindful of request frequency
HTTPS – Cloudflare Workers provide HTTPS automatically – no extra configuration needed
Security – The sample code above only ensures the service runs. In production, you should modify the code to add authentication, rate limiting, etc. – all easily achievable with a few code changes

WARNING

Using Cloudflare Workers to proxy GitHub may violate Cloudflare's Terms of Service and could result in your account being banned. Please take necessary precautions – limit request frequency, add authentication, etc. Most account bans are caused by excessive request rates or the proxy endpoint being leaked without authentication, leading to abuse by malicious parties. We recommend using this solution only in test environments. For production use, proceed with caution and ensure proper security measures are in place. If possible, self‑host your own proxy server. This article is for learning and reference purposes only – we take no responsibility for any account bans or other issues arising from use of this solution.

The Complete Guide to FRP (Fast Reverse Proxy) – Internal Network Penetration

Sat, 06 Jun 2026 00:00:00 GMT

Introduction

Have you ever run into any of these frustrating situations?

You set up a NAS or home media server at home, but can't access it from outside
You developed a web app that only works on your local network, and you can't share it with friends over the internet
You have a Raspberry Pi or an internal server, but you can't remotely control it because you don't have a public IP

All of these problems can be solved with FRP (Fast Reverse Proxy). This article will walk you through the principles, use cases, and a complete step‑by‑step setup guide for FRP.

1. What is FRP and What Can It Do?

1.1 What is FRP?

FRP (Fast Reverse Proxy) is a high‑performance reverse proxy application focused on internal network penetration. It is written in Go and supports TCP, UDP, HTTP, HTTPS, and many other protocols. It can expose internal services to the public internet securely and conveniently via a node with a public IP address.

Official GitHub repository: https://github.com/fatedier/frp

1.2 What Can FRP Do?

FRP has a wide range of use cases:

Scenario	Description
Remote access to home NAS	Access your home NAS from outside to view files or stream media
Remote development & debugging	Expose your local development service to the internet for third‑party testing
Build a private VPN	Combine FRP with SSH for secure remote access
Access LAN devices	Such as IP cameras, printers, smart home devices, etc.
WeChat/Alipay webhook development	Receive external callbacks to your local service
Game servers / multiplayer hosting	Such as hosting a Minecraft server for friends

2. Why Do We Need FRP?

2.1 The Problem with Internal Network Devices

For hosts on the internet to communicate with each other, they need to know each other's public IP addresses. But in reality:

IPv4 address exhaustion – Global IPv4 addresses have long run out, making it difficult for regular users to obtain a public IP
NAT (Network Address Translation) – Most users are behind a network topology like this:

User Device → Home Router → ISP NAT → Internet

Your home router assigns a private IP (192.168.x.x, 172.16.x.x, or 10.x.x.x), and your ISP may further use CGNAT (Carrier‑Grade NAT), sharing a small pool of public IPs among many users.

The result is: the outside world cannot initiate a connection to any of your devices.

2.2 Comparison of Common Solutions

Solution	Pros	Cons
Commercial tools (like Ngrok, Hamachi)	Works out of the box	Free tiers are rate‑limited/throttled, contain ads, require payment
IPv6	Direct addressing	Some ISPs restrict it, older devices don't support it
Apply for a public IP	Truly public access	Some ISPs (e.g., China Telecom) may offer it; others (China Mobile/Unicom) generally don't
FRP	Fully self‑controlled, supports multiple protocols	Requires a cloud server (VPS)

2.3 Advantages of FRP

Compared to other solutions, FRP offers:

Full control – Only a cheap cloud server (VPS) is needed, costing around $3‑5/month
Multi‑protocol support – TCP, UDP, HTTP, HTTPS, etc.
High performance – Uses TCP connection multiplexing to save resources
Security – Supports TLS encryption and authentication
Cross‑platform – Works on Linux, Windows, macOS, Raspberry Pi, and more

3. How FRP Works – Detailed Explanation

3.1 Core Components

FRP uses a C/S (Client/Server) architecture with two core programs:

Component	Runs On	Function
frps (Server)	Public server (VPS)	Receives external requests and forwards them to the internal client
frpc (Client)	Internal device (PC/NAS/RPi)	Actively connects to frps and registers local services

3.2 Workflow

FRP works as follows:

┌─────────────────────────────────────────────────────────────┐
│                    Public Server (frps)                     │
│  ┌─────────┐                                                │
│  │  frps   │◄────── Control connection (TCP long‑live) ────► frpc │
│  │         │                                                │
│  │   Port  │◄──────── Data forwarding ────────────────── Internal Service │
│  │   Mapped │                                                │
│  └─────────┘                                                │
└─────────────────────────────────────────────────────────────┘

Outbound connection from internal network: When frpc starts, it actively connects to frps's listening port (default 7000) and establishes a persistent TCP long connection
Service registration: frpc tells frps: "I can proxy the local HTTP service at 127.0.0.1:8080"
External request arrives: A user visits http://public_ip:port or http://domain, and DNS resolves to the frps server
Smart routing: frps finds the corresponding frpc based on the port number or HTTP Host header
Data forwarding: frps notifies frpc via the existing connection, the two establish a data channel, and all subsequent traffic is forwarded bidirectionally

3.3 Supported Proxy Types

FRP supports multiple proxy types for different use cases:

Proxy Type	Use Case	Features
TCP	General port mapping	SSH, RDP, game servers, etc.
UDP	UDP port mapping	DNS, video streaming, etc.
HTTP/HTTPS	Web services	Supports virtual hosting, automatic Host header handling
STCP/SUDP	Secure internal proxies	No public port exposure; requires frpc on both ends
XTCP	P2P direct connection	After hole punching, traffic bypasses the server to save bandwidth

4. FRP Setup – Step by Step

4.1 Prerequisites

Before you start, you'll need:

A Linux public server (VPS) – Ubuntu or CentOS recommended, minimum 1 vCPU 1 GB RAM
An internal device (Windows/Linux/macOS/Raspberry Pi)
A domain name (optional, but required for HTTP/HTTPS mode)

4.2 Download FRP

Download the latest binary from the GitHub Releases page:

# Download (choose the version for your system architecture)
wget https://github.com/fatedier/frp/releases/download/v0.62.0/frp_0.62.0_linux_amd64.tar.gz

# Extract
tar -xzf frp_0.62.0_linux_amd64.tar.gz
cd frp_0.62.0_linux_amd64

After extraction, you'll see the following files:

.
├── frpc          # Client binary
├── frpc.toml     # Client configuration
├── frps          # Server binary
├── frps.toml     # Server configuration
└── LICENSE

The server and client use the same binary – just different config files. Let's separate them:

# Create directories for server and client
mkdir -p /opt/frp/server
mkdir -p /opt/frp/client

# Server gets frps and frps.toml
mv frps frps.toml /opt/frp/server/

# Client gets frpc and frpc.toml
mv frpc frpc.toml /opt/frp/client/

4.3 Configure the Server (frps)

Edit the server config file /opt/frp/server/frps.toml:

[common]
# Server listening port – clients connect here
bindPort = 7000

# HTTP entry port (if using HTTP proxies)
vhostHttpPort = 8080

# HTTPS entry port (if using HTTPS proxies)
# vhostHttpsPort = 8443

# Authentication
auth.method = "token"
auth.token = "your_strong_token_here"  # Use a strong password

# Dashboard config (web admin interface)
webServer.addr = "0.0.0.0"
webServer.port = 7500
webServer.user = "admin"
webServer.password = "your_dashboard_password"

# Logging
log.to = "/var/log/frp/frps.log"
log.level = "info"
log.maxDays = 7

# Enable Prometheus metrics (optional)
enablePrometheus = true

Note: Remember to open ports 7000, 7500, 8080, etc., in your cloud server's security group.

Start the server:

# Run in foreground (for testing)
./frps -c ./frps.toml

# Run in background
nohup ./frps -c ./frps.toml &

Using systemd (recommended):

# /etc/systemd/system/frps.service
[Unit]
Description = frp server
After = network.target

[Service]
Type = simple
ExecStart = /opt/frp/server/frps -c /opt/frp/server/frps.toml
Restart = always
RestartSec = 5

[Install]
WantedBy = multi-user.target

sudo systemctl daemon-reload
sudo systemctl enable frps
sudo systemctl start frps
sudo systemctl status frps

Visit http://your_server_ip:7500, enter the Dashboard username and password, and you'll see the admin interface.

4.4 Configure the Client (frpc)

Edit the client config file /opt/frp/client/frpc.toml:

[common]
# Server address and port
serverAddr = "your_server_ip"
serverPort = 7000

# Authentication – must match the server
auth.method = "token"
auth.token = "your_strong_token_here"

# Logging
log.to = "./frpc.log"
log.level = "info"
log.maxDays = 7

# ==== TCP Proxy Example ====
# Map local port 22 to server port 6000
[[proxies]]
name = "ssh"
type = "tcp"
localIP = "127.0.0.1"
localPort = 22
remotePort = 6000

# ==== HTTP Proxy Example ====
# Map local web service at port 8080 to the server
[[proxies]]
name = "web"
type = "http"
localIP = "127.0.0.1"
localPort = 8080
customDomains = ["your-domain.com"]

Start the client:

# Run in foreground
./frpc -c ./frpc.toml

# Run in background
nohup ./frpc -c ./frpc.toml &

4.5 Advanced: Access via Domain Name

If you have a domain, you can configure HTTP/HTTPS proxies to access internal services via the domain.

Additional server config:

# In frps.toml, make sure this is set
vhostHttpPort = 8080

Client config:

[[proxies]]
name = "web"
type = "http"
localIP = "127.0.0.1"
localPort = 8080
customDomains = ["web.your-domain.com"]

Then add a DNS record at your domain registrar:

web.your-domain.com  →  your_server_ip

Now visiting http://web.your-domain.com:8080 will automatically forward to your internal service.

4.6 Advanced: Load Balancing

FRP supports load balancing across proxy groups, which is great for multi‑instance deployments:

[[proxy groups]]
name = "web_group"
type = "load-balance"
selector = "round-robin"
servers = [
    { name = "web1", localIP = "192.168.1.100", localPort = 8080 },
    { name = "web2", localIP = "192.168.1.101", localPort = 8080 },
]

5. Security Hardening Recommendations

FRP is powerful, but misconfiguration can introduce security risks. Here are some hardening tips:

5.1 Mandatory Security Measures

Use a strong token – Don't use simple passwords; generate a random, strong one
Restrict Dashboard access – Use firewall rules to allow only specific IPs to access port 7500
Enable TLS encryption – Add transport.tls.tlsCertFile and transport.tls.tlsKeyFile in the [common] section
Avoid running as root – Create a dedicated frp user

5.2 TLS Encryption Example

# Server frps.toml
[common]
transport.tls.tlsCertFile = "/path/to/cert.pem"
transport.tls.tlsKeyFile = "/path/to/key.pem"
transport.tls.force = true

# Client frpc.toml
[common]
transport.tls.tlsCertFile = "/path/to/cert.pem"
transport.tls.tlsKeyFile = "/path/to/key.pem"
transport.tls.force = true

6. Common Issues

Q1: Connection fails with "login to server failed"

Check that the server is running properly
Verify that serverAddr, serverPort, and auth.token match the server config
Check firewall and security group rules to ensure the ports are open

Q2: HTTP service is not accessible

Confirm that vhostHttpPort is set correctly on the server
Check that the customDomains in the client config resolves to the server IP
Confirm DNS records have propagated

Q3: How do I check connection status?

Visit the Dashboard (default http://server_ip:7500)
Check the log files
Use sudo systemctl status frps to see the service status

7. Summary

FRP is a powerful and flexible internal network penetration tool. Through this article, you should now understand:

The basic concepts and use cases of FRP
How FRP works under the hood
Complete configuration for both the server and client
Common security hardening measures

Compared to commercial internal‑penetration tools, FRP offers greater control and flexibility – and it's completely open‑source and free. We hope this guide helps you better understand and use FRP.

References:

FRP Official GitHub: https://github.com/fatedier/frp
FRP Official Documentation: https://gofrp.org/en/docs/

The Complete New Laptop Inspection Guide – From Unboxing to Stress Testing

Sat, 11 Apr 2026 00:00:00 GMT

Introduction

After purchasing a new laptop, a proper inspection process is critical. It not only ensures that you've received a brand‑new, genuine product, but also helps you protect your rights if any issues are found. This guide starts from the unboxing and walks you through every step of the inspection, helping you thoroughly check your new machine.

Step 1: Unboxing – Record Everything

Preparation

Before unsealing, prepare the following:

A phone or camera (make sure you have enough storage space)
A well‑lit, stable filming environment
A tripod or fixed mount (optional, but recommended)

Key Filming Points

Show the outer shipping box
- Capture the tracking number clearly on camera
- Record the condition of the shipping box – any damage, crushing, or compression marks
- Film all angles of the package
The unsealing process
- Record continuously from the moment you open the outer box
- Film the product box's seals and stickers – check if they are intact
- Note the location of the serial number label for later verification
Unboxing details
- Take out each accessory one by one and show them on camera
- Film the laptop, power adapter, manuals, warranty card, etc.
- Verify that the accessory list matches the official specifications
- Pay special attention to the screen protective film and keyboard spacer paper: New laptops usually have a paper spacer between the screen and keyboard to prevent the keyboard from leaving marks on the screen during shipping. This paper should be flat and undamaged – any missing or damaged spacer is a red flag

Important: Keep the recording uninterrupted from start to finish. If any problems are found, this video will be critical evidence for after‑sales support.

Step 2: Physical Inspection

External Chassis Inspection

Check the casing
- Carefully inspect the A‑cover (lid), C‑cover (keyboard deck), and D‑cover (bottom panel) for scratches, dents, or scuffs
- Check corners for wear or deformation
- Look for uneven coating, bubbles, or peeling
Screen inspection
- Check the screen surface for scratches, fingerprints, or dust
- Check if the screen bezel is evenly aligned and if there's any separation
- Test the hinge – it should open and close smoothly without any unusual noises
- Check the screen protective film: New screens usually have a film with branding or warning labels – confirm it's intact and free of bubbles
- Check for keyboard imprints: Look closely at the screen surface for any marks or imprints left by the keyboard keys – this is a key indicator of a used or refurbished unit
Port inspection
- Check all USB, HDMI, Type‑C, and other ports for signs of use
- Look inside the ports – they should be clean, with no oxidation or debris
- Confirm that the ports are properly aligned and not loose
Bottom panel inspection
- Check the bottom screws for any signs of tampering (paint chipping, scratches)
- Verify that the rubber feet are intact and complete
- Cross‑check the model and serial number on the bottom label against the box

Keyboard and Touchpad Inspection

Keyboard
- Check that all keycaps are level – no crooked or missing keys
- Press every key – feel for normal tactile feedback, no sticking or unresponsive keys
- Test the backlight (if available) – it should be uniform, with no dead LEDs or uneven brightness
- Pay special attention to common keys: WASD, Space, Enter, etc. – check for visible wear or shine
- Check keyboard cleanliness: There should be no dust, hair, or debris between the keycaps
Touchpad
- The surface should be smooth – no scratches or indentations
- Test multi‑touch gestures – two‑finger scrolling, pinch‑to‑zoom, etc. should be responsive
- Click feel should be crisp – left and right buttons should work properly
- Test edge gestures (e.g., three‑finger swipe to switch tasks)

Step 3: Critical Pre‑Power‑On Checks

Power Adapter Check (Extremely Important)

This is one of the most crucial ways to determine if the laptop has ever been activated:

Try to power on without plugging in the adapter
- New laptops ship with the battery in a protective state – charge is extremely low or completely drained
- A brand‑new laptop should NOT power on without the AC adapter connected
- If it turns on without being plugged in, the laptop has been previously activated and used
- This is a strong indication of a display unit, a return, or a used machine – contact the seller immediately
Connect the power adapter
- Plug in the adapter first, then press the power button
- Check that the charging indicator lights up
- Some models may take a while to charge enough to boot for the first time – be patient
Inspect the power adapter
- Check for wear, scratches, or scuffs on the adapter body
- Inspect the power cord for kinks or signs of use
- Confirm that the model and wattage match the official specifications
- Check the plug for insertion marks (the metal contacts should be bright and free of oxidation)

Before Powering On

Do NOT connect to the internet! This is the most critical step. Windows will automatically activate once it goes online, and activation will affect your return/exchange rights.

What to expect on first boot:

After connecting power and pressing the power button
The screen lights up and the brand logo appears
The Windows initial setup screen (OOBE – Out‑of‑Box Experience) appears
If the system boots directly to the desktop or shows an existing user account, the device has been used before

How to Skip Network Activation

Windows 11

Method 1: Use a command to bypass
- When you reach the "Let's connect you to a network" screen, press Shift + F10 (on some models, Shift + Fn + F10)
- In the command prompt that appears, type: OOBE\BYPASSNRO
- The system will restart and return to the setup screen
- You will now see the option "I don't have internet" – select it
- Click "Continue with limited setup" to complete the offline setup
Method 2: End the process via Task Manager
- At the network connection screen, press Shift + F10 to open Command Prompt
- Type taskmgr to open Task Manager
- Find the "Network Connection Flow" process (or similar)
- Right‑click and select "End Task" – this will skip the network setup

Windows 10

Windows 10 usually offers the option "I don't have internet" directly. If not, the Windows 11 method (using the command) also works.

Initial System Checks

System information verification
- Right‑click "This PC" → "Properties" – check processor, RAM, and system type
- Press Win + R, type msinfo32 – check detailed system information
- Confirm that the CPU model and RAM size match your purchase configuration
Disk check
- Right‑click the Start menu → "Disk Management"
- Check the partition layout and confirm that the total capacity matches the advertised size
- Note: The advertised capacity and the actual usable capacity will differ (due to different calculation methods) – this is normal
Battery information
- Press Win + X and select "Windows PowerShell (Admin)" or "Terminal (Admin)"
- Enter the command: powercfg /batteryreport
- The system will generate a battery report, usually saved as C:\Users\[username]\battery‑report.html
- Open the report and check the following key metrics:
  - "Design Capacity" – the battery's rated capacity
  - "Full Charge Capacity" – the current maximum charge the battery can hold
  - "Cycle Count" – the number of charge/discharge cycles
- What to expect for a new laptop:
  - Design Capacity and Full Charge Capacity should be very close (within 5% difference)
  - Cycle Count should be 0 or a single digit (1‑5 cycles is within normal factory testing range)
  - If Cycle Count exceeds 10, or there is a significant capacity difference, it may be a display unit or a return
BIOS/UEFI Check
- Restart the laptop and press the designated key to enter BIOS (common keys: F2, F10, Del, Esc – varies by brand)
- Check the system information in BIOS:
  - Serial Number – should match the label on the chassis and the box
  - BIOS Version – confirm it's the latest version
  - Hardware Configuration – CPU, RAM, HDD/SSD info should match your configuration
- Check the system time – it may be inaccurate on a new unit (this is normal)
- Important: Do not change any BIOS settings – just review and exit

Step 4: Hardware Testing

Screen Testing

Dead/bright pixel test
- Visit online screen test websites (such as testufo.com, eizo.cn/monitor‑test, or screen.burnin.test)
- Cycle through solid‑color screens: white, black, red, green, blue, grey, etc.
- Carefully look for bright dots, dark dots, or colour anomalies
- Test method:
  - White screen: check for dark spots and colour deviations
  - Black screen: check for bright spots and light bleed
  - Colour screens: check for colour anomalies
- Industry standard: Grade‑A panels allow up to 3 defective pixels, but it's reasonable to expect a perfect screen
- Note: Different brands have different dead‑pixel policies – understand them before purchasing
Backlight bleed test
- Display a solid black screen in a dark room
- Inspect the edges of the screen from different angles for obvious light bleeding
- Some light bleed is normal on IPS panels, but it should not be excessive
Colour and brightness
- Check colour gradients – they should be smooth with no banding
- Adjust brightness – the backlight should respond smoothly with no flickering across the range
- Check for uneven colour temperature (yellow or blue tinting on one side)
- Viewing angle test: Tilt the screen up/down/left/right to check colour shift
- Refresh rate check: Right‑click desktop → Display settings → Advanced display settings – confirm the available refresh rate matches the advertised specification

Recommended Diagnostic Tools

CPU‑Z – Detailed processor information
GPU‑Z – Detailed graphics card information
CrystalDiskInfo – Hard drive health and power‑on time
Tuba Toolbox (图吧工具箱) – An all‑in‑one toolkit bundling many diagnostic utilities

Hard Drive Power‑On Time Check

Using CrystalDiskInfo:

Check "Power On Count" and "Power On Hours"
A brand‑new drive should have power‑on time under 10 hours (factory testing)
Power‑on count should typically be under 50
If the numbers are significantly higher, the unit may be a display model or a return

Step 5: Stress Testing (Burn‑In)

The purpose of stress testing is to verify the laptop's stability and cooling performance under high load. The recommended test duration is 30 minutes to 1 hour.

Preparation

Install necessary software
- AIDA64 – system stability test
- FurMark – GPU stress test
- HWMonitor or HWiNFO – real‑time temperature and frequency monitoring
Test environment
- Keep the room well‑ventilated
- Place the laptop on a hard, flat surface to ensure bottom air intakes are unobstructed
- Close unnecessary background applications

CPU Stress Test

Use AIDA64's FPU stress test:

Open AIDA64, click "Tools" → "System Stability Test"
Check "Stress FPU" (leave other options unchecked to test only the CPU)
Click "Start" to begin the test
Monitor the following:
- CPU Temperature: Should generally stay below 95°C. Ultra‑portables may reach 90‑100°C, while gaming laptops should be in the 85‑95°C range
- CPU Frequency: Should remain stable without significant throttling
- Power Consumption: Check if the CPU reaches its rated TDP
- System Stability: Any blue screens, freezes, reboots, or application crashes?
Test duration: At least 15‑30 minutes to allow temperatures to stabilise
Record data: Take screenshots of temperature and frequency curves for comparison

GPU Stress Test

Using FurMark:

Open FurMark, select an appropriate resolution (1080p recommended) and anti‑aliasing level
Click "GPU stress test" to begin
Monitor the following:
- GPU Temperature: Should generally stay under 85°C, and should never exceed 90°C
- FPS (Frames Per Second): Should remain stable with no wild fluctuations
- Visual Quality: Any artifacts, screen tearing, flickering, or graphical corruption?
- Power Consumption: Check if the GPU reaches its rated TGP/TDP
- System Stability: Any crashes, driver resets, or blue screens?
Test duration: 15‑30 minutes recommended
Important:
- FurMark imposes a very high load and may cause temperatures to rise quickly – this is normal
- If the temperature exceeds 90°C and continues to rise, stop the test immediately
- On laptops with a MUX switch, ensure the dedicated GPU is being used for the test

Combined Stress Test (Advanced)

Run AIDA64 FPU and FurMark simultaneously:

This pushes the cooling system to its absolute limit, simulating the worst‑case scenario
Test for 15‑30 minutes – no need to go longer
Key observations:
- CPU and GPU temperature curves
- Whether severe throttling occurs (frequency drops significantly)
- Whether total power consumption reaches the system's thermal design limit
- If the system shuts down due to overheating or blue‑screens, there is a serious cooling problem
Expected behaviour:
- Gaming laptops: CPU 85‑95°C, GPU 80‑85°C, with full performance delivery
- Ultra‑portables: CPU 90‑100°C with noticeable throttling – this is typical
- Fan speeds should reach maximum with noticeable noise, but it should be steady without any rattling

Keyboard Surface Temperature Test

During and after the stress test:

Touch the keyboard area with your hand, especially the WASD keys, arrow keys, and the centre‑upper section
Feel if any area is uncomfortable to touch (above 45°C becomes noticeably hot)
Normal: Warm to the touch (35‑42°C) – should not affect normal use
Abnormal: If any area is too hot to touch, it may indicate a cooling or heat‑pipe issue
Recommendation: Use a thermometer or infrared thermometer for more accurate temperature readings

Fan Noise Test

Pay attention to the fan sound during stress testing
Normal: A steady whooshing sound that increases with temperature
Abnormal:
- Grinding, squealing, or clicking sounds – may indicate bearing issues or debris
- Surging – the fan speed ramps up and down erratically
- Rattling accompanied by vibration – the fan may be loose or damaged
Idle noise: At low load, the fan should be nearly silent or completely off (on models with zero‑fan/stop‑start technology)

Step 6: Functional Testing

Port Testing

USB ports
- Connect a USB drive or external hard drive – test data transfer
- Test each USB port individually – all should recognise devices
- Test the speed difference between USB 2.0 and 3.0/3.1/3.2 ports
- Test hot‑plug functionality – remove and re‑insert to see if the device is re‑recognised
Audio jacks
- Plug in headphones and test left/right channel separation
- Play stereo music to confirm proper channel balance
- Test the microphone input – record and play back to check quality
- If there are multiple audio jacks, test each one
Video output
- If available (HDMI, DisplayPort, Type‑C with video out), connect to an external monitor or projector
- Confirm external display works and resolution/refresh rates are adjustable
- Test extended and mirrored display modes
- If a Thunderbolt port is available, test both data transfer and video output
Card reader (if available)
- Insert an SD or TF card and test reading functionality
- Copy files to check read/write speeds
- Confirm supported card types (SD/SDHC/SDXC, etc.)
Ethernet port (if available)
- Connect a network cable and test wired connectivity
- Confirm the link indicator lights are blinking correctly
- Test network speed and stability

Wireless Network Testing

Wi‑Fi test
- Connect to Wi‑Fi and test network speed (use speedtest.net)
- Test signal strength – move to different locations and check signal variation
- Test both 2.4GHz and 5GHz bands (if supported)
- Test Wi‑Fi 6/6E/7 features (if your router supports them)
- Test network stability – download a large file to check for disconnections
Bluetooth test
- Pair Bluetooth devices (mouse, keyboard, headphones, speakers, etc.)
- Test connection stability and effective range
- Test audio quality and latency with Bluetooth audio devices
- Test simultaneous connection of multiple devices
- Check the Bluetooth version matches the advertised spec (view in Device Manager)

Camera and Microphone

Camera
- Open the Camera app and test the front‑facing camera (some models also have rear‑facing)
- Check image clarity and colour accuracy
- Test auto‑focus (if supported)
- Check low‑light performance
- Test any privacy shutter or physical cover (if available)
- Confirm the camera resolution matches the advertised spec (check in Camera settings)
Microphone
- Use the Voice Recorder app to record audio
- Play back to check quality, volume, and clarity
- Test noise cancellation (if supported)
- If there are multiple microphones, test the array pickup quality
- Test in a video call for real‑world performance

Speaker Test

Play different music genres (pop, classical, rock, etc.)
Test left/right channel balance and stereo imaging
Check for distortion, rattling, or buzzing at maximum volume
Test bass and treble response
If there are audio enhancements (Dolby Atmos, DTS, etc.), test the difference with them on/off
Test at different volume levels to ensure no noticeable background hiss

Other Feature Testing

Fingerprint reader (if available)
- Enrol a fingerprint and test recognition speed and accuracy
- Test with wet or dirty fingers – see how it handles real‑world conditions
- Test Windows Hello login functionality
Face recognition (if available)
- Set up Windows Hello face recognition
- Test recognition under different lighting conditions
- Test with glasses, hats, etc.
Backlit keyboard
- Test brightness adjustment (if multiple levels are available)
- Check backlight uniformity – any dark spots?
- Test automatic backlight timeout
Sensor testing
- Gyroscope/Accelerometer: Rotate the screen to test auto‑rotation (on 2‑in‑1s or tablets)
- Ambient light sensor: Test automatic brightness adjustment
- Hall sensor: Close the lid to check if the system goes to sleep
Extended features
- If there's a numeric keypad, test all keys
- If there are dedicated function keys, test media controls, airplane mode, etc.
- Test fingerprint‑power‑button combo (if supported)

Step 7: Software Activation & Driver Updates

Activate Windows After Confirmation

Once all tests above are complete and no issues are found, you can connect to the internet and activate Windows:

Connect to Wi‑Fi or wired network
Windows will activate automatically, or you can manually enter your product key
Sign in with your Microsoft account to sync settings

Office Activation (if included)

Open any Office application (Word, Excel, etc.)
Sign in with your Microsoft account
Follow the prompts to activate the pre‑installed Office

Driver Updates

Official channels first
- Visit the laptop brand's official support website
- Download the latest drivers for your specific model
- Prioritise updates for: chipset, graphics, audio, and network drivers
Windows Update
- Run Windows Update and install recommended driver updates
- Note: Some drivers offered via Windows Update may not be the latest versions
Avoid third‑party driver tools
- Do not use tools like Driver Booster, Driver Easy, or similar third‑party software
- They often bundle adware or install unnecessary content

Common Issues & Important Notes

About Returns and Exchanges

7‑day no‑questions‑asked return policy
- An unactivated system can be returned under the 7‑day policy
- Once activated, returns are generally not accepted unless there is a quality defect
- Keep the packaging and all accessories
Quality defect criteria
- Screen pixel defects exceeding industry standards
- Hardware failures (unable to boot, frequent blue screens, etc.)
- Cosmetic defects (severe scratches, dents)
- Configuration does not match the advertised specifications

About Warranty

Warranty start date
- Usually based on the invoice date or the system activation date
- Keep your proof of purchase
Warranty coverage
- Understand the official warranty policy
- Note that accidental damage is typically not covered
- Some brands offer accidental damage protection (ADP) as an add‑on

Inspection Timeline

Day 1: Physical inspection, basic functional tests
Within 1 week: Stress testing, daily‑use experience
Ongoing: Battery life monitoring, long‑term stability

Conclusion

While the full inspection process may seem lengthy, it's the most effective way to protect your rights. Here are the key takeaways:

Record the entire unboxing process on video as evidence
Scrutinise the physical condition – don't overlook any detail
Complete all hardware checks before connecting to the internet
Skip network activation to preserve your return rights
Run thorough stress tests to verify cooling and stability
Test every feature to ensure full functionality
Only activate your software after confirming everything is fine

We hope this guide helps you complete your new laptop inspection smoothly and enjoy a worry‑free experience. If you discover any issues, contact your seller or brand support promptly.

Complete Guide to Building a Website on Linux Using aaPanel (BT Panel)

Sat, 11 Apr 2026 00:00:00 GMT

Introduction

:::warning Important Note The content of this article may differ from the actual situation due to aaPanel version updates. This guide is for reference and learning purposes only. For actual operations, please refer to the latest tutorials and documentation provided on the official aaPanel website. :::

For developers or webmasters who want to set up their own websites, traditional command-line operations can be complex and error-prone. aaPanel (formerly known as BT Panel) is an excellent server management tool that greatly simplifies the website deployment process. This article will walk you through installing aaPanel on a Linux server and using it to build a complete website.

Prerequisites

Server Requirements

Before you begin, you'll need the following:

A Linux server (CentOS 7.x, Ubuntu 18.04+, or Debian 9+ recommended)
Root or sudo access to the server
A stable internet connection
Recommended specs: 1 CPU core, 1 GB RAM, 20 GB+ disk space

Domain Name (Optional)

While you can access your site via IP address, a domain name provides a better user experience. You can purchase one from providers like Alibaba Cloud, Tencent Cloud, or any domain registrar.

Installing aaPanel

Connect to Your Server

First, connect to your Linux server via SSH:

ssh root@your_server_ip

Enter your password to log in.

Run the Installation Command

:::tip Note The security entry parameter in the installation command below (e.g., ed8484bec) is just an example. The system will generate a random security entry automatically during actual installation. Always use the latest official installation command from the aaPanel website. :::

Choose the appropriate installation command for your operating system:

CentOS / AlmaLinux / Rocky Linux

yum install -y wget && wget -O install.sh https://download.bt.cn/install/install_6.0.sh && sh install.sh

Ubuntu / Debian

wget -O install.sh https://download.bt.cn/install/install-ubuntu_6.0.sh && sudo bash install.sh

Other Linux Distributions

wget -O install.sh https://download.bt.cn/install/install_6.0.sh && bash install.sh

:::info Getting the Latest Command Visit the official aaPanel website homepage to get the latest installation command for your specific operating system. :::

Installation Process

After running the installation command, you'll be prompted to confirm. Type y and press Enter to proceed. The installation typically takes 5–15 minutes, depending on your server's performance and network speed.

Once completed, the terminal will display something like:

==================================================================
Congratulations! Installed successfully!
==================================================================
External Panel URL: http://your_server_ip:random_port/random_security_entry
Internal Panel URL: http://your_local_ip:random_port/random_security_entry
username: random_username
password: random_password
Warning:
If you cannot access the panel,
release the following port (panel_port|80|443|20|21) in the security group
==================================================================

:::warning Important Newer versions of aaPanel use random ports and random security entries by default, rather than the fixed 8888 port. Please carefully record the full access URL, username, and password displayed after installation. :::

Save this information somewhere safe – especially the access URL, username, and password.

Logging Into aaPanel

Open Required Ports

Before logging in for the first time, ensure your server's security group has the necessary ports open:

aaPanel Port: The random port displayed after installation (not fixed at 8888)
80: HTTP
443: HTTPS
20/21: FTP (if you need FTP access)
3306: MySQL (only if remote database access is required)

:::tip Port Notes Newer aaPanel versions use random ports for improved security. Check the installation output for your specific port number. If you're using a cloud server (e.g., Alibaba Cloud, Tencent Cloud), you'll need to open these ports in the cloud console's security group settings. :::

Access the Panel

In your browser, enter the External Panel URL displayed after installation, for example:

http://your_server_ip:random_port/random_security_entry

Enter the username and password you recorded earlier to log in.

:::warning Important If you can't access the panel, check:

Whether the server firewall has opened the panel port
Whether the cloud security group has opened the panel port
Whether the access URL is correct (including port number and security entry) :::

Initial Configuration

On first login, aaPanel will recommend installing a set of common software packages. For website hosting, we recommend the LNMP stack:

Nginx – Web server
MySQL – Database
PHP – Scripting language
phpMyAdmin – Database management tool

Choose the appropriate versions based on your needs, then click "One‑Click Install". The installation may take some time, so please be patient.

Creating a Website

Add a Site

Click "Website" in the left sidebar
Click the "Add Site" button
Fill in the required information:
- Domain: Enter your domain name (or use your IP address if you don't have a domain yet)
- Root Directory: The path for your website files – the default is fine
- Database: Choose whether to create a database and its type
- PHP Version: Select the appropriate PHP version
- Remark: Optional – for identifying the site's purpose
Click "Submit" to create the site

Upload Website Files

Method 1: Upload via File Manager

Click "Files" in the left sidebar
Navigate to the site root directory (usually /www/wwwroot/your_domain)
Click the "Upload" button and select your local website files
Wait for the upload to complete

Method 2: Upload via FTP

Click "FTP" in the left sidebar
View the FTP account details
Connect to the server using an FTP client (e.g., FileZilla)
Upload your files to the corresponding directory

Method 3: Clone via Git

If your project is hosted on GitHub or another Git platform:

Open a terminal in your site root directory
Run the git clone command:

cd /www/wwwroot/your_domain
git clone https://github.com/username/repo.git .

Configure Your Site

Basic Settings

Click "Settings" next to your site in the list
You can modify the following:
- Domain Management: Add or remove domains
- Directory Binding: Change the site root directory
- Traffic Limiting: Set bandwidth limits
- Rewrite Rules: Configure URL rewrite rules (pseudo‑static)

SSL Certificate Configuration

To enable HTTPS access for your site:

In the site settings, click "SSL"
Select "Let's Encrypt" for a free certificate
Check the domains you want to secure
Click the "Apply" button
Once issued, enable "Force HTTPS"

Domain Name Resolution

Get Your Server IP Address

You can find your public IP address on the aaPanel dashboard or in your server information.

Configure DNS Resolution

Log in to your domain registrar's control panel (e.g., Alibaba Cloud, Tencent Cloud)
Find the DNS resolution settings
Add an A record:
- Host Record: @ (for the root domain) or www
- Record Value: Your server's IP address
- TTL: Default is fine
Wait for DNS propagation (typically takes anywhere from a few minutes to a few hours)

Verify Resolution

Use the ping command to check if the domain resolves correctly:

ping your_domain.com

If it returns your server's IP address, the resolution is successful.

Security Optimisation

Change the Panel Port

To improve security, it's recommended to periodically change the aaPanel access port:

After logging in, click "Panel Settings" in the left sidebar
Find the "Panel Port" option
Change it to an uncommon high port (e.g., between 10000‑65535)
Save the settings and restart the panel service

:::warning Important After changing the port, make sure to:

Open the new port in your cloud security group
Close the old port's security group rules
Access the panel using the new port address :::

Set a Strong Password

Ensure your password is complex – include uppercase and lowercase letters, numbers, and special characters.

Enable the Firewall

Click "Security" in the left sidebar
Enable the system firewall
Only open necessary ports
Consider setting up IP whitelists to restrict access

Keep Everything Updated

Regularly update aaPanel and all installed software:

Click "Update" on the dashboard to check for the latest version
Apply security patches promptly
Update web servers, databases, and other components

Backup Strategy

Click "Scheduled Tasks" in the left sidebar
Add regular backup tasks:
- Back up website files
- Back up databases
- Set retention periods
You can choose to back up locally or to remote storage

Performance Optimisation

Nginx Optimisation

Find Nginx in the App Store
Click "Settings" to optimise:
- Enable gzip compression
- Adjust worker_processes to match your CPU cores
- Configure caching strategies

PHP Optimisation

Adjust PHP settings:
- memory_limit – increase appropriately
- max_execution_time – set a reasonable value
- opcache – enable OPcache for speed

Database Optimisation

In MySQL settings:
- Adjust innodb_buffer_pool_size
- Enable query caching
- Regularly optimise table structures

Enable CDN (Optional)

For high‑traffic websites, consider using a CDN:

Sign up for a CDN provider (e.g., Alibaba Cloud CDN, Tencent Cloud CDN)
Point your domain's CNAME to the CDN‑provided address
Configure your origin server settings in the CDN console
Set caching policies and HTTPS certificates

A CDN effectively reduces server load and improves user access speed, especially for static content delivery.

Monitoring & Maintenance

Monitor Server Status

aaPanel provides comprehensive monitoring features:

CPU usage
Memory usage
Disk space
Network traffic
System load

Regularly check these metrics to catch potential issues early.

Log Management

View access logs and error logs in your site settings
Analyse logs to identify problems
Periodically clean old logs to free up space

Common Issues & Fixes

Website Not Accessible

Check if the server is running
Confirm the security group ports are open
Verify domain resolution is correct
Check Nginx/Apache error logs

Database Connection Failed

Confirm the database service is running
Check database username and password
Verify database permissions
Check database error logs

Pages Loading Slowly

Check server resource usage
Optimise website code and assets
Enable caching mechanisms
Consider upgrading server configuration

Advanced Features

Scheduled Tasks

Use aaPanel's scheduled tasks feature to:

Periodically clean temporary files
Automate backups
Schedule service restarts
Run custom scripts

Multi‑Site Management

aaPanel supports hosting multiple websites on a single server:

Each site has independent configuration
Different domains can point to different directories
Assign different PHP versions per site
Separate databases and FTP accounts for each site

Docker Support

aaPanel also includes Docker container management:

Install Docker from the App Store
Easily deploy various containerised applications
Manage images and containers
Configure container networking and storage

Conclusion

By following this guide, you should now have a solid understanding of the complete workflow for using aaPanel to build a website on a Linux server – from installation and site creation to domain configuration, security, and performance optimisation. aaPanel's intuitive graphical interface significantly lowers the barrier to website deployment.

Key takeaways:

Ensure your server meets the basic requirements before installation
After first installation, promptly change default settings to improve security
Configure your website parameters properly for optimal performance
Establish regular backup and maintenance habits
Choose the right optimisation strategies based on your actual needs

As you become more familiar with aaPanel, you can explore advanced features like load balancing and cluster deployment. We hope this tutorial helps you successfully launch your own website!

The Ultimate Windows Guide to Hosting a Minecraft Bedrock Server

Tue, 07 Apr 2026 00:00:00 GMT

The Ultimate Windows Guide to Hosting a Minecraft Bedrock Server

Hey everyone! Following up on the Java Edition server guide, today we're covering Minecraft Bedrock Edition server hosting. Bedrock's biggest advantage is cross‑platform play – it lets players on Windows 10/11, mobile (iOS/Android), tablets, Xbox, Switch, and more all play together on the same server!

Bedrock vs Java – What's the Difference?

Before we start, here's a quick comparison:

Feature	Java Edition	Bedrock Edition
Platforms	PC only (Windows/Mac/Linux)	Multi‑platform (PC/mobile/consoles)
Language	Java	C++
Mod Support	Rich (Forge/Fabric)	Limited (plugins/Add‑ons)
Performance	Java‑dependent optimisation	Native, better performance
Redstone	Original mechanics	Slightly different
Commands	Traditional syntax	Newer syntax
Best For	PC players, mod enthusiasts	Cross‑platform, casual players

Which one should you choose?

If you have friends on mobile or consoles → Bedrock

If you want lots of mods → Java

If everyone is on PC and loves technical gameplay → Java

Prerequisites

1. System Requirements

OS: Windows 10 or Windows 11 (64‑bit)
RAM: At least 4 GB total (Bedrock is lighter than Java)
Storage: Minimum 5 GB free space
Network: Stable internet connection
Microsoft Account: Required for authentication

2. Software You'll Need

Minecraft Bedrock Dedicated Server (official)
Text editor (Notepad++ or VS Code recommended)
Optional: Port forwarding tools (for online play)

Note: Bedrock servers do NOT require Java – that's the biggest difference from Java Edition!

Step 1: Download the Bedrock Server

Option A: Official Bedrock Dedicated Server (Recommended)

Visit the official download page
https://www.minecraft.net/en-us/download/server/bedrock
Choose the Windows version
- Look for "Bedrock Dedicated Server for Windows"
- Click the "Download" button – you'll get a .zip file
Extract the files
- Create a folder on D: or E: drive, e.g., D:\MC_Bedrock_Server
- Right‑click the .zip → Extract Here
- Copy all files into D:\MC_Bedrock_Server

File structure – after extraction, you should see:

D:\MC_Bedrock_Server\
├── bedrock_server.exe          # Main executable
├── bedrock_server_how_to.html  # Instructions
├── permissions.json            # Permission config
├── server.properties           # Server config
├── whitelist.json              # Whitelist
├── worlds\                     # World folder
└── ... other files

Option B: Third‑Party Server Software (More Features)

If you want plugins or extra features, try these:

Popular alternatives:

PocketMine‑MP: https://pmmp.io/ (plugin‑supported)
Nukkit: https://cloudburstmc.org/ (Java‑based, cross‑platform)
LeviLamina: https://lamina.levimc.org/ (lightweight)

For beginners: Start with the official server – it's the simplest and most stable.

Step 2: First Server Launch

1. Direct Launch (Easiest)

Navigate to D:\MC_Bedrock_Server
Double‑click bedrock_server.exe
A command prompt window will open

Wait until you see something like:

Server started.
IPv4 supported, port: 19132
IPv6 supported, port: 19133

Your server is running!

Note: The first launch generates a default world – this may take a minute or two.

2. Create a Startup Script (Recommended)

For easier management, create a start.bat file:

In the server folder, right‑click → New → Text Document
Rename it to start.bat
Right‑click and edit with Notepad, paste this:

@echo off
title Minecraft Bedrock Server
echo Starting Minecraft Bedrock server...
echo.
bedrock_server.exe
pause

Save it – from now on, just double‑click start.bat to launch.

3. Firewall Configuration

On first launch, Windows Firewall may pop up:

You'll see a "Windows Security Alert" window
Check both "Private" and "Public" networks
Click "Allow access"

If you missed the popup, set it manually:

Control Panel → Windows Defender Firewall → Advanced Settings
Inbound Rules → New Rule
Port → UDP → Specific local ports: 19132,19133
Allow the connection
Name the rule (e.g., "Minecraft Bedrock Server") and save

Important: Bedrock uses UDP protocol, not TCP!

Step 3: Configure the Server

1. Edit server.properties

Shut down the server (press Ctrl + C in the command window), then open server.properties with Notepad.

Here are the key settings and what they do:

# ===== Basic Settings =====

# Server name (shown in the server list)
server-name=Minecraft Bedrock Server

# Game mode: survival, creative, adventure
gamemode=survival

# Difficulty: peaceful, easy, normal, hard
difficulty=normal

# Allow cheats (commands)
allow-cheats=false

# Maximum number of players
max-players=10

# Server port (UDP)
server-port=19132

# IPv6 port
server-portv6=19133

# Online mode (premium authentication)
online-mode=true

# Enable whitelist
white-list=false

# Allow player damage
player-damage=true

# Drop items on death
drop-items-on-death=true

# ===== World Settings =====

# World name
level-name=Bedrock level

# World seed (leave blank for random)
level-seed=

# World type: default, flat, legacy
level-type=default

# Default permission level: member, operator, visitor
default-player-permission-level=member

# Generate structures (villages, fortresses, etc.)
generate-structures=true

# Bonus chest at spawn
bonus-chest-enabled=false

# Spawn protection radius (blocks)
spawn-protection=16

# ===== Performance Settings =====

# View distance (in chunks)
tick-distance=4

# Max threads (0 = auto)
max-threads=8

# Enable education features
education-feature-enabled=false

# ===== Other Settings =====

# MOTD (message of the day)
motd=A Minecraft Bedrock Server

# Emit server telemetry
emit-server-telemetry=true

# Content log file
content-log-file-enabled=false

# Compression threshold
compression-threshold=1

# Allow flight
allow-flight=false

# Client-side chunk generation
client-side-chunk-generation-enabled=true

# Encryption
encryption-enabled=true

# Texture pack required
texturepack-required=false

# Server authoritative movement
server-authoritative-movement=server-auth

# Correct player movement
correct-player-movement=true

# Server authoritative block breaking
server-authoritative-block-breaking=true

2. Important Settings Explained

Game Mode

gamemode=survival    # Survival (recommended)
gamemode=creative    # Creative
gamemode=adventure   # Adventure (can't break blocks)

Player Management

# Max players (adjust based on your upload speed)
# LAN: 10‑20
# Online: 5‑10 (depends on upload bandwidth)
max-players=10

# Default permission level
# member = regular player
# operator = admin (can use commands)
# visitor = spectator only
default-player-permission-level=member

Online Mode (Authentication)

# true = only Microsoft‑authenticated accounts (recommended)
# false = offline accounts allowed (not recommended – security risk)
online-mode=true

World Generation

# Custom seed – find nice ones online
level-seed=-1234567890

# Superflat world (great for creative building)
level-type=flat

# View distance (affects performance)
# Low‑end PC: 2‑3
# Mid‑range: 4‑6
# High‑end: 8‑12
tick-distance=4

Flight Settings

# Allow flight in survival mode
# Creative mode always allows flight
allow-flight=false

3. Save and Restart

Save server.properties (Ctrl + S)
Restart the server (double‑click start.bat or bedrock_server.exe)
New settings take effect

Step 4: Connect to Your Server

Local Connection (Same PC)

Open Minecraft Bedrock client
Click Play → Servers
Scroll to the bottom and click Add Server
Fill in:
- Server Name: Anything, e.g., "My Server"
- Server Address: localhost or 127.0.0.1
- Port: 19132 (if unchanged)
Click Save
Click the server name to join

LAN Connection (Same WiFi)

Bedrock makes LAN discovery super easy!

Method A: Auto‑Discovery (Easiest)

Make sure all devices are on the same WiFi
Start your server
On mobile/tablet, open Minecraft
Click Play → Friends tab
Your server should appear automatically
Tap to join

Method B: Manual Addition

If auto‑discovery doesn't work:

Find your PC's local IP
- Press Win + R, type cmd, hit Enter
- Run ipconfig and press Enter
- Look for IPv4 Address – e.g., 192.168.1.100
Add server on other devices
- Open Minecraft
- Play → Servers → Add Server
- Server Address: 192.168.1.100
- Port: 19132
- Save and join

Tip: Make sure all devices are logged into Microsoft accounts.

Online Connection (Let Friends Join from the Internet)

Method A: Router Port Forwarding (Requires Public IP)

Check if you have a public IP
- Visit https://ip.cn to see your public IP
- Log into your router and check the WAN IP
- If they match, you have a public IP
Log into your router
- Open browser to 192.168.1.1 or 192.168.0.1
- Enter admin credentials
Set up port forwarding
- Find "Port Forwarding", "Virtual Server", or "NAT"
- Add a rule:
  - External Port: 19132
  - Internal Port: 19132
  - Internal IP: your PC's IP
  - Protocol: UDP (important!)
- Also add port 19133 (IPv6)
- Save
Give your public IP to friends
- They add: your.public.ip:19132
- Example: 123.45.67.89:19132

Important:

Bedrock uses UDP, not TCP!

Make sure your router forwards UDP ports

Public IP may change – consider using DDNS

Method B: Tunneling Tools (No Public IP Needed)

If you don't have a public IP, use tunneling:

Recommended tools:

SakuraFrp: https://www.natfrp.com/
Ngrok: https://ngrok.com/
ZeroTier: https://www.zerotier.com/ (virtual LAN)

Using SakuraFrp:

Create an account and download the client
Create a tunnel:
- Type: UDP
- Local IP: 127.0.0.1
- Local Port: 19132
Start the tunnel – get a domain name
Friends connect using that domain

Using ZeroTier (great for small groups):

All players install ZeroTier
Create a network and share the network ID
Everyone joins the network
Connect using the ZeroTier virtual IP
No port forwarding needed – secure and easy

Method C: Minecraft Realms (Official Solution)

If self‑hosting is too much hassle, consider Minecraft Realms:

Pros: Official, stable, one‑click invites
Cons: Paid subscription (~$7.99/month)
Purchase: Directly in‑game

Step 5: Server Administration

Common Console Commands

In the server command window, type commands directly:

# Show help
help

# Stop the server
stop

# Save the world
save hold      # Pause auto‑save
save resume    # Resume auto‑save
save query     # Check save status

# Grant OP (admin) privileges
op PlayerName

# Remove OP
deop PlayerName

# Ban a player
ban PlayerName

# Unban a player
pardon PlayerName

# List online players
list

# Change game mode (in‑game or console)
gamemode survival PlayerName
gamemode creative PlayerName
gamemode adventure PlayerName

# Teleport
tp Player1 Player2
tp PlayerName x y z

# Broadcast a message
say Hello everyone, welcome to the server!

# Set world spawn
setworldspawn
setworldspawn x y z

# List all game rules
gamerule

# Keep inventory on death
gamerule keepinventory true

# Set time
time set day
time set night
time set noon

# Set weather
weather clear
weather rain
weather thunder

# Change difficulty
difficulty peaceful
difficulty easy
difficulty normal
difficulty hard

# Kick a player
kick PlayerName

# Check server performance
tickingarea list

Managing Player Permissions

1. Using OP (Operator) Status

OPs have full command access.

# In console:
op Steve

# In‑game (if already OP):
/op Steve

OP players can:

Use all commands
Ignore spawn protection
Fly in any mode
Break and place any block

2. Using a Whitelist

Allow only specific players to join:

# Enable whitelist
whitelist on

# Add players (use Xbox Gamertag or Microsoft account name)
whitelist add Steve
whitelist add Alex

# Remove a player
whitelist remove Steve

# List whitelisted players
whitelist list

# Reload the whitelist
whitelist reload

You can also enable in server.properties:

white-list=true

Then manually add players in whitelist.json.

3. Editing permissions.json

Bedrock has a more granular permission system.

Open permissions.json:

[
  {
    "permission": "operator",
    "xuid": "1234567890123456"
  },
  {
    "permission": "member",
    "xuid": "9876543210987654"
  }
]

Permission levels:

operator – full admin access
member – regular player
visitor – spectator only

Getting a player's XUID:

When a player joins, their XUID shows in the console
Or use online lookup tools

Backing Up World Data

Regular backups are essential!

Manual Backup

Stop the server (type stop)
Copy the worlds folder
Paste to another location as backup
Restart the server

Automated Backup Script

Create backup.bat:

@echo off
set backup_folder=D:\MC_Bedrock_Backup
set server_folder=D:\MC_Bedrock_Server
set date_str=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
set date_str=%date_str: =0%

if not exist "%backup_folder%" mkdir "%backup_folder%"

echo Backing up Bedrock server...
xcopy "%server_folder%\worlds" "%backup_folder%\worlds_%date_str%" /E /I /Y

echo Backup complete!
pause

Run this weekly, or before any major changes.

Step 6: Cross‑Platform Play Guide

Bedrock's biggest strength is cross‑platform play! Here are platform‑specific notes:

Windows 10/11 PC

Buy Minecraft for Windows from the Microsoft Store
Log in with a Microsoft account
Fully compatible with mobile versions

Mobile / Tablet (iOS / Android)

Download from App Store or Google Play
Paid purchase required (~$6.99‑$7.99)
Touch‑optimised interface
Fully cross‑plays with PC

Xbox One / Series

Requires Xbox Live Gold (for online play)
Controller‑based
Full cross‑platform support

Nintendo Switch

Requires Nintendo Switch Online membership
Portable mode works too
Slightly lower performance than other platforms

PlayStation 4 / 5

Requires PlayStation Plus membership
Supports cross‑play (requires linking Sony account to Microsoft account)

Tip: All players need a Microsoft account for cross‑platform play.

Step 7: Performance Optimisation

1. server.properties Optimisation

# Lower view distance (biggest impact)
tick-distance=4

# Limit max players
max-players=10

# Adjust threads (based on CPU cores)
max-threads=4

# Disable unnecessary features
education-feature-enabled=false
content-log-file-enabled=false

2. System Optimisation

Close background apps:

Browsers, video software, etc. – they steal CPU and RAM

Set process priority:

Open Task Manager (Ctrl + Shift + Esc)
Find bedrock_server.exe
Right‑click → Go to Details
Right‑click the process → Set Priority → High

Power plan:

Control Panel → Power Options
Select "High Performance"

3. Network Optimisation

Check upload speed:

Test at https://speedtest.net
Bedrock requires ~0.1‑0.5 Mbps per player
10 players = at least 5‑10 Mbps upload

Use a wired connection:

Ethernet is better than WiFi – less lag and packet loss

Port forwarding best practices:

Only open necessary ports (19132 UDP)
Don't open extra ports on your firewall

Common Issues & Solutions

1. Server Won't Start

Symptom: bedrock_server.exe opens and closes instantly.

Fixes:

Check if antivirus is blocking it – add folder to whitelist
Ensure no other server instance is running
Check if the port is already in use
Run as Administrator

2. Friends Can't Connect

Symptom: "Can't connect to world" or "Connection timed out".

Fixes:

Is the server actually running?
Double‑check IP address and port
Confirm firewall allows UDP ports 19132‑19133
LAN: make sure everyone is on the same WiFi
Online: check port forwarding (UDP!)

Test connectivity:

# On friend's PC
ping your.ip.address

# Test port (requires telnet)
telnet your.ip.address 19132

3. LAN Auto‑Discovery Not Working

Symptom: Server doesn't show up on mobile devices.

Fixes:

Ensure all devices are on the same WiFi network
Check if router has AP isolation enabled (turn it off)
Try manually adding the server (use IP)
Restart router and all devices
Make sure everyone is logged into Microsoft accounts

4. Server is Laggy / High Ping

Symptom: Low TPS, players experience delays.

Fixes:

Lower tick-distance (view distance)
Reduce max-players
Disable redstone machines near spawn
Clear excess entities (animals, monsters, items)
Check upload bandwidth
Use a wired network connection

Clear entities command:

# Remove all dropped items
kill @e[type=item]

# Remove experience orbs
kill @e[type=xp_orb]

# Remove entities within a radius (except players)
kill @e[r=50,type=!player]

5. Authentication Issues

Symptom: Friends get "Failed to authenticate" errors.

Fixes:

Ensure all players own a legitimate copy of Minecraft
Make sure everyone is logged into Microsoft accounts
If everyone uses offline/cracked launchers, set online-mode=false
Warning: Offline mode is insecure – anyone can join with any name

6. Port Already in Use

Symptom: Error says port is already occupied.

Fixes:

Change server-port in server.properties to another value (e.g., 19134)

Or find and kill the process using the port:

netstat -ano | findstr :19132
taskkill /PID process_id /F

7. World Corruption

Symptom: Server fails to load the world.

Fixes:

Restore from a backup
Delete the corrupted world and generate a new one
Use third‑party repair tools (e.g., Universal Minecraft Editor)

8. Xbox Live Connectivity

Symptom: Error about Xbox Live service being unavailable.

Fixes:

Check Xbox Live status: https://support.xbox.com/
Ensure all players are signed into Microsoft accounts
Restart the game and server
Check network connection

Final Words

Congratulations! You've successfully set up your own Minecraft Bedrock Edition server!

Quick recap:

Downloaded the official Bedrock server
Extracted and launched it for the first time
Configured server.properties
Set up firewall (UDP ports)
Connected (local / LAN / online)
Managed players and permissions
Set up backups and optimisation

Bedrock's strengths:

Cross‑platform: Mobile, PC, and consoles all together
Performance: C++ native code, lightweight
Easy setup: Simple config, auto‑discovers LAN servers
Mobile friendly: Play anywhere, anytime

What you can do next:

Invite friends from different platforms
Explore Bedrock‑specific Add‑ons and Marketplace content
Learn more Bedrock commands and mechanics
Build mini‑game maps
Grow your cross‑platform community

Java vs Bedrock – which one to pick?

Want mods and technical gameplay → Java
Want cross‑platform with friends → Bedrock
Casual players / kids → Bedrock
Redstone engineers / builders → Java
Why not try both? They each have their own charm!

If you run into issues:

Check the console logs
Visit Minecraft forums
Join Bedrock player communities
Look for video tutorials online

Have a blast in the world of Minecraft – no matter what device you're on, may you create wonderful memories with your friends!

Useful Links:

Minecraft Official Site: https://www.minecraft.net/
Bedrock Server Download: https://www.minecraft.net/en-us/download/server/bedrock
Minecraft Wiki (Bedrock): https://minecraft.wiki/w/Bedrock_Edition
Xbox Service Status: https://support.xbox.com/
PocketMine‑MP: https://pmmp.io/
ZeroTier: https://www.zerotier.com/

The Ultimate Windows Guide to Hosting a Minecraft Java Server

Tue, 07 Apr 2026 00:00:00 GMT

The Ultimate Windows Guide to Hosting a Minecraft Java Server

Hey everyone! This is a super‑detailed, beginner‑friendly tutorial on setting up your very own Minecraft Java Edition server on a Windows PC. Whether you want to play with a few friends or run a small community server, this guide has you covered – zero prior experience needed.

Prerequisites

1. System Requirements

OS: Windows 10 or Windows 11 (64‑bit)
RAM: At least 8 GB total (system + server usage)
Storage: Minimum 10 GB free space
Network: Stable internet connection

2. Software You'll Need

Java Runtime Environment (JRE / JDK)
Minecraft server core file (jar)
Text editor (Notepad++ or VS Code recommended, but regular Notepad works too)

Step 1: Install Java

Minecraft Java servers require Java to run. Here are two ways to get it.

Option A: Official JDK (Recommended)

Visit the Oracle download page
- https://www.oracle.com/java/technologies/downloads/
- Alternatively, use the open‑source OpenJDK: https://adoptium.net/
Pick the right version
- Minecraft 1.17 and below → Java 8 or Java 11
- Minecraft 1.18 – 1.20.4 → Java 17
- Minecraft 1.20.5 and above → Java 21
Recommendation: If you're unsure, download Java 17 – it's the most compatible across versions.
Download the installer
- Choose Windows x64 Installer (.exe) and download it.
Run the installer
- Double‑click the .exe, click "Next", and keep the default installation path.

Verify the installation

Press Win + R, type cmd, and hit Enter.
Run this command:
```
java -version
```

You should see something like:

java version "17.0.x" 202x-xx-xx LTS
Java(TM) SE Runtime Environment (build 17.0.x+x-LTS-x)
Java HotSpot(TM) 64-Bit Server VM (build 17.0.x+x-LTS-x, mixed mode, sharing)

If you see that, Java is installed correctly.

Option B: Microsoft Store (Simpler)

Open the Microsoft Store.
Search for "Java" or "OpenJDK".
Pick a version and install it.
Verify with java -version as above.

Step 2: Download the Minecraft Server Jar

1. Choose Your Game Version

Decide which Minecraft version you want to run. Popular choices:

1.20.1 – Stable, lots of mod support
1.19.4 – Relatively new
1.16.5 – Classic, well‑optimised
1.12.2 – Old but with a huge mod ecosystem

For beginners: Go with 1.20.1 – it has new features and is rock‑solid.

2. Download the Server Core File

Option A: Official Vanilla Server (Pure, unmodded)

Visit the official Minecraft server download page:
https://www.minecraft.net/en-us/download/server
(The page auto‑detects your OS – just click the jar link.)
Click the download button for minecraft_server.x.xx.x.jar (e.g., minecraft_server.1.20.1.jar).
Save it to a new folder – preferably on a drive other than C: (e.g., D:\MC_Server). Put the jar file inside that folder.

Option B: Paper Server (Recommended – Better Performance)

Paper is an optimised fork of the vanilla server – better performance, fewer bugs.

Go to the PaperMC downloads page:
https://papermc.io/downloads/paper
Select your Minecraft version (e.g., 1.20.1).
Download the latest build .jar file.
Place it in your server folder (e.g., D:\MC_Server).

Option C: Forge / Fabric Server (For Mods)

If you want to run mods, you'll need a mod‑loader server.

Forge Server:

Visit https://files.minecraftforge.net/
Choose your version and download the "Installer".
Run the installer, select "Install server", and point it to your server folder.

Fabric Server:

Visit https://fabricmc.net/use/server/
Download the server jar for your version.

Newbie tip: Start with Paper – it's the easiest to set up and manage. Move to Forge/Fabric later when you're comfortable.

Step 3: First Server Launch

1. Create a Startup Script

Inside your server folder (e.g., D:\MC_Server):

Right‑click in an empty space → New → Text Document.
Rename it to start.bat (make sure the extension is .bat, not .txt).
If you don't see file extensions, go to the folder's View tab and check "File name extensions".
Right‑click start.bat and open it with Notepad.
Paste the appropriate content based on your server type:

For Vanilla server:

@echo off
java -Xms2G -Xmx4G -jar minecraft_server.1.20.1.jar nogui
pause

For Paper server:

@echo off
java -Xms2G -Xmx4G -jar paper-1.20.1-xxx.jar nogui
pause

What these parameters mean:

-Xms2G – initial RAM allocation (2 GB)

-Xmx4G – maximum RAM (4 GB) – adjust based on your system

-jar – the jar file to run

nogui – disables the graphical interface (saves resources)

pause – keeps the window open so you can see any errors

Memory allocation tips:

8 GB RAM total → -Xms2G -Xmx4G

16 GB RAM total → -Xms4G -Xmx8G

32 GB RAM total → -Xms8G -Xmx16G

Don't allocate more than 70% of your physical RAM – otherwise your system will slow down.

2. Accept the EULA

Double‑click start.bat to run it.
The server will start and then immediately close – that's normal.
A new file called eula.txt will appear in the folder.
Open eula.txt in Notepad.
Find this line:
```
eula=false
```
Change it to:
```
eula=true
```
Save and close the file.

3. Launch the Server Properly

Double‑click start.bat again.
This time the server will start normally and stay running.

Wait until you see a message like:

[Server thread/INFO]: Done (xx.xxxs)! For help, type "help"

That means your server is up and running!

Note: The first launch generates the world files – this may take a few minutes. Please be patient.

Step 4: Configure the Server

Your server folder now contains many files. The most important one is server.properties – the main configuration file.

1. Edit server.properties

Shut down the server (press Ctrl + C in the command window, or type stop and press Enter).
Open server.properties with Notepad or any text editor.
Here are the key settings and what they do:

# ===== Basic Settings =====
# Server port (default 25565 – usually leave as is)
server-port=25565

# Server name (shown in the server list)
motd=A Minecraft Server

# Maximum number of players
max-players=20

# Game mode: survival, creative, adventure, spectator
gamemode=survival

# Difficulty: peaceful, easy, normal, hard
difficulty=normal

# Enable PvP (player vs player)
pvp=true

# Allow flying (creative mode always allows it)
allow-flight=false

# World folder name
level-name=world

# World seed (leave blank for random)
level-seed=

# ===== Advanced Settings =====
# View distance (in chunks – higher = more lag)
view-distance=10

# Simulation distance
simulation-distance=10

# Generate structures (villages, fortresses, etc.)
generate-structures=true

# Enable command blocks
enable-command-block=false

# OP permission level (1–4)
op-permission-level=4

# Enable whitelist
white-list=false

# Online mode (authentication)
# true = only premium accounts can join
# false = offline / cracked accounts allowed
online-mode=true

# Max build height
max-build-height=256

# Allow the Nether
allow-nether=true

# Allow the End
allow-end=true

# Spawn animals
spawn-animals=true

# Spawn monsters
spawn-monsters=true

# Spawn NPC villagers
spawn-npcs=true

# Spawn protection radius
spawn-protection=16

# Player idle timeout (minutes, 0 = disabled)
player-idle-timeout=0

# Network compression threshold
network-compression-threshold=256

# Enable query (for server status tools)
enable-query=false

# RCON remote console password (if you need remote control)
rcon.password=
rcon.port=25575

2. Important Settings Explained

Game Mode

gamemode=survival          # Survival
gamemode=creative          # Creative
gamemode=adventure         # Adventure
gamemode=spectator         # Spectator

Player Management

# Max players (adjust based on your PC and internet)
max-players=10

# Whitelist mode (only allowed players can join)
white-list=true

# After enabling whitelist, add players via console or in‑game:
# /whitelist add PlayerName

Online Mode (Authentication)

# If all players have premium accounts, keep this on
online-mode=true

# If some friends use cracked launchers, turn this off
online-mode=false

Warning: Setting online-mode=false allows anyone to join with any username – be aware of the security implications.

World Generation

# Custom seed (you can find nice ones online)
level-seed=123456789

# Or use a text seed
level-seed=village_at_spawn

# View distance (affects performance and visibility)
# Low‑end PC: 6–8
# Mid‑range: 10–12
# High‑end: 15–20
view-distance=10

3. Save and Apply Changes

Save server.properties (Ctrl + S).
Restart your server by double‑clicking start.bat.
The new settings will take effect.

Step 5: Connect to Your Server

Local Connection (Same PC)

Launch the Minecraft Java client.
Click Multiplayer.
Click Add Server or Direct Connect.
Enter localhost or 127.0.0.1 as the server address.
Click Done or Join Server.
You're in!

LAN Connection (Other Devices on the Same Network)

Find your local IP address:
- Press Win + R, type cmd, and hit Enter.
- Run ipconfig and press Enter.
- Look for the IPv4 Address – it will look like 192.168.1.100.
On the other device:
- Open Minecraft.
- Go to Multiplayer → Add Server.
- Enter your IPv4 address as the server address, e.g., 192.168.1.100.
- If you changed the port, use 192.168.1.100:25565.

Online Connection (Let Friends Join from the Internet)

This requires port forwarding or tunneling – it's a bit more advanced.

Method A: Router Port Forwarding (Requires a Public IP)

Check if you have a public IP
- Visit https://ip.cn to see your public IP.
- Log into your router's admin page and check the WAN IP.
- If they match, you have a public IP.
Log into your router
- Open a browser and go to your router's address (usually 192.168.1.1 or 192.168.0.1).
- Enter the admin username and password.
Set up port forwarding / virtual server
- Find "Port Forwarding", "Virtual Server", or "NAT Settings".
- Add a new rule:
  - External Port: 25565
  - Internal Port: 25565
  - Internal IP: your PC's IP (e.g., 192.168.1.100)
  - Protocol: TCP & UDP (or select "Both")
- Save the settings.
Give your public IP to friends
- They connect using: your.public.ip:25565
- Example: 123.45.67.89:25565

Notes:

Your public IP may change after each router reboot – consider using DDNS.

Make sure your Windows firewall allows port 25565.

Method B: Tunneling Tools (No Public IP Needed)

If your ISP doesn't give you a public IP, use a tunneling service:

Popular tools:

SakuraFrp: https://www.natfrp.com/ (free tier available)
Ngrok: https://ngrok.com/
Hamachi (LogMeIn)

Using SakuraFrp as an example:

Create an account.
Download and run the client.
Create a tunnel:
- Type: TCP
- Local IP: 127.0.0.1
- Local Port: 25565
Start the tunnel – you'll get a domain name.
Friends connect using that domain.

Method C: Third‑Party Hosting Platforms

If all of this sounds like too much work, try free hosting services:

Aternos: https://aternos.org/ (completely free)
Minehut: https://minehut.com/
Or rent a VPS from cloud providers like AWS, Google Cloud, or Alibaba Cloud.

Step 6: Server Administration

Common Console Commands

In the server command window, you can type commands directly:

# Show help
help

# Stop the server gracefully
stop

# Save the world
save-all

# Disable auto‑save (for backups)
save-off
save-on        # Re‑enable auto‑save

# Grant OP (admin) privileges
op PlayerName

# Remove OP
deop PlayerName

# Ban a player
ban PlayerName

# Unban a player
pardon PlayerName

# List online players
list

# Change game mode (in‑game or console)
gamemode survival PlayerName
gamemode creative PlayerName

# Teleport
tp Player1 Player2

# Broadcast a message
say Hello everyone, welcome to the server!

# Set world spawn
setworldspawn

# List all game rules
gamerule

# Keep inventory on death
gamerule keepInventory true

# Set time
time set day
time set night
time set noon

# Set weather
weather clear
weather rain
weather thunder

Managing Player Permissions

1. Using OP (Operator) Status

OPs have full administrative control.

# In console:
op Steve

# In‑game (if you're already OP):
/op Steve

2. Using a Whitelist

Allow only specific players to join:

# Enable whitelist
whitelist on

# Add players
whitelist add Steve
whitelist add Alex

# Remove a player
whitelist remove Steve

# List whitelisted players
whitelist list

# Reload the whitelist
whitelist reload

You can also enable the whitelist in server.properties:

white-list=true

Then manually edit whitelist.json to add player UUIDs – but the in‑game commands are easier.

Installing Plugins (Paper / Spigot Only)

If you're using Paper or Spigot, you can add plugins to extend functionality.

Find plugins
Popular repositories:
- SpigotMC Resources
- Bukkit Plugins
Recommended plugins for beginners:
- EssentialsX – essential admin commands
- WorldEdit – world editing tools
- CoreProtect – block logging (anti‑griefing)
- LuckPerms – advanced permissions
- Vault – economy system foundation
- Multiverse‑Core – multi‑world management
Installation:
- Shut down the server.
- Place the downloaded .jar plugin files into the plugins folder.
- Start the server – plugins load automatically and generate config files.
Configuration:
- Each plugin creates its own config folder inside plugins.
- Edit the config files as needed.
- Reload plugins: /reload or /pluginname reload (check the plugin's docs).

Backing Up World Data

Regular backups are essential!

Manual Backup

Stop the server (type stop).
Copy the entire server folder, or just the world, world_nether, and world_the_end folders.
Paste them somewhere safe (another drive, cloud storage, etc.).
Restart the server.

Automated Backup Script

Create a backup.bat file in your server folder:

@echo off
set backup_folder=D:\MC_Backup
set server_folder=D:\MC_Server
set date_str=%date:~0,4%%date:~5,2%%date:~8,2%_%time:~0,2%%time:~3,2%%time:~6,2%
set date_str=%date_str: =0%

if not exist "%backup_folder%" mkdir "%backup_folder%"

echo Backing up server...
xcopy "%server_folder%\world" "%backup_folder%\world_%date_str%" /E /I /Y
xcopy "%server_folder%\world_nether" "%backup_folder%\world_nether_%date_str%" /E /I /Y
xcopy "%server_folder%\world_the_end" "%backup_folder%\world_the_end_%date_str%" /E /I /Y

echo Backup complete!
pause

Run this script weekly, or before any major changes.

Step 7: Performance Optimisation

1. JVM Tuning

Modify your start.bat with advanced JVM flags for better performance:

@echo off
java -Xms4G -Xmx8G ^
     -XX:+UseG1GC ^
     -XX:+ParallelRefProcEnabled ^
     -XX:MaxGCPauseMillis=200 ^
     -XX:+UnlockExperimentalVMOptions ^
     -XX:+DisableExplicitGC ^
     -XX:+AlwaysPreTouch ^
     -XX:G1NewSizePercent=30 ^
     -XX:G1MaxNewSizePercent=40 ^
     -XX:G1HeapRegionSize=8M ^
     -XX:G1ReservePercent=20 ^
     -XX:G1HeapWastePercent=5 ^
     -XX:G1MixedGCCountTarget=4 ^
     -XX:InitiatingHeapOccupancyPercent=15 ^
     -XX:G1MixedGCLiveThresholdPercent=90 ^
     -XX:G1RSetUpdatingPauseTimePercent=5 ^
     -XX:SurvivorRatio=32 ^
     -XX:+PerfDisableSharedMem ^
     -XX:MaxTenuringThreshold=1 ^
     -jar paper-1.20.1-xxx.jar nogui
pause

2. server.properties Optimisation

# Lower view distance (biggest performance impact)
view-distance=8
simulation-distance=8

# Tune entity spawning
spawn-animals=true
spawn-monsters=true
spawn-npcs=true

# Disable unnecessary features
enable-command-block=false
spawn-protection=0

# Adjust tick rate (change with caution)
max-tick-time=60000

3. Pre‑Generate the World

Use a plugin to pre‑generate chunks, reducing lag when players explore:

Chunky plugin: https://www.spigotmc.org/resources/chunky.81534/

Commands:

/chunky radius 1000
/chunky start

4. Clean Up Unused Data

Regularly remove:

Offline player inventory data
Old log files
Outdated backups

Common Issues & Solutions

1. Server Fails to Start

Symptoms: start.bat opens and closes instantly.

Fixes:

Verify Java is installed: java -version
Check that the jar filename in start.bat matches the actual file name.
Look at the log files in the logs folder.
Make sure you've set eula=true in eula.txt.

2. Can't Connect to Server

Symptoms: "Connection timed out" or "Can't connect to server".

Fixes:

Is the server actually running?
Double‑check the IP address and port.
Is Windows Firewall blocking port 25565?
Test locally with localhost first.

To open Windows Firewall port:

Control Panel → Windows Defender Firewall → Advanced Settings.
Inbound Rules → New Rule.
Port → TCP → Specific local ports: 25565.
Allow the connection.
Name the rule and save.

3. Server is Laggy

Symptoms: TPS (ticks per second) drops below 20.

Fixes:

Reduce view‑distance.
Lower max‑players.
Limit redstone and mob farms.
Install optimisation plugins like ClearLag.
Allocate more RAM (but not too much).
Pre‑generate the world with Chunky.

4. Authentication Problems

Symptom: Friends can't join, getting "Failed to authenticate".

Fixes:

If everyone has premium accounts: keep online-mode=true.
If some use cracked launchers: set online-mode=false.
Remember that offline mode is less secure.

5. Port Already in Use

Symptom: "Address already in use" error.

Fixes:

Change server-port in server.properties to another number (e.g., 25566).

Or find and kill the process using the port:

netstat -ano | findstr :25565
taskkill /PID process_id /F

6. World Corruption

Symptom: Server fails with world loading errors.

Fixes:

Restore from a backup.
Use repair tools like MCEdit (older versions).
Delete corrupted region files (last resort).

Final Words

Congratulations! You've now successfully set up your very own Minecraft Java Edition server.

Quick recap of what you've done:

Installed Java
Downloaded a server core
Created a startup script and accepted the EULA
Configured server.properties
Started the server and connected
Set up port forwarding (for online play)
Installed plugins and optimised performance

What you can do next:

Invite your friends to play
Explore plugins and mods
Learn advanced server admin commands
Create custom gameplay experiences
Build your own community

If you get stuck, here's where to look for help:

Check the server logs (logs folder)
Visit Minecraft forums (like MCBBS or Reddit)
Join Discord communities or QQ groups dedicated to server hosting

Have a blast in the world of Minecraft!

Useful Links:

Minecraft Official Site: https://www.minecraft.net/
PaperMC Downloads: https://papermc.io/
SpigotMC Plugins: https://www.spigotmc.org/
Minecraft Wiki: https://minecraft.wiki/

Inside the Maimai DX Arcade Cabinet – A Technical Breakdown

Tue, 24 Feb 2026 00:00:00 GMT

The Screen Setup

Many players assume that Maimai DX uses a single large screen split into two halves. In reality, it's two separate 43-inch displays mounted vertically side by side, as shown in the image above (credit: BV1kt4y1h7QW). I haven't been able to track down the exact manufacturer or model of these panels – if anyone knows, feel free to chime in.

The Base Hardware (System Board)

Maimai DX runs on SEGA's ALLS HX2 arcade board, which is essentially a custom‑spec PC. Here's what's inside:

Motherboard: Gigabyte MCH31AM‑S2 (Intel H310 chipset)
CPU: Intel Core i3‑8100
RAM: 2× 4 GB DDR4
GPU: NVIDIA GeForce GTX 1050 Ti
OS: Windows 10 IoT LTSC 2019

The Buttons

A common misconception is that the buttons are simple mechanical microswitches – but they're actually optical switches.

When you press a button, it compresses a spring leaf that blocks an infrared sensor, registering the hit. This design offers longer lifespan, faster response, and lower cost compared to traditional microswitches. In fact, this optical setup is standard across most rhythm game arcade cabinets, not just Maimai.

The Touchscreen

The touchscreen on Maimai DX is a capacitive touch panel – similar in principle to what you'd find on a smartphone, but with one key difference:

Instead of pinpointing exact touch coordinates like a phone does, the screen is divided into 34 touch zones. Touching anywhere within a zone triggers that specific zone, essentially acting like 34 independent touch switches. The underlying sensing technology, however, is still capacitive.

The touch panel is manufactured using magnetron sputtering, which slightly reduces light transmittance – that's why the screen can look a bit washed out or hazy from certain angles. On the plus side, this coating helps hide the internal sensor traces, so you don't see the zone wiring under normal viewing conditions.

The entire touch surface is a single piece of glass – reportedly 5 mm thick rare‑earth glass.

Interestingly, the touch panel material on Japanese cabs seems to differ from the ones used in Chinese (export) units. Japanese panels have a bit more flexibility and sit closer to the display, which gives them better image clarity and less noticeable parallax. The trade‑off is durability – they're more prone to showing white distortion under heavy hits, but they also suffer fewer issues like "ghost notes" or missed star inputs compared to the export versions.

Button Maintenance & Tuning

If you're looking to restore or maintain button feel, your best bet is to buy official replacement parts from Wahlap Technology – they're the only ones guaranteed to match the original factory feel.

Over time, heavy play can cause the spring leaf inside a button to deform, making it feel "mushy" or too soft. The button housing itself can also develop cracks or even have its retaining clips snap off. That leads to loose buttons, rattling noises on impact, or in worst‑case scenarios, the broken clip pieces can jam the button mechanism entirely, preventing it from returning or even pressing down.

For minor spring leaf deformation, you can carefully bend it back into shape – but don't overdo it, or you risk metal fatigue and a snapped leaf. If the housing is cracked or the clips are broken, just replace the button casing entirely.

Touchscreen Optimization & Fixes

There's no one‑size‑fits‑all cure for the dreaded "ghost touches," missed star inputs, or unresponsive zones – but in most cases, the root cause is one of these three:

Poor cabinet grounding
Dust buildup over time
Screen aging / wear

Poor grounding is by far the most common culprit. Without a solid ground connection, internal devices can create common‑mode interference, which messes with the capacitive touch controller – leading to phantom touches, dead zones, or erratic behavior. The fix is simple: make sure the cabinet is properly grounded.

As for the rumor that "unstable voltage causes missed star inputs" – I'm skeptical. Most modern arcades and shopping malls are already equipped with industrial‑grade power conditioning and filtering. Voltage fluctuations severe enough to affect the touchscreen are extremely rare in practice.

If you've already tried grounding, cleaning, and basic maintenance but still have issues, there's one more trick you can try: add an RS‑232 optical isolator.

Here's how:

Open the rear maintenance panel on the 2P side of the cabinet.
Locate the two serial cables labeled COM3 and COM4 on the system board – these carry the touch signals for Player 1 and Player 2, respectively.
Unplug both cables, then plug each one into an RS‑232 optical isolator, and plug the isolators back into the COM ports on the board.

This has been known to reduce or even eliminate ghost touch / missed input problems in some cases, by breaking ground loops and isolating electrical noise.

lanying's blog

DeepSeek-V4-Pro Production Release – Performance Closing in on Fable 5 at 1/60th the Price

Overview

V4 Series Parameter Recap

Ten Benchmark Comparisons: Preview vs Production

Horizontal Comparison: Against Flash Production and Competitors

The Technical Path Behind the Performance Gains

New Capabilities: Multimodal Support, Responses API, and Codex Integration

Pricing & Concurrency: 6 CNY/MT Output, Three Times Flash

Flash vs Pro – Selection Guide

Concurrency, API Migration, and Practical Tips

Concurrency Limits

API Migration Guide

Codex Integration

Cost Optimisation Tips

Pricing Trends and Industry Context

Summary

References

EdgeOne vs ESA – How Strong Is EdgeOne, What's the Difference, and Which One Should You Choose?

Product Positioning and Background

Tencent Cloud EdgeOne (EO)

Alibaba Cloud ESA

Free Tier Comparison

Getting the Free Tier

Site and Domain Quota Differences

Rate Limiting Details

Real‑World Performance and Node Coverage

Security Capabilities Comparison

Edge Ecosystem and Additional Capabilities

Static Site Hosting

Edge Functions and Edge Containers

Layer‑4 Proxy (L4)

DNS Hosting and Certificates

Logging and Other Features

Onboarding and Common Pitfalls

Paid Plan Pricing Reference

Selection Recommendations

Important Notes

Summary

References

DeepSeek-V4-Flash Production API Public Beta

Introduction

A Quick Recap of the V4 Series

Nine Benchmark Results

Same Model Architecture, Stronger Performance

Native Responses API Support with Codex Adaptation

Pricing

Peak/Off‑Peak Pricing

Upgrade Scope

More Technical Insights into the Maimai DX Cabinet – V2

Previous Article

Changes in Newer Batch Cabinets

Buttons, Screen, and Touch Panel

1. Screen

2. Buttons and Touch Panel

2.1 Buttons

2.2 Touch Panel

Build Your Own QQ Chatbot with Napcat & AstrBot!

WARNING!!!

Introduction

What are AstrBot and NapCat?

Prerequisites

Deployment

1. Create the Project Directory

2. Write docker-compose.yml

3. Start the Containers

Configuring NapCat (Logging into QQ)

1. Get the WebUI Token

2. Log into the WebUI

Configuring AstrBot

1. Access the Admin Dashboard

2. Connect an AI Model

3. Create a Bot Instance

Verify It Works

Common Issues

1. Docker image pull is too slow?

2. WebSocket connection fails?

3. QQ account got flagged/restricted?

4. How to update?

5. How to back up?