Professional independent Nvidia gpu tech. - AS403083 GPU assembly and rework in the USA. Repairing nvidia cards since 2021.
Grand Rapids, MI
Joined November 2024
- Tweets632
- Following19
- Followers8K
- Likes148
Pinned Tweet
Currently in China for 2-3 weeks making deals and talking to pcb/memory partners.
This trip got delayed while I was finishing augusts 4090 orders.
The people who purchased cards and upgrades for Nov 7th, those are delayed by 3 weeks until I begin upgrading again. Please reach out via email or dm if you have any questions or requests.
It is my honor to serve the community.
Greg Sky retweeted
🤯炸了！RTX 40系显卡算力被破解了，性能近乎翻倍！从4090到4060 Ti全都能用！
不刷BIOS、不改驱动，直接把被锁的Tensor算力放出来，跑AI矩阵乘快一倍左右。
实测亮点：
🔹4090：FP16/BF16 从165→331 TFLOPS，接近翻倍
🔹4080 SUPER：FP16涨到1.83倍，FP8也翻倍
🔹4070 Ti SUPER：FP16/BF16翻倍，FP8矩阵乘涨到1.95倍
🔹4060 Ti：还能涨近40%，Windows也能试
简单使用步骤（Linux）：
1️⃣下载安装：git clone github.com/JustAResearcher/g… && cd gpu-unlock-ada && make install
2️⃣运行 python3 tools/print_boot_option.py 拿到你的卡参数
3️⃣把参数加到开机启动项，关掉Secure Boot
4️⃣关机、拔电源等几秒再开机（必须冷启动）
敢试的今晚拔电源试试，评论区报结果！
Windows用户去搜：huggingface.co/TOTORONG/AD10…
仓库：github.com/JustAResearcher/g…
#RTX4090 #GPU解锁 #AI算力 #40系解锁 #本地AI
💥卧槽！6年前的3090被“破解”了！Tensor算力直接翻倍，官方锁了一半的性能被开源一键释放！
NVIDIA长期把30系显卡的FP16/BF16 Tensor指令发射速率限制在专业卡一半左右。
开源项目 gpu-unlock 用一个Linux内核模块，在驱动加载前改寄存器解锁，不刷BIOS、不改驱动、更新无忧。
牛逼之处：
🔹3090 Ti：BF16 87.5→157 TFLOPS，TF32 42.6→82.6 TFLOPS
🔹3080：FP16/BF16 约65→110–113 TFLOPS
🔹vLLM长上下文prefill快33–36%，首字延迟降25%
🔹结果比特一致，社区已在3070/3080/3090复现
关键步骤（仅Linux）：
1️⃣git clone github.com/Zagitta/gpu-unloc… && cd gpu-unlock && make install
2️⃣内核参数加 gpu_unlock.targets=all，关Secure Boot
3️⃣更新引导并重启
4️⃣journalctl -k | grep gpu_unlock 看到 unlocked 即成功
GitHub：github.com/Zagitta/gpu-unloc…
#RTX3090 #NVIDIA #AI算力 #开源 #GPU解锁 #本地AI #vLLM #黄仁勋
Repurposing a FLIR boson for imaging BGA rework areas under a microscope and bga station. This is the video that will be streamed to my AR surgeon glasses.
By end of year GPU LAB goes from "we upgrade your GPU" to "we host it too." Tier 2 datacenter in Grand Rapids, MI.
Bring your own GPU server, or have us build one. Upgraded 48GB 4090s or SXM-converted A100/H100s. GPU-density power (5 kW+ per node)
We rack it, power it, and support it on-site. Redundant network core, SLA-backed contracts, and hosting that knows what's in the box down to the BGA level.
Greg Sky retweeted
GLM-Legal uncensored:
Introducing GLM-Legal
Built on GLM-5.3-Flash, GLM-Legal matches frontier models on mainstream legal benchmarks.
Designed for real legal workflows: contract review, evidence analysis, compliance, and more.
Tech Blog: z.ai/blog/glm-legal
Apply for API Access: z.ai/consultation/legal
Try AutoClaw: autoclaw.z.ai
Only 1 of 2 in the world 4090 "Ti" with D8BZF memory - faster than the stock memory can ever hit on overclocks.
Work was done to a binned 4090 core on a MSI SUPRIM hybrid gpu.
Custom BGA Rework done on NVIDIA GPU's gpulab.net
Who wants an 8x GPU, PCIe Gen5, EPYC 9004/9005, DDR5 server custom built and validated for their local AI workflows?
Supermicro AS-4125GS-TNRT2 platform, configured to your workload, burned in and tested before it ships.
10-12 week leadtimes. DMs open or gpulab.net
Version 5 CMP Unlocker has been released.
ECC support enabled.
github.com/amoghmunikote/cmp…
These 170HX cards are one-NVLink unlock away from becoming more compelling than 40G A100's
