noc.social is part of the decentralized social network powered by Mastodon.
This instance is focused on technology, networking, linux, privacy, security, infosec, engineering, but open to anyone. Civil discourse, polite and open. Managed by the noc.org / trunc.org team.

Administered by:

Server stats:

674
active users

Learn more

#gaming #pcgaming #nvidia

Ich muss Thermaltake's ToughPower Serie wirklich loben😃

Ich habe meine 5090 mal mithilfe von Silent Hill 2 Remake auf 550+ getrieben um das 600W Kabel von TT einem Stresstest zu unterziehen.

Das Kabel wurde natürlich etwas warm, aber nicht heiß.

Fiepen tut meine Karte übrigens weiterhin nicht.

🌘 在NixOS WSL上運行Nvidia和Ollama:全天候啟動您的遊戲PC上的LLM
➤ 在你的遊戲PC上輕鬆運行大型語言模型
yomaq.github.io/posts/nvidia-o
這篇文章詳細介紹了作者如何在遊戲PC上使用NixOS和WSL環境,成功配置Nvidia GPU和Ollama,實現LLM模型的持續運行。作者解決了vram鎖定、WSL自動關閉以及NixOS對Nvidia的初始支持不足等問題,並分享了詳細的配置步驟,包括保持WSL運行的設置、NixOS的安裝配置、Nvidia Container Toolkit的設置以及Ollama Docker容器的配置,並整合了Tailscale以簡化網絡連接。
+ 這篇文章真的很有用,我一直想試試本地LLM,但設定起來太麻煩了。這個方法看起來更可行!
+ NixOS看起來很強大,但是學習曲線有點陡峭。不過為了能在本地運行LLM,我
#NixOS #WSL #Nvidia #Ollama #LLM #Docker

Nvidia on Nixos WSL - Ollama up 24/7 on your gaming PCyomaq

Upgrading the "slim" base container for our #Deepstream app would be a lot simpler if #Nvidia didn't keep restructuring how they package these things.

All they need to do is package this stuff in clean Debian packages and it would be trivial to upgrade, but no, instead they use a mix of packages (including packages that unpack a local repo they then install packages from!), tar balls, scripts and then they move & delete files randomly too. It's a mess.

Optcast is an implementation of a reduction server written in #Rust, specifically designed for enhanced performance in distributed machine learning environments utilizing the #NVIDIA NCCL library for the AllReduce collective operation. Although still in its prototype stage, Optcast has achieved a 50% speed improvement in NCCL's AllReduce operation under certain conditions.
cloud.google.com/blog/topics/d
Question is there any better way to share a url from #eww to here w/o bookmarking it 1st and then copying it?

Optimize training performance with Reduction Server on Vertex AI | Google Cloud BlogGoogle Cloud Blog