Hacker News Reader

Comment by davidy123 | Hacker News Reader

This is such a good idea. You can get "old" Thinkpads with NVidia GPUs and 16+ GB RAM for a very decent price, and use them for home server purposes like encoding and object detection. Much smaller and lower power and heat than a desktop, yet much more powerful than a Pi, Jetson, &c. There are even workstation class models with Xeons and ECC RAM.

View on HN

> Much smaller and lower power and heat than a desktop, yet much more powerful than a Pi, Jetson

This brings me to my biggest complain about (all kinds of) Pi(s): while running a Pi 24/7 CAN save some power compare to a 24/7 running T430, the T430 can do Wake Up On LAN, which means you can just turn it off when it's not in-use and turn it back on remotely based on demand. Many (if not all) Pi(s) can not do that, you have to run it 24/7.

Of course there are Wake Up On LAN power switches, but they are hacky and require extra care to setup properly. And if you did it wrong, you might risk losing data or damage the storage on the Pi.

> "old" Thinkpads with NVidia GPU

> with NVidia GPU

*Linux*...

What about Linux? Are you saying running an nvidia driver is an extra setup step? Or something else entirely?

It’s gotten a lot easier to run NVidia GPUs on Linux since Linus first gave Nvidia the finger all those years ago. It’s usually a matter of running a single package manager command.

I’m old enough to remember when it was the NVidia drivers that were great (because it shared most of its code with the windows driver so supported “nice things”), and the ATI driver (fglrx) was a total dumpster fire that never worked right.

Correct me if I’m wrong, but I don’t think the Nvidia Linux drivers were ever bad, they just weren’t open-source and Linus Torvalds didn’t like that the open-source nouveau drivers sucked.

No idea what you're talking about. From what I've read "Optimus" (switching between onboard graphics and GPU on demand) can be a pain, but otherwise NVidia's driver, while proprietary, is very good, and the open source driver is pretty good.

Which, if you're using it as a home server then setting it to nvidia-only mode seems the easiest route forwards instead of futzing with Optimus graphics stuff.

It's not Nvidia + Linux that sucks, it's Nvidia + Linux on a desktop. CUDA thrives on Linux. Most AI SaaS runs on Nvidia + Linux (if it isn't on a TPU or alike)

> It's not Nvidia + Linux that sucks, it's Nvidia + Linux on a desktop.

It's not just Nvidia + Linux on a desktop, its might be an *old mobile Nvidia + Linux on an old laptop* which results in a lot of restrictions & bugs nowadays:[0]

[0] https://www.thinkwiki.org/wiki/NVIDIA_Quadro_2000M

[1] https://forums.debian.net/viewtopic.php?t=143606

Well, recent Nvidia GPUs also have (had) issues with desktop, the Wayland GBM/EGLStreams debacle comes to mind. See --my-next-gpu-wont-be-nvidia

Reply on news.ycombinator.com