5 ms·

I'm in the market for a new desktop (it's been nearly nine years since I purchased my gaming rig). My current plan is to leave Windows behind finally. I'll be

I'm in the market for a new desktop (it's been nearly nine years since I purchased my gaming rig). My current plan is to leave Windows behind finally. I'll be getting a Linux rig of some sort. And it won't just be for gaming, it will be for running and training local LLMs, which is far more brutal on the hardware than any games I've seen; I can still play most titles on my 1440p screen with my current "potato." I've been running Microsoft operating systems for decades: Windows, and DOS before that since DOS 3.3. With all the attention sinks stuck into the new Windows (pops on the menu, pops from the dock, ads everywhere on my machine) it's time to leave it. The two viable alternatives are Mac OS and Linux. Work has me on Mac, but the hardware I want to run... Linux is the right choice. I'll game on the system I have. I have to wonder how many PC gamers will follow suit as they get older. Most of the systems I've been pricing out for Linux desktops are AMD Threadripper builds. Intel Xeons would be better for AI workloads, but they're way more expensive.

skellyclock 3y ago
You might need a nvidia GPU for your LLM training case tbh.
qudat 3y ago
Steam and wine has definitely pushed Linux into “good enough” territory for Linux gaming. I, like you, switched from windows to Linux for gaming. This is the first year I haven’t owned a windows machine. End of an era. I don’t miss it at all.
- nicce 3y ago
  Basically the state is getting to be that almost every game exluding some, which require specific anti-cheat engines, will work. In some cases it is just a matter of enabling EAC linux support by game publishers. Future is here!
  - Arubis 3y ago
    That’s a pretty big “just”, tbh. It would absolutely require the sort of proprietary kernel drivers most distros (understandably!) consider anathema, and I’m not sure game publishers want to shoulder that burden entirely themselves.
    - nicce 3y ago
      In one sense, the problem is "centralized". Instead that there is a need to fix something special in many games, we can make many games working by fixing selected few entities. EAC has already support for linux, so it should not be impossible.
  - Shekelphile 3y ago
    There's definitely a lot games that don't run as well as they do on windows despite being generally playable on linux. It's nice to have a dual boot for games like that even if you play the vast majority of games on the linux side.
dimaor 3y ago
Exactly what I did 6 months ago, abandoning windows was an amazing experience. Some games don't work though, but it's a price I am willing to pay.
solardev 3y ago
Ever considered cloud gaming? I built dozens of gaming desktops as a kid and young adult, but lately the time demands of adulthood means it's hard to keep up with the latest trends. Like you I got sick of fighting Windows just to launch Steam. But I also had numerous bad experiences with desktop Linux, and with not wanting to deal with Proton and Wine. When I want to relax I just want to get in a quick game for an hour or two, not go through a ton of driver updates etc. To that end, GeForce Now has been... no pun intended... game changing. It works miraculously well, and has almost top of the line hardware. Paired with my ultrawide QHD screen, most games run beautifully. Shooters are out due to the input lag (small but definitely noticeable) but all the other genres feel fantastic. No fan noise, no heat, no giant tower taking up space, no worrying about upgrades... It's an amazing service at a terrific price point. I have a Mac but just plug in a Windows keyboard and mouse and it works flawlessly. Same with an Xbox controller. People pooh poohed cloud gaming for so long, but these days it's so easy and seamless I'm honestly worried about the future of the gaming PC market. Nvidia seems well on their way to providing data center and AI cards and just renting games as a service. (Honestly, that's fine with me...).
- thejazzman 3y ago
  (I'm not suggesting anyone be like me, just chiming in!) I feel like a combination of a few comments here. Itching to put together something that can bring justice to Zelda and tinker with AI Apparently there is a way to stream games from your own machine similar to GeForce now. It's built into their GPUs -- https://arstechnica.com/gaming/2023/04/nvidias-gamestream-is-dead-sunshine-and-moonlight-are-better-replacements/ https://arstechnica.com/gaming/2023/04/nvidias-gamestream-is...
  - solardev 3y ago
    Yeah, you've been able to stream from your hardware doe a while now. It never works quite as well as just plugging in a HDMI cord though. There are also services like Shadow that just rent you a GPU enabled Windows box, and you can run whatever emulator on there you want. It's cheaper than renting GPU hours from the enterprise providers (who mostly target AI workloads, I guess?)
Arubis 3y ago
If you’re going to be investing much time and energy into AI/ML on your GPU, I urge you to consider going the nvidia route rather than an AMD GPU. (CPU-wise, either AMD or Intel is just fine.) Using ROCm instead of CUDA may be possible, but it’s unloved, unsupported on consumer hardware, doesn’t even build with PyTorch at its current minor version, and generally a frustrating time-suck. The proprietary nvidia driver situation is unfortunate, and if you’re sufficiently driven off by it, ROCm can work, sometimes, eventually. But do consider the effort and time it’ll cost.
- BLKNSLVR 3y ago
  I think Arubis didn't push the point hard enough. Nvidia is a long way ahead of AMD in AI/ML hardware capability, from my very limited experience. I got a Stable Diffusion thing up and running locally on a relatively old AMD RX 580 with 8GB of memory, and default settings of 512x512 with 10 (I think) passes, and it takes about 10 minutes to generate an image. On a similarly aged Nvidia 1060 with 6GB of memory that a colleague of mine uses for their local Stable Diffusion instance, it takes about 20 seconds to generate an image. I don't know how apples and oranges this comparison is, my OS is Linux and his is Windows, for example, but 10 minutes versus 20 seconds is, for me, a massively disappointing difference. I'm sure the gap has decreased in the intervening 6-7 years, but I don't know how much - such that I'd be opting for a second hand Nvidia over a new AMD (and I philosophically prefer AMD). This may be a worthwhile read if you're looking to squeeze the most performance: https://timdettmers.com/2023/01/30/which-gpu-for-deep-learning/ https://timdettmers.com/2023/01/30/which-gpu-for-deep-learni...
  - Aeolun 3y ago
    That doesn’t seem like a difference caused by the chip, but by whatever integration is used.
  - nicopappl 3y ago
    To clarify this. On my RX5700 XT, with ROCm, Stable Diffusion generates images on the default sampler at 512x512 20 passes in about 2 seconds. It's definitively slower than a similar NVIDIA card, but I've no concrete numbers. It's still a pain in other ways. AMD does not support ROCm on consumer-grade GPUs, it only "happens to work". I had to explicitly install the 3.10 python interpreter and use an older torch version, because seemingly, the newer ROCm releases do not support the RX5700 XT. Also ROCm only works on Linux.
Veraticus 3y ago
I left for NixOS instead of upgrading to Windows 11. Honestly it can play anything I can throw at it. Including FFXIV which has zero official Linux support. And it might play even better than it does on Windows. I wish I had migrated way earlier.
franknord23 3y ago
Don't buy nvidia. The driver situation is terrible.
jwrallie 3y ago
Microsoft is losing this particular market, but they don't seem to care. I would pay good money for a Windows version that had a more professional and consistent interface (Win 2K comes to mind, but Win 7 interface was still OK) no advertisement inside the OS, no telemetry, automatic update notification but manual installation, no unexplained HDD usage when idle that goes away as soon as I open the system monitor, WSL support. Closest thing is LTSC but it is discouraged and hard to legally buy. Linux can do all of this. Compatibility with games is good enough and better in some edge cases. KVM performance running a VM is good enough for running MS Office for work, and still a better experience than in bare metal as you can limit core usage and the computer will run cooler and use less power.

skellyclock 3y ago

You might need a nvidia GPU for your LLM training case tbh.

qudat 3y ago

Steam and wine has definitely pushed Linux into “good enough” territory for Linux gaming. I, like you, switched from windows to Linux for gaming. This is the first year I haven’t owned a windows machine. End of an era. I don’t miss it at all.

nicce 3y ago
Basically the state is getting to be that almost every game exluding some, which require specific anti-cheat engines, will work. In some cases it is just a matter of enabling EAC linux support by game publishers. Future is here!
- Arubis 3y ago
  That’s a pretty big “just”, tbh. It would absolutely require the sort of proprietary kernel drivers most distros (understandably!) consider anathema, and I’m not sure game publishers want to shoulder that burden entirely themselves.
  - nicce 3y ago
    In one sense, the problem is "centralized". Instead that there is a need to fix something special in many games, we can make many games working by fixing selected few entities. EAC has already support for linux, so it should not be impossible.
- Shekelphile 3y ago
  There's definitely a lot games that don't run as well as they do on windows despite being generally playable on linux. It's nice to have a dual boot for games like that even if you play the vast majority of games on the linux side.

dimaor 3y ago

Exactly what I did 6 months ago, abandoning windows was an amazing experience. Some games don't work though, but it's a price I am willing to pay.

solardev 3y ago

Ever considered cloud gaming? I built dozens of gaming desktops as a kid and young adult, but lately the time demands of adulthood means it's hard to keep up with the latest trends. Like you I got sick of fighting Windows just to launch Steam. But I also had numerous bad experiences with desktop Linux, and with not wanting to deal with Proton and Wine. When I want to relax I just want to get in a quick game for an hour or two, not go through a ton of driver updates etc. To that end, GeForce Now has been... no pun intended... game changing. It works miraculously well, and has almost top of the line hardware. Paired with my ultrawide QHD screen, most games run beautifully. Shooters are out due to the input lag (small but definitely noticeable) but all the other genres feel fantastic. No fan noise, no heat, no giant tower taking up space, no worrying about upgrades... It's an amazing service at a terrific price point. I have a Mac but just plug in a Windows keyboard and mouse and it works flawlessly. Same with an Xbox controller. People pooh poohed cloud gaming for so long, but these days it's so easy and seamless I'm honestly worried about the future of the gaming PC market. Nvidia seems well on their way to providing data center and AI cards and just renting games as a service. (Honestly, that's fine with me...).

thejazzman 3y ago
(I'm not suggesting anyone be like me, just chiming in!) I feel like a combination of a few comments here. Itching to put together something that can bring justice to Zelda and tinker with AI Apparently there is a way to stream games from your own machine similar to GeForce now. It's built into their GPUs -- https://arstechnica.com/gaming/2023/04/nvidias-gamestream-is-dead-sunshine-and-moonlight-are-better-replacements/ https://arstechnica.com/gaming/2023/04/nvidias-gamestream-is...
- solardev 3y ago
  Yeah, you've been able to stream from your hardware doe a while now. It never works quite as well as just plugging in a HDMI cord though. There are also services like Shadow that just rent you a GPU enabled Windows box, and you can run whatever emulator on there you want. It's cheaper than renting GPU hours from the enterprise providers (who mostly target AI workloads, I guess?)

Arubis 3y ago

If you’re going to be investing much time and energy into AI/ML on your GPU, I urge you to consider going the nvidia route rather than an AMD GPU. (CPU-wise, either AMD or Intel is just fine.) Using ROCm instead of CUDA may be possible, but it’s unloved, unsupported on consumer hardware, doesn’t even build with PyTorch at its current minor version, and generally a frustrating time-suck. The proprietary nvidia driver situation is unfortunate, and if you’re sufficiently driven off by it, ROCm can work, sometimes, eventually. But do consider the effort and time it’ll cost.

BLKNSLVR 3y ago
I think Arubis didn't push the point hard enough. Nvidia is a long way ahead of AMD in AI/ML hardware capability, from my very limited experience. I got a Stable Diffusion thing up and running locally on a relatively old AMD RX 580 with 8GB of memory, and default settings of 512x512 with 10 (I think) passes, and it takes about 10 minutes to generate an image. On a similarly aged Nvidia 1060 with 6GB of memory that a colleague of mine uses for their local Stable Diffusion instance, it takes about 20 seconds to generate an image. I don't know how apples and oranges this comparison is, my OS is Linux and his is Windows, for example, but 10 minutes versus 20 seconds is, for me, a massively disappointing difference. I'm sure the gap has decreased in the intervening 6-7 years, but I don't know how much - such that I'd be opting for a second hand Nvidia over a new AMD (and I philosophically prefer AMD). This may be a worthwhile read if you're looking to squeeze the most performance: https://timdettmers.com/2023/01/30/which-gpu-for-deep-learning/ https://timdettmers.com/2023/01/30/which-gpu-for-deep-learni...
- Aeolun 3y ago
  That doesn’t seem like a difference caused by the chip, but by whatever integration is used.
- nicopappl 3y ago
  To clarify this. On my RX5700 XT, with ROCm, Stable Diffusion generates images on the default sampler at 512x512 20 passes in about 2 seconds. It's definitively slower than a similar NVIDIA card, but I've no concrete numbers. It's still a pain in other ways. AMD does not support ROCm on consumer-grade GPUs, it only "happens to work". I had to explicitly install the 3.10 python interpreter and use an older torch version, because seemingly, the newer ROCm releases do not support the RX5700 XT. Also ROCm only works on Linux.

Veraticus 3y ago

I left for NixOS instead of upgrading to Windows 11. Honestly it can play anything I can throw at it. Including FFXIV which has zero official Linux support. And it might play even better than it does on Windows. I wish I had migrated way earlier.

franknord23 3y ago

Don't buy nvidia. The driver situation is terrible.

jwrallie 3y ago

Microsoft is losing this particular market, but they don't seem to care. I would pay good money for a Windows version that had a more professional and consistent interface (Win 2K comes to mind, but Win 7 interface was still OK) no advertisement inside the OS, no telemetry, automatic update notification but manual installation, no unexplained HDD usage when idle that goes away as soon as I open the system monitor, WSL support. Closest thing is LTSC but it is discouraged and hard to legally buy. Linux can do all of this. Compatibility with games is good enough and better in some edge cases. KVM performance running a VM is good enough for running MS Office for work, and still a better experience than in bare metal as you can limit core usage and the computer will run cooler and use less power.