5 ms·

I'm glad the AMD GPU option still exists, I don't have great experience with NVIDIA on Linux. The rest of the upgrades, like the new top cover and keyboard, are

I'm glad the AMD GPU option still exists, I don't have great experience with NVIDIA on Linux. The rest of the upgrades, like the new top cover and keyboard, are very welcome

rauli_ 1y ago
It's so weird to hear people who have problems with NVIDIA GPUs on Linux, because for me it's always been the opposite. I have had problems with AMD but never with NVIDIA.
- lotharcable 1y ago
  Things have changed a lot since Steam deck. Especially in the last 3 or 4 years. Mobile users suffer more problems then people with dedicated desktop GPUs, but it still gotten a lot better. The one thing to be careful about AMD GPUs is that for most GPU OEMs AMD is just a after thought. So they get sub-par QA and heatsinks compared to their more popular Nvidia models. It is best to go with card makers that only sell AMD GPUs, like Sapphire, PowerColor, and XFX. I am partial to Sapphire.
  - seanw444 1y ago
    Had good experiences with XFX so far as well.
- aidenn0 1y ago
  When the AMD driver was named "radeon" nvidia was better, but since "amdgpu" came out things have flipped.
  - TomLisankie 1y ago
    Is this because the driver itself has changed in its operation or just from the name change breaking lots of code that referenced the "radeon" name?
    - rnhmjoj 1y ago
      It's a completely different driver for a different architecture. The biggest reason it works so much better is that it's open source (with some blobs, of course) and part of the mainline kernel, unlike its predecessor which was developed downstream and fully proprietary.
      - account42 1y ago
        amdgpu replaced both the in-kernel open-source "radeon" driver, which was already open source, and the proprietary "fglrx" driver. But the user-space portions are probably more significant for performance than the kernel drivers. Here we have: - r300 and r600 (open source OpenGL backend for older hardware, sits on top of the radeon kernel driver, not much development happening) - radeonsi (open source OpenGL backend for newer hardware, sits on top of either the radeon or amdgpu kernel drivers depending on hardware version and kernel configuration) - fglrx (closed source OpenGL driver on top of the fglrx kernel driver, both obsolete now) - radv (open source Vulkan driver on top of amdgpu) - amgpu-pro (closed source Vulkan driver on top of amdgpu) - not sure if there is also still a proprietary OpenGL driver but if there is no one should care since radeonsi works well enough - amdvlk (open source dumps of amdgpu-pro without proprietary shader compiler on top of amdgpu) Then you have different shader compilers which also significantly affect both shader compile time and runtime performance: - internal compiler used by r600 - LLVM (used by radeonsi and amdvlk) - ACO (used by radv and possibly radeonsi these days) - AMD's proprietary compiler (used by fglrx and amdgpu-pro) And for X.org you also have different display drivers (fglrx, radeon, modesetting).
    - chao- 1y ago
      They were referencing a time, in the middle of the 2010's, when "amdgpu" was released. It is a completely rewritten, different driver, and is mostly open source [0]. Before that, the driver was named "radeon" and it was very shaky. I can speak to this personally. I had desktop Linux systems with both AMD/ATI and Nvidia GPUs, and while there were some issues with Nvidia, the AMD/ATI drivers gave me nightmares. Once the rewritten "amdgpu" driver came out, things got much better. The first few cards created after that (IIRC the Polaris GPUs, RX 400's), the situation reversed. I still have had occasional issues with various Nvidia cards (normally driver updates breaking things), but for almost a decade now, I have not had issues with AMD GPUs under Linux. [0] Except for pro features while using workstation cards. You need to use a proprietary driver for those, but even those share a lot of code with the open source driver.
- bodge5000 1y ago
  Maybe things have improved, or support was just never that good for older NVIDIA GPU's (for reference, last time I used Nvidia on Linux I was running Fedora on a Thinkpad P50, which I think has a Quadro M1000M gpu), but it'd be a costly experiment to find
- Kudos 1y ago
  Nvidia driver 580 (latest stable, but not lts) was just released and is widely freezing people's systems right now.
- kcb 1y ago
  I have a 5070 ti running Kubuntu 25.04 and it's a mess. Animations repeating, half the desktop disappears when waking from sleep, HDMI audio cuts out... I swapped to a 7900xt and it is absolutely flawless.
- extraisland 1y ago
  AMD cards are plug and play for 99% of cases with Linux now. Everything just works out of the box. The only issues you may run into if you distro doesn't include the firmware. e.g. This was the case with Debian 11 and you had to enable the non-free repo. The only other problem you can conceivably have is card isn't supported by the kernel because it is too new. This can be fixed by upgrading the kernel. In Debian you can use a backports kernel, I am sure there are similar options in other distros. When I was using my old 1080Ti, I had constant issues with the NVIDIA drivers. Acceleration didn't work on the second screen sometimes. There was some magic setting that would unset itself.
- ankurdhama 1y ago
  The problem is not NVIDIA GPU, it is laptops that have iGPU (amd or intel) and Nvidia dgpu. In such a configuration the experience is really really bad in both X11 and wayland.
  - account42 1y ago
    It's both. With Nvidia you still need a proprietary driver for anything close to full performance which causes all kinds of issues.
lotharcable 1y ago
With the advent of Steam deck and Valve putting time and effort into AMD GPU drivers the AMD GPU is really the best option for Linux when it comes to general desktop stuff and gaming. The days of Nvidia proprietary drivers being a safe bet is long gone. Especially for any sort of Wayland desktop, but it still applies to X11. Intel drivers should be good as well, since they use the same Mesa code base. With the ROCm stuff no longer depending on AMD Pro then there is not going to be any reason to step away from the default GPU drivers provided by your distro, provided they are relatively new. While I am sure that there are still going to be professional-grade proprietary apps that recommend Nvidia... for most of us the only reason to actually go and choose Nvidia on Linux is because of CUDA. And, personally, I would rather lease time on the cloud or have a second GPU work horse PC separate from my desktop for that. Unfortunately Nvidia is, by far, the most popular option for Windows users. Over 4:1 ratio according to Steam statistics. So most new Linux users are still going to have to suffer through dealing with their GPU drivers.
- account42 1y ago
  > Intel drivers should be good as well, since they use the same Mesa code base. They use the same front end but that says very little about the quality of the overall driver. Performance is mostly determined by the shader compiler and other hardware-specific parts which obviously differ between Intel and AMD.
jvanderbot 1y ago
It'd be nice if I could upgrade my old Framework into this spec. Infinitely upgradeable is nice on a per product line basis. But new product lines still lead to obsolescence and in this case regret.
- chpatrick 1y ago
  You can upgrade your old Framework 16 to this. Framework 13 wouldn't work anyway because it's a different chassis.
  - jvanderbot 1y ago
    That's exactly my point. When I bought the 13 I figured there would be these kinds of upgrades down the road. You're right to say that was stupid and it was. And next there will be the framework 17, a 16 that's not backwards compatible or something?
    - adgjlsfhk1 1y ago
      There are a bunch of upgrades that have come out for the 13, new gens of intel and AMD boards, new displays, new keyboard options, etc.
    - bcrosby95 1y ago
      Like the other commenter mentioned, you can upgrade the 13". It's pretty common for laptops to come in two classes: smaller with a focus on portability and larger with a focus on performance. There's a reason why a lot of us sat on the sidelines and were looking forward to the 16". There is no slippery slope here, the differentiated product lines 100% make sense. Edit: there is another class I could see making sense - desktop replacement. Those chassis' tend to be pretty chunky because they put desktop parts into a laptop. Think 10 lb laptop with a battery that lasts 20-30 minutes. But I'm not sure if the market is large enough for them to pursue it.
    - ffsm8 1y ago
      As framework doesn't produce their own hardware, they're also forced to live with the reality that generations are also bound by the whims of the producers. E.g sockets and chipsets change and will force incompatible changes, no matter how much framework would like to keep things stable.
bigstrat2003 1y ago
I've never had anything but positive experience with Nvidia on Linux (which I've been using for 5 years or so now). That said, I'm on a desktop and not a laptop, so the hardware isn't the same. My experience might not be representative of what laptop users see.
- danudey 1y ago
  On a laptop with multiple GPUs (Intel and nvidia Quadro) running Ubuntu and Wayland, trying to get the nvidia card working has been a nightmare. Until a recent reinstall, I couldn't load the nvidia driver or I wouldn't be able to log in to my system (graphically, I mean). If anything changed on my system to remove the blacklist I had for those modules I'd have to spend an hour trying to figure out what changed so that I could get back to work. Now that I have it working I see random glitches here and there that I can't pin down. Some Electron apps I have to turn off GPU acceleration or they won't get any windows showing up - they launch, the process exists, they're in the dock as active, but the window doesn't appear at all. Getting a new laptop from work to replace this one and I'm really hoping it won't have nvidia hardware - or at least, if it does I can disable it and the Intel GPU will work fine also.
drcongo 1y ago
I'm one month into owning a GMKTec Evo-X2 with the new mad AMD gpu in it and so far I've managed to get Ollama running. ROCm doesn't officially support the GPU and everything seems to be hacky workaround on top of hacky workaround. Starting to wish I'd just waited for NVIDIA to actually release the project digits thing.
- Tepix 1y ago
  So, how did you get it running? I‘m about to get a Bosgame M5 (same CPU).
  - drcongo 1y ago
    Installing the ROCm nightlies from TheRock and upgrading the linux kernel to 6.16. However the kernel upgrade breaks a ton of other things, notably dkms. I've been trying for weeks to get ComfyUI to not explode without success.
  - nicolaslem 1y ago
    I would recommend not wasting time with ollama, https://github.com/containers/ramalama https://github.com/containers/ramalama just works on that chip.
fooker 1y ago
FYI All current and future Nvidia drivers are open source, since blackwell.
- kcb 1y ago
  Kernel driver only.
  - cromka 1y ago
    Is AMD different, do they OpenSource firmware as well?
    - ItsHarper 1y ago
      They have an open source userspace component.
- jonkoops 1y ago
  But they are not in the actual Linux kernel, and are a separate module that does not adhere to Linux kernel code conventions. It also has no user-space driver that isn't NVIDIA's proprietary driver. Anything further open-source such as NVK is not being worked on by NVIDIA, but by other 3rd parties. Compared to AMD and Intel, NVIDIA is very much not an 'out of the box', or stable experience.
- mdaniel 1y ago
  Saying "and future" is like taunting fate Anyway, in case someone was interested it seems the code itself is cited as MIT, however it has a "when it becomes a Linux .ko it becomes GPLv2" clause https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.76.05/COPYING https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.7... and they do go out of their way to say "lol, needs binary blobs" https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.76.05/README.md?plain=1#L18-L19 https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.7... That XFree86.run has always struck me as "you're gonna what*?"
  - account42 1y ago
ethersteeds 1y ago
I came to say the same thing. In the late 2010s I ran Linux on a work-issued Lenovo P50 with a Nvidia Quadro M2000M. It was such a miserable experience that I swore to never own another NVIDIA product again. Given both Framework and NVIDIA's checkered histories around Linux driver support, I see no reason to revisit that, but it is interesting to see the voices in this thread with positive NVIDIA experience.
- bodge5000 1y ago
  I said in another comment but exact same experience here (though about a decade later), same laptop and gpu. It's a shame because the P50 is so nice otherwise. Maybe its just a problem with older Nvidia gpu's, but its not a gamble I want to take

I'm glad the AMD GPU option still exists, I don't have great experience with NVIDIA on Linux. The rest of the upgrades, like the new top cover and keyboard, are very welcome

rauli_ 1y ago

It's so weird to hear people who have problems with NVIDIA GPUs on Linux, because for me it's always been the opposite. I have had problems with AMD but never with NVIDIA.

lotharcable 1y ago
Things have changed a lot since Steam deck. Especially in the last 3 or 4 years. Mobile users suffer more problems then people with dedicated desktop GPUs, but it still gotten a lot better. The one thing to be careful about AMD GPUs is that for most GPU OEMs AMD is just a after thought. So they get sub-par QA and heatsinks compared to their more popular Nvidia models. It is best to go with card makers that only sell AMD GPUs, like Sapphire, PowerColor, and XFX. I am partial to Sapphire.
- seanw444 1y ago
  Had good experiences with XFX so far as well.
aidenn0 1y ago
When the AMD driver was named "radeon" nvidia was better, but since "amdgpu" came out things have flipped.
- TomLisankie 1y ago
  Is this because the driver itself has changed in its operation or just from the name change breaking lots of code that referenced the "radeon" name?
  - rnhmjoj 1y ago
    It's a completely different driver for a different architecture. The biggest reason it works so much better is that it's open source (with some blobs, of course) and part of the mainline kernel, unlike its predecessor which was developed downstream and fully proprietary.
    - account42 1y ago
      amdgpu replaced both the in-kernel open-source "radeon" driver, which was already open source, and the proprietary "fglrx" driver. But the user-space portions are probably more significant for performance than the kernel drivers. Here we have: - r300 and r600 (open source OpenGL backend for older hardware, sits on top of the radeon kernel driver, not much development happening) - radeonsi (open source OpenGL backend for newer hardware, sits on top of either the radeon or amdgpu kernel drivers depending on hardware version and kernel configuration) - fglrx (closed source OpenGL driver on top of the fglrx kernel driver, both obsolete now) - radv (open source Vulkan driver on top of amdgpu) - amgpu-pro (closed source Vulkan driver on top of amdgpu) - not sure if there is also still a proprietary OpenGL driver but if there is no one should care since radeonsi works well enough - amdvlk (open source dumps of amdgpu-pro without proprietary shader compiler on top of amdgpu) Then you have different shader compilers which also significantly affect both shader compile time and runtime performance: - internal compiler used by r600 - LLVM (used by radeonsi and amdvlk) - ACO (used by radv and possibly radeonsi these days) - AMD's proprietary compiler (used by fglrx and amdgpu-pro) And for X.org you also have different display drivers (fglrx, radeon, modesetting).
  - chao- 1y ago
    They were referencing a time, in the middle of the 2010's, when "amdgpu" was released. It is a completely rewritten, different driver, and is mostly open source [0]. Before that, the driver was named "radeon" and it was very shaky. I can speak to this personally. I had desktop Linux systems with both AMD/ATI and Nvidia GPUs, and while there were some issues with Nvidia, the AMD/ATI drivers gave me nightmares. Once the rewritten "amdgpu" driver came out, things got much better. The first few cards created after that (IIRC the Polaris GPUs, RX 400's), the situation reversed. I still have had occasional issues with various Nvidia cards (normally driver updates breaking things), but for almost a decade now, I have not had issues with AMD GPUs under Linux. [0] Except for pro features while using workstation cards. You need to use a proprietary driver for those, but even those share a lot of code with the open source driver.
bodge5000 1y ago
Maybe things have improved, or support was just never that good for older NVIDIA GPU's (for reference, last time I used Nvidia on Linux I was running Fedora on a Thinkpad P50, which I think has a Quadro M1000M gpu), but it'd be a costly experiment to find
Kudos 1y ago
Nvidia driver 580 (latest stable, but not lts) was just released and is widely freezing people's systems right now.
kcb 1y ago
I have a 5070 ti running Kubuntu 25.04 and it's a mess. Animations repeating, half the desktop disappears when waking from sleep, HDMI audio cuts out... I swapped to a 7900xt and it is absolutely flawless.
extraisland 1y ago
AMD cards are plug and play for 99% of cases with Linux now. Everything just works out of the box. The only issues you may run into if you distro doesn't include the firmware. e.g. This was the case with Debian 11 and you had to enable the non-free repo. The only other problem you can conceivably have is card isn't supported by the kernel because it is too new. This can be fixed by upgrading the kernel. In Debian you can use a backports kernel, I am sure there are similar options in other distros. When I was using my old 1080Ti, I had constant issues with the NVIDIA drivers. Acceleration didn't work on the second screen sometimes. There was some magic setting that would unset itself.
ankurdhama 1y ago
The problem is not NVIDIA GPU, it is laptops that have iGPU (amd or intel) and Nvidia dgpu. In such a configuration the experience is really really bad in both X11 and wayland.
- account42 1y ago
  It's both. With Nvidia you still need a proprietary driver for anything close to full performance which causes all kinds of issues.

lotharcable 1y ago

With the advent of Steam deck and Valve putting time and effort into AMD GPU drivers the AMD GPU is really the best option for Linux when it comes to general desktop stuff and gaming. The days of Nvidia proprietary drivers being a safe bet is long gone. Especially for any sort of Wayland desktop, but it still applies to X11. Intel drivers should be good as well, since they use the same Mesa code base. With the ROCm stuff no longer depending on AMD Pro then there is not going to be any reason to step away from the default GPU drivers provided by your distro, provided they are relatively new. While I am sure that there are still going to be professional-grade proprietary apps that recommend Nvidia... for most of us the only reason to actually go and choose Nvidia on Linux is because of CUDA. And, personally, I would rather lease time on the cloud or have a second GPU work horse PC separate from my desktop for that. Unfortunately Nvidia is, by far, the most popular option for Windows users. Over 4:1 ratio according to Steam statistics. So most new Linux users are still going to have to suffer through dealing with their GPU drivers.

account42 1y ago
> Intel drivers should be good as well, since they use the same Mesa code base. They use the same front end but that says very little about the quality of the overall driver. Performance is mostly determined by the shader compiler and other hardware-specific parts which obviously differ between Intel and AMD.

jvanderbot 1y ago

It'd be nice if I could upgrade my old Framework into this spec. Infinitely upgradeable is nice on a per product line basis. But new product lines still lead to obsolescence and in this case regret.

chpatrick 1y ago
You can upgrade your old Framework 16 to this. Framework 13 wouldn't work anyway because it's a different chassis.
- jvanderbot 1y ago
  That's exactly my point. When I bought the 13 I figured there would be these kinds of upgrades down the road. You're right to say that was stupid and it was. And next there will be the framework 17, a 16 that's not backwards compatible or something?
  - adgjlsfhk1 1y ago
    There are a bunch of upgrades that have come out for the 13, new gens of intel and AMD boards, new displays, new keyboard options, etc.
  - bcrosby95 1y ago
    Like the other commenter mentioned, you can upgrade the 13". It's pretty common for laptops to come in two classes: smaller with a focus on portability and larger with a focus on performance. There's a reason why a lot of us sat on the sidelines and were looking forward to the 16". There is no slippery slope here, the differentiated product lines 100% make sense. Edit: there is another class I could see making sense - desktop replacement. Those chassis' tend to be pretty chunky because they put desktop parts into a laptop. Think 10 lb laptop with a battery that lasts 20-30 minutes. But I'm not sure if the market is large enough for them to pursue it.
  - ffsm8 1y ago
    As framework doesn't produce their own hardware, they're also forced to live with the reality that generations are also bound by the whims of the producers. E.g sockets and chipsets change and will force incompatible changes, no matter how much framework would like to keep things stable.

bigstrat2003 1y ago

I've never had anything but positive experience with Nvidia on Linux (which I've been using for 5 years or so now). That said, I'm on a desktop and not a laptop, so the hardware isn't the same. My experience might not be representative of what laptop users see.

danudey 1y ago
On a laptop with multiple GPUs (Intel and nvidia Quadro) running Ubuntu and Wayland, trying to get the nvidia card working has been a nightmare. Until a recent reinstall, I couldn't load the nvidia driver or I wouldn't be able to log in to my system (graphically, I mean). If anything changed on my system to remove the blacklist I had for those modules I'd have to spend an hour trying to figure out what changed so that I could get back to work. Now that I have it working I see random glitches here and there that I can't pin down. Some Electron apps I have to turn off GPU acceleration or they won't get any windows showing up - they launch, the process exists, they're in the dock as active, but the window doesn't appear at all. Getting a new laptop from work to replace this one and I'm really hoping it won't have nvidia hardware - or at least, if it does I can disable it and the Intel GPU will work fine also.

drcongo 1y ago

I'm one month into owning a GMKTec Evo-X2 with the new mad AMD gpu in it and so far I've managed to get Ollama running. ROCm doesn't officially support the GPU and everything seems to be hacky workaround on top of hacky workaround. Starting to wish I'd just waited for NVIDIA to actually release the project digits thing.

Tepix 1y ago
So, how did you get it running? I‘m about to get a Bosgame M5 (same CPU).
- drcongo 1y ago
  Installing the ROCm nightlies from TheRock and upgrading the linux kernel to 6.16. However the kernel upgrade breaks a ton of other things, notably dkms. I've been trying for weeks to get ComfyUI to not explode without success.
- nicolaslem 1y ago
  I would recommend not wasting time with ollama, https://github.com/containers/ramalama https://github.com/containers/ramalama just works on that chip.

fooker 1y ago

FYI All current and future Nvidia drivers are open source, since blackwell.

kcb 1y ago
Kernel driver only.
- cromka 1y ago
  Is AMD different, do they OpenSource firmware as well?
  - ItsHarper 1y ago
    They have an open source userspace component.
jonkoops 1y ago
But they are not in the actual Linux kernel, and are a separate module that does not adhere to Linux kernel code conventions. It also has no user-space driver that isn't NVIDIA's proprietary driver. Anything further open-source such as NVK is not being worked on by NVIDIA, but by other 3rd parties. Compared to AMD and Intel, NVIDIA is very much not an 'out of the box', or stable experience.
mdaniel 1y ago
Saying "and future" is like taunting fate Anyway, in case someone was interested it seems the code itself is cited as MIT, however it has a "when it becomes a Linux .ko it becomes GPLv2" clause https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.76.05/COPYING https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.7... and they do go out of their way to say "lol, needs binary blobs" https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.76.05/README.md?plain=1#L18-L19 https://github.com/NVIDIA/open-gpu-kernel-modules/blob/580.7... That XFree86.run has always struck me as "you're gonna what*?"
- account42 1y ago

ethersteeds 1y ago

I came to say the same thing. In the late 2010s I ran Linux on a work-issued Lenovo P50 with a Nvidia Quadro M2000M. It was such a miserable experience that I swore to never own another NVIDIA product again. Given both Framework and NVIDIA's checkered histories around Linux driver support, I see no reason to revisit that, but it is interesting to see the voices in this thread with positive NVIDIA experience.

bodge5000 1y ago
I said in another comment but exact same experience here (though about a decade later), same laptop and gpu. It's a shame because the P50 is so nice otherwise. Maybe its just a problem with older Nvidia gpu's, but its not a gamble I want to take