Thread up
@vajdao Hi, and you got your answers 
Yepp, I was replying to The_Istar so that s/he knows its up if s/he wants to join the talk
@vajdao I would suggest the cuda run file, it’s what I use to start with and then update driver as required… but it does mean a manual update on a new kernel or use dkms (which should work).
@malcolmlewis thanks for reply, unfortunately the CUDA approach did not work. I’m not even sure if its good for gaming. Isn’t CUDA meant for calculation tasks?
@vajdao compute, which will improve desktop performance as applications will use this if present. More and more take advantage.
I guess I’ve been using the run file for too long But in saying that at this point it’s only used for Prime Render Offload, but not had issues as primary graphics when I did use it for this.
@malcolmlewis ok, so I knew right. So its not optimized for gaming. CUDA is not for me. I need gaming driver, not computing 
I’m not a developer, nor designer. I use my computer as an entertainment center, gaming hub 'n such.
@vajdao No, desktop applications will take advantage of it and improve gaming performance for sure they will use cuda.
Okay, I’ll give it a 3rd try then, but I don’t think its gonna work, because I already tried and I was stuck with a very low resolution 
I guess it doesn’t support my card 
@vajdao Start a new thread so I can walk through it all with you.
Tremendous situation. I can only understand (and probably I didn’t understand well enough) that 560 cannot arrive as version 560 has bugs with Xorg (and, also, on the Nvidia forum seems that GSP has performance issues with Wayland). Also, Linux’s devs have (maybe???) removed a particular “function” (or whatever’s the word) necessary for Nvidia. Sheesh.
Also, careful with that CUDA. You DO need the compute driver. I think that malcolmlewis mentioned some .run file (is that the usual blob or another one?), which is not the Nvidia CUDA repo. Also, dkms inside that .run file has never worked for me, so manual reinstallation is necessary at every kernel update. Honestly, it’s quick and easy, it’s okay. Better open that thread as they said and resolve your issue.
@shishimaru Yes, there is a cuda run file as well as a driver run file, two different things
This includes the driver as well as the cuda bits…
I use this first, then use the driver run file to update as appropriate.
nvidia-smi
Wed Sep 18 10:58:49 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA T400 Off | 00000000:02:00.0 Off | N/A |
| 58% 49C P8 N/A / 31W | 3MiB / 2048MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
nvidia_check
libcudart.so.12 -> libcudart.so.12.6.68
libcuda.so.1 -> libcuda.so.560.35.03
libcudadebugger.so.1 -> libcudadebugger.so.560.35.03
libcuda.so.1 -> libcuda.so.560.35.03
libcuda is installed
libnccl.so.2 -> libnccl.so.2.20.3
libnccl is installed
libcudnn_ops.so.9 -> libcudnn_ops.so.9.2.0
libcudnn_heuristic.so.9 -> libcudnn_heuristic.so.9.2.0
libcudnn_graph.so.9 -> libcudnn_graph.so.9.2.0
libcudnn_engines_runtime_compiled.so.9 -> libcudnn_engines_runtime_compiled.so.9.2.0
libcudnn_engines_precompiled.so.9 -> libcudnn_engines_precompiled.so.9.2.0
libcudnn_cnn.so.9 -> libcudnn_cnn.so.9.2.0
libcudnn_adv.so.9 -> libcudnn_adv.so.9.2.0
libcudnn.so.9 -> libcudnn.so.9.2.0
libcudnn is installed
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Wed_Aug_14_10:10:22_PDT_2024
Cuda compilation tools, release 12.6, V12.6.68
Build cuda_12.6.r12.6/compiler.34714021_0As malcomlewis said: open a thread, explain what goes wrong, someone can try to tackle your issues. Unless you are okay with the Nouveau drivers or if Nvidia 550 works good. In that case, you can stay with them.
Yepp its ok, I’ll stay with 550, and waiting on 560’s official release.
@malcolmlewis I can install 560 from OBS though, and it works, I just don’t prefer using OBS.
Hey Vajdao,
I did join the talk and my question has been answered.
Let me share what I learned.
The reason why we do not have the New feature branch is because it seems it is technically it is not really possible to include both drivers unless we use different driver package names (see mailing list for details).
However, this would cause such an extra workload on the single maintainer of the NVIDIA package that he can not manage this.
However he did say that if anyone from the community would step forward and package and maintain the feature branch this would be possible.
Unfortunately I myself am not a linux package maintainer. I would honestly not even know where to start. But it does make me wonder if the/a person hosting it already on the obs repo could not mainline it.
But maybe that person simply does not want this responsibility.
Anyway, this was enlightening.
2 Likes
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.