Blender wont detect nvidia cuda

English Applications
,
1

been having some issues and spend 2 days trying to figure out this and first thought my nvidia isint loaded, but drivers are installed, MOK is signed, Nvidia-smi shows nvidia, glxinfo shows OpenGL as intel, prime-select is set to offload and modules loaded and htop shows both cards so that seems not be driver issue since on my side nvidia is loaded and working, but Blender on RPM install wont detect cuda even cuda tools are installed

using KDE Plasma on wayland

2

¿What does this script outputs?

3

seems like i am still missing something on cuda side

phatle@tumbleweed:~> #!/usr/bin/bash
phatle@tumbleweed:~> 
phatle@tumbleweed:~> function lib_installed() { /sbin/ldconfig -N -v $(sed 's/:/ /' <<< $LD_LIBRARY_PATH) 2>/dev/null | grep $1; }
phatle@tumbleweed:~> function check() { lib_installed $1 && echo "$1 is installed" || echo "ERROR: $1 is NOT installed"; }
phatle@tumbleweed:~> 
phatle@tumbleweed:~> check libcuda
        libcuda.so.1 -> libcuda.so.550.90.07
        libcudadebugger.so.1 -> libcudadebugger.so.550.90.07
        libcuda.so.1 -> libcuda.so.550.90.07
libcuda is installed
phatle@tumbleweed:~> 
phatle@tumbleweed:~> check libcudart
ERROR: libcudart is NOT installed
phatle@tumbleweed:~> 
phatle@tumbleweed:~> check libnccl
ERROR: libnccl is NOT installed
phatle@tumbleweed:~> 
phatle@tumbleweed:~> check libcudnn 
ERROR: libcudnn is NOT installed
phatle@tumbleweed:~> 
phatle@tumbleweed:~> nvcc -V
4

Yeah, libcudart is missing. ¿How did you installed CUDA?

5

from yast software took all G06 compute ones and KDE installed rest automatic

6

any ideas how to fix? or re-install all downloading now the cuda toolkit, but that is version 12.5.1_555.42.06 and tumbleweed i have 550.90.07 and cuda 12.4

7

also hows the open drivers? i dont play on this laptop as physical i run steam just to buy games and games i play is done from cloud streaming. open drivers has cuda packages done already so it might be good switch, but never used open ones so do i still need to sign everyhting to kernel

8

Fixed the blender no cuda removing all propriatery nvidia and installed open nvidia drivers with cuda atleast now i can run blender and cuda still some libraries is missing when checking the script

phatle@tumbleweed:~/Documents> ./nvidia-cuda.sh
        libcudart.so.12 -> libcudart.so.12.4.127
        libcuda.so.1 -> libcuda.so.550.90.07
        libcudadebugger.so.1 -> libcudadebugger.so.550.90.07
        libcuda.so.1 -> libcuda.so.550.90.07
libcuda is installed
        libcudart.so.12 -> libcudart.so.12.4.127
libcudart is installed
ERROR: libnccl is NOT installed
ERROR: libcudnn is NOT installed
./nvidia-cuda.sh: line 14: nvcc: command not found
1 Like
9

These libraries are optional, you can extract them and move to /usr/lib64

libnccl: NVIDIA Collective Communications Library (NCCL) | NVIDIA Developer (You’ll need to make a account and fill the forms fro this one)
libcudnn: CUDA Deep Neural Network (cuDNN) | NVIDIA Developer

1 Like
10

Last time I used the blender rpm version, I installed the cuda from nvidia site.
I haven’t done it for years so not really sure if what the op did was enough for blender to recognize cuda. I am using the blender .tar.xz now, less hassle for cuda because it’s built-in in the .tar.xz installer.

11

Blender works on my system rpm needs cuda files etc to be installed, but easy way is just install normal drivers with compute and use Flatpak version and then all done.

On my current daily AEON I haven’t installed cuda just
transactional-update -i pkg in nvidia-drivers-G06 nvidia-driver-G06-kmp-default nvidia-video-G06 nvidia-gl-G06 nvidia-compute-G06

Flatpaks works, cuda works, blender works, openCL works

12

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.