exo icon indicating copy to clipboard operation
exo copied to clipboard

Nvidia GPU not detected on Windows 11

Open BolverBlitz opened this issue 11 months ago • 4 comments

My System got CPU: Ryzen 9 7950X3D + Integrated GPU (128GB) GPU: RTX 3090 (24GB)

I see no manual config option, so i run it with python3 ./exo/main.py, it seems to work and downloads models. But it claims it uses 128GB and 0TFLOPS so its not using the RTX Card.

I´ve also added CUDA=1 to my env as some people on linux sugested in issues $env:CUDA = "1"

nvcc --version

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Wed_Apr_17_19:36:51_Pacific_Daylight_Time_2024
Cuda compilation tools, release 12.5, V12.5.40
Build cuda_12.5.r12.5/compiler.34177558_0

nvidia-smi

Wed Jan  8 16:31:11 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.94                 Driver Version: 560.94         CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                  Driver-Model | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 3090      WDDM  |   00000000:01:00.0  On |                  N/A |
|  0%   44C    P8             47W /  370W |    2744MiB /  24576MiB |     30%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI        PID   Type   Process name                              GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A      6464    C+G   C:\Windows\explorer.exe                     N/A      |
|    0   N/A  N/A     11168    C+G   ...m Files\Mozilla Firefox\firefox.exe      N/A      |
|    0   N/A  N/A     13700    C+G   ...nt.CBS_cw5n1h2txyewy\SearchHost.exe      N/A      |
|    0   N/A  N/A     13736    C+G   ...2txyewy\StartMenuExperienceHost.exe      N/A      |
|    0   N/A  N/A     15680    C+G   ...Roaming\64Gram Desktop\Telegram.exe      N/A      |
|    0   N/A  N/A     15980    C+G   ...t.LockApp_cw5n1h2txyewy\LockApp.exe      N/A      |
|    0   N/A  N/A     17672    C+G   ...oration\NvContainer\nvcontainer.exe      N/A      |
|    0   N/A  N/A     20440    C+G   ...m Files\Mozilla Firefox\firefox.exe      N/A      |
|    0   N/A  N/A     21336    C+G   ...GeForce Experience\NVIDIA Share.exe      N/A      |
|    0   N/A  N/A     22320    C+G   ...CBS_cw5n1h2txyewy\TextInputHost.exe      N/A      |
|    0   N/A  N/A     25152    C+G   ...es\Elgato\StreamDeck\StreamDeck.exe      N/A      |
|    0   N/A  N/A     27580    C+G   ...\AMD\CNext\CNext\RadeonSoftware.exe      N/A      |
|    0   N/A  N/A     27812    C+G   ...les\AMD\CNext\CNext\AMDRSSrcExt.exe      N/A      |
|    0   N/A  N/A     28540    C+G   ...les\Microsoft OneDrive\OneDrive.exe      N/A      |
|    0   N/A  N/A     28860    C+G   ...\cef\cef.win7x64\steamwebhelper.exe      N/A      |
|    0   N/A  N/A     30516    C+G   ...al\Discord\app-1.0.9175\Discord.exe      N/A      |
|    0   N/A  N/A     32904    C+G   ...\Programs\signal-desktop\Signal.exe      N/A      |
|    0   N/A  N/A     34100    C+G   ... Stream\101.0.3.0\GoogleDriveFS.exe      N/A      |
|    0   N/A  N/A     35300    C+G   ...38.0_x64__zpdnekdrzrea0\Spotify.exe      N/A      |
|    0   N/A  N/A     38268    C+G   ...tBakPCAgent\App\NetBak PC Agent.exe      N/A      |
|    0   N/A  N/A     67568    C+G   ...esktop\app-3.4.12\GitHubDesktop.exe      N/A      |
|    0   N/A  N/A     68620    C+G   ...Mozilla Thunderbird\thunderbird.exe      N/A      |
|    0   N/A  N/A     76684    C+G   ...5n1h2txyewy\ShellExperienceHost.exe      N/A      |
|    0   N/A  N/A     81580    C+G   ...crosoft\Edge\Application\msedge.exe      N/A      |
|    0   N/A  N/A     84284    C+G   ...__8wekyb3d8bbwe\WindowsTerminal.exe      N/A      |
+-----------------------------------------------------------------------------------------+

BolverBlitz avatar Jan 08 '25 15:01 BolverBlitz

hi there, an issue that i've had myself a lot of times is that sometimes the casing inside the exo/exo/topology/device_capabilities.py might not be the same as what gets detected. For your case: Your GPU name is: NVIDIA GeForce RTX 3090 but inside said file it's called:

"NVIDIA GEFORCE RTX 3090": DeviceFlops(fp32=35.6*TFLOPS, fp16=71.2*TFLOPS, int8=142.4*TFLOPS),
"NVIDIA GEFORCE RTX 3090 TI": DeviceFlops(fp32=40.0*TFLOPS, fp16=80.0*TFLOPS, int8=160.0*TFLOPS),
# RTX 20 series
"NVIDIA GEFORCE RTX 2060": DeviceFlops(fp32=6.45*TFLOPS, fp16=12.9*TFLOPS, int8=25.8*TFLOPS),

so maybe try to modify that file yourself, exo/exo/topology/device_capabilities.py LINE: 86 to "NVIDIA GeForce RTX 3090": DeviceFlops(fp32=35.6*TFLOPS, fp16=71.2*TFLOPS, int8=142.4*TFLOPS)

RickyRAV avatar Jan 09 '25 23:01 RickyRAV

same here on Ubuntu 24.04, CPU: Ryzen 9 5950X + Integrated GPU (128GB) GPU: RTX 3090 Ti(24GB)

jli113 avatar Jan 12 '25 08:01 jli113

Make sure you install CUDA drivers if you're using WSL and Ubuntu. Immediately fixed it for me:

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=WSL-Ubuntu&target_version=2.0&target_type=deb_local

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda-repo-wsl-ubuntu-12-8-local_12.8.0-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-12-8-local_12.8.0-1_amd64.deb
sudo cp /var/cuda-repo-wsl-ubuntu-12-8-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-8

unstatusthequo avatar Jan 28 '25 17:01 unstatusthequo

Si son usuarios de windows y presentan problema con uvloop, pueden usar wsl.

  1. Abre una terminal de wsl Si no lo tienes puedes instalarlo con wsl --install siguiendo los pasos de Instalación de Linux en Windows con WSL, en mi caso ya tenía instalado la distribución Ubuntu 22.04.5 LTS.

  2. Verifica la instalación de los drivers de cuda con nvcc --version y nvidia-smi Si no responden posiblemente necesites instalar los drivers con sudo apt install nvidia-cuda-toolkit

  3. Ejecuta exo y deberías mirar los TFLOPS Si no tienes instalado exo entonces lo puedes instalar con los pasos sugeridos por EXO

git clone https://github.com/exo-explore/exo.git
cd exo
pip install -e .
# alternatively, with venv
source install.sh

nelsonboada avatar Feb 02 '25 17:02 nelsonboada