Minimum CUDA driver version which works with R2024a

MATLAB displays the error saying that "Graphics driver is out of date. Download and install the latest graphcis driver for your GPU from NVIDIA." despite we have relatively new driver version (530.30.02, Feb2023).
As the page https://www.mathworks.com/help/releases/R2024a/parallel-computing/gpu-computing-requirements.html did not list the driver version required, I would like to know how to check the minimum driver version required?

답변 (2개)

Matt J
Matt J 2024년 9월 23일

2 개 추천

The driver version you need is a function of your GPU card. Why not just install the latest one as the error message suggests?
Ayush
Ayush 2024년 9월 23일
I understand that you need to run MATLAB on your GPU, but you are facing issues with the version of the installed GPU driver.
To resolve your issue, you can refer to the “Note” section of the documentation : https://www.mathworks.com/help/releases/R2024a/parallel-computing/gpu-computing-requirements.html.
It says MATLAB supports NVIDIA GPU architectures with compute capability 5.0 to 9.x.
To determine the "compute capability" of your GPU, please visit: https://developer.nvidia.com/cuda-gpus. Additionally, you can check the maximum CUDA support of your installed driver by executing the following command:
$ Nvidia-smi
This information will help you identify the appropriate driver to install, ensuring compatibility with your MATLAB version.
Hope this helps!

댓글 수: 5

MATLAB R2023b supports by GPU, but R2024a not. Just wanna know what is the minimum version required as it is not written in the R2024a documentation (the link which I have shared).
nvidia-smi output (slightly redacted):
Mon Sep 23 17:16:52 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 530.30.02 Driver Version: 530.30.02 CUDA Version: 12.1 |
|-----------------------------------------+----------------------+----------------------+
| 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 Quadro RTX 6000 Off| 00000000:86:00.0 Off | Off |
| 33% 32C P8 10W / 260W| 33MiB / 24576MiB | 0% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
+---------------------------------------------------------------------------------------+
R2023b output:
>> gpuDeviceTable
ans =
1x5 table
Index Name ComputeCapability DeviceAvailable DeviceSelected
_____ _________________ _________________ _______________ ______________
1 "Quadro RTX 6000" "7.5" true false
Ayush
Ayush 2024년 9월 23일
편집: Ayush 2024년 9월 23일
If possible, can you share your output of "gpuDeviceTable" in MATLAB R2024a. Also, if any other errors you are encountering apart from the one you mentioned above.
Ayush
Ayush 2024년 9월 23일
편집: Ayush 2024년 9월 23일
Your issue might be because MATLAB R2024a supports CUDA toolkit version 12.2 but as per your ouput, your driver version supports CUDA version 12.1 which might be the cause of issue in MATLAB R2024a.
You can read more about it in the "Install CUDA Toolkit" section here https://www.mathworks.com/help/parallel-computing/run-mex-functions-containing-cuda-code.html
You can also use "canUseGPU" command to verify compatibility. You can read more about it here: https://www.mathworks.com/help/matlab/ref/canusegpu.html
Yau Hing Tuen
Yau Hing Tuen 2024년 9월 23일
편집: Yau Hing Tuen 2024년 9월 23일
Understood. So by having libcudart.so.12.2.140 in the installation, R2024a requires CUDA 12.2, which means driver has to be >=535.54.03
(Cross-ref: https://docs.nvidia.com/cuda/archive/12.2.1/cuda-toolkit-release-notes/index.html )
(Edit: Not the minimum version supported by CUDA 12.2, but the driver associating with the release of CUDA 12.2)
Glad you understood. if you find it useful, you can "accept" the answer so that it helps the other people having similar issue.

댓글을 달려면 로그인하십시오.

카테고리

도움말 센터File Exchange에서 GPU Computing에 대해 자세히 알아보기

제품

릴리스

R2024a

질문:

2024년 9월 23일

답변:

2024년 9월 23일

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by