Matlab throws and error while using gpuDevice. It asks for CUDA driver newer than 10.1. I have 10.2 installed.

If I run the command gpuDevice, I get the following error,"Error using gpuDevice (line 26)

CUDA driver version 9 is not supported. Update CUDA driver to version 10.1 or greater. For more information on GPU support,

see GPU Support by Release."

I am positive that I have installed CUDA driver 10.2.

Answer 1

[Jason Ross]

If you run nvidia-smi it will tell you the driver version you are running. On Linux it should be on the path, on Windows it's in "C:\Program Files\Nvidia corporation\nvsmi\nvidia-smi.exe"

Output will look something like the following:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.33.01    Driver Version: 440.33.01    CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K80           Off  | 00000000:04:00.0 Off |                    0 |
| N/A   37C    P8    29W / 149W |     11MiB / 11441MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  Tesla K80           Off  | 00000000:05:00.0 Off |                    0 |
| N/A   46C    P0    71W / 149W |      0MiB / 11441MiB |     99%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

[Tamoghna Bhattacharya]

I tried this, a pop up window appears and then instantly disappears. May be because my GPU is pretty low key, I am running a Geforce GT 820M.

[Jason Ross]

You need to start a Command Prompt and run it in there -- I assume you are running on Windows?

That said, looking at the latest driver available for this card from nVidia, (423.31), it was released on April 11, 2019. According to the Wikipedia article on the GeForce 800M series, it looks like this particular GPU uses the GF117, which makes it a Fermi class GPU. In 2018, nVidia moved these cards to "legacy status", for which the final CUDA toolkit version is 8.0 -- so anything requiring a tooklit version > 8 is going to not work with this GPU.

According to our GPU support matrix, the latest release to support CUDA toolkit 8.0 is R2017b. Anything later will require a newer GPU to work with it.

I assume when you run nvidia-smi, it's going to show you have CUDA 8.

[Tamoghna Bhattacharya]

Yes, I got the impression that my GPU is an old one. Thanks for the help though.

[Gaetan Richard]

I have a similar issue with my GPU, I have a NVIDIA Quadro 600, the pilot was released in january 2018. I can't open the nvidia-smi.exe (the pop up open and close immediately), and after installing the CUDA toolkit version 10.1 (when I look on my prompt the version is correct I have the 10.1) but when assessing the gpuDevice command on Matlab I have :

Error using gpuDevice (line 26)

CUDA driver version 8 is not supported. Update CUDA driver to version 10.1 or greater. For more

information on GPU support, see GPU Support by Release.

Is the GPU Quadro 600 too old?

Thank you in advance for your help,

Gaetan

[Walter Roberson]

@Gaetan Richard

Quadro T600: compute capacity 7.5

Quadro P600: compute capacity 6.1

Quadro K600: compute capacity 3.1

Quadro M600M: compute capacity 5.0

MATLAB® supports NVIDIA® GPU architectures with compute capability 5.0 to 9.x.

So you should be able to get it to work provided that you do not have the K600