GPU not recognized (r2020a and cuda 10.2)

5 views (last 30 days)
Ben Mc Larney
Ben Mc Larney on 19 Mar 2020
Commented: Yanhui Guo on 11 Jan 2021
Hi there,
Recently have a new PC with a titan xp with cuda 10.2 installed and matlab 2019b. I've done stress tests etc, and the GPU is recognized by the PC (GPUZ and in the task manager). Also ran nvidia-smi.exe (screenshot attached) with all the details. From what I've read the insufficient permissions line is not relevant to matlab seeing the GPU or not.
Matlab still doesn't see the GPU, any idea why? Should I use CUDA 10.1? From what I understand 10.2 should still have all the drivers needed from 10.1 so not sure if that would fix it. I can't currently install 2020a due to licensing access but if that would be the solution I can try and figure it out.
Many thanks for any help!
  5 Comments
Ben Mc Larney
Ben Mc Larney on 23 Mar 2020
Hi Alison,
Looks like you could be right. Here's the output from that code:
cd ([matlabroot '/bin/win64'])
!CertUtil -hashfile gpushim.dll SHA1
CertUtil: -hashfile command FAILED: 0x80070002 (WIN32: 2 ERROR_FILE_NOT_FOUND)
CertUtil: The system cannot find the file specified.
Many thanks,
Ben
Yanhui Guo
Yanhui Guo on 11 Jan 2021
Hi Alison,
I have the same problem in R2020b and CUDA10.2. I already install CUDA 10.2.
When I run gpuDevice()
There is error as
Error using gpuDevice (line 26)
CUDA driver version 10.1 is not supported. Update CUDA driver to
version 10.2 or greater. For more information on GPU support, see GPU
Support by Release.
After I run the code
cd ([matlabroot '/bin/win64'])
!CertUtil -hashfile gpushim.dll SHA1
I got that:
SHA1 hash of file gpushim.dll:
f5 39 63 20 25 d7 2a df 1d 00 5a 8a 74 f1 a2 26 5a b7 4b ff
CertUtil: -hashfile command completed successfully.
Please help.
Thanks a lot.

Sign in to comment.

Answers (0)

Community Treasure Hunt

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

Start Hunting!