0

On my computer I have an NVIDIA Quadro KM2100M graphic card with the following main parameters (as reported by Cuda-Z tool):

Name: Quadro K2100M
Compute Capability: 3.0
Clock Rate: 666.5 MHz
PCI Location: 0:1:0
Multiprocessors: 3 (576 Cores)
Threads Per Multiproc.: 2048
Warp Size: 32
Regs Per Block: 65536
Threads Per Block: 1024
Threads Dimensions: 1024 x 1024 x 64
Grid Dimensions: 2147483647 x 65535 x 65535
Watchdog Enabled: Yes
...

I found out that installing PyTorch 1.8.1 results in answering that CUDA is enabled:

import torch
print(torch.__version__)
# returns 1.8.1+cu101

print(torch.cuda.is_available())
# returns True

At the first attempt to really use CUDA in PyTorch, however, I get immediately

UserWarning: 
Found GPU0 Quadro K2100M which is of cuda capability 3.0.
PyTorch no longer supports this GPU because it is too old.
The minimum cuda capability that we support is 3.5.

and

RuntimeError: CUDA error: no kernel image is available for execution on the device

Is there a possibility to run ANY PyTorch version with NVIDIA Quadro K2100M? (Compute Capability being the problem, I assume)?

Stanislav Koncebovski
  • 462
  • 5
  • 10
  • 1
    https://stackoverflow.com/questions/62437918/which-pytorch-version-is-cuda-3-0-compatible – Robert Crovella Jan 25 '22 at 17:01
  • The linked duplicate has an accepted answer that was written in a broad way to cover questions like yours. Please refer to section 4 of that answer. To use a prebuilt torch, built by the torch maintainers, would require you to revert back to torch 0.3.1 or earlier, and this is generally not recommended, as it is quite old. The other options for this situation are listed there as well. – Robert Crovella Jan 25 '22 at 17:08
  • Thanks @RobertCrovella, I really failed to find that old question and, really, it is quite identical with mine. – Stanislav Koncebovski Jan 26 '22 at 07:29

0 Answers0