RTX A4500 showing as RTX 3090

This isn’t a major issue or anything, more of a GUI/QoL issue, but CryoSPARC shows RTX A4500 GPUs as RTX 3090s. I realize they’re virtually the same card, but since the RTX A5000 shows up in CryoSPARC, it would be nice to have the A4500 show up as well.

Does the installed version of the CUDA toolkit or the nvidia driver predate the release of the A4500 GPU?

No, it does not. We are using Cuda 11.7.1 which was released in August 2022. The A4500 was released the year prior. When you run nvidia-smi they show up as A4500s.

Are the outputs of
/path/to/cryosparc_worker/bin/cryosparcw gpulist
and
cryosparcm cli "get_scheduler_targets()"
consistent w.r.t GPU models?
What are the outputs of
cat /proc/driver/nvidia/version
and

/path/to/cryosparc_worker/bin/cryosparcw call python -c "import pycuda.driver; print(pycuda.driver.get_version())"

gpulist provides the following:

bin$ ./cryosparcw gpulist
  Detected 4 CUDA devices.

   id           pci-bus  name
   ---------------------------------------------------------------
       0      0000:81:00.0  NVIDIA RTX A4500
       1      0000:82:00.0  NVIDIA RTX A4500
       2      0000:C1:00.0  NVIDIA RTX A4500
       3      0000:C2:00.0  NVIDIA RTX A4500
   ---------------------------------------------------------------

cryosparcm cli lists them as RTX 3090s however:

{'id': 0, 'mem': 25447170048, 'name': 'NVIDIA GeForce RTX 3090'}, {'id': 1, 'mem': 25447170048, 'name': 'NVIDIA GeForce RTX 3090'}, {'id': 2, 'mem': 25447170048, 'name': 'NVIDIA GeForce RTX 3090'}, {'id': 3, 'mem': 25447170048, 'name': 'NVIDIA GeForce RTX 3090'}],

All other servers are consistent.

nvidia version info:

bin$ cat /proc/driver/nvidia/version 
NVRM version: NVIDIA UNIX x86_64 Kernel Module  515.65.01  Wed Jul 20 14:00:58 UTC 2022
GCC version:  gcc version 11.3.0 (Ubuntu 11.3.0-1ubuntu1~22.04) 

pycuda driver version:

bin$ ./cryosparcw call python -c "import pycuda.driver; print(pycuda.driver.get_version())"
(11, 7, 0)

You might try
cryosparcw connect with the --update option and the other options you used when you last ran that command (guide) and see of the card info is shown correctly after that.

Just tried it and unfortunately it didn’t work. Still claims they are RTX 3090s

Due to an unrelated issue, I restarted the database and that seems to have fixed the issue. The GPUs are now showing up in CryoSPARC as they should be, and everything seems to be running fine.

1 Like