Just in case this might be of use to the wider community…
Added two new GPUs to two of our servers yesterday.
With CryoSPARC 4.4.1, following the guide didn’t work quite as expected: while it detected the two new GPUs, it didn’t count past the original number detected (8, or rather 7 as it starts from 0). Further, something seems to have got a little stuck as even after a restart of CryoSPARC, I got CUDA errors (illegal access) if trying to use either of the new GPUs, or, for some reason, the last two detected “old” GPUs.
To get all GPUs detected (and no CUDA errors) I needed the following:
./cryosparcw connect --master [hostname] --worker [hostname] --update --gpus 0,1,2,3,4,5,6,7,8,9
cryosparcm restart
Sort of a note-to-self I thought I’d share just in case anyone else stumbles over a similar quirk.