I finally got a chance to try this and it didn’t work for me. The job didn’t spend any time in the ‘building phase’, so as soon as I hit ‘queue’ it starting going and I didn’t have any opportunity to change from 1 to 2 GPUs. On the other hand, I didn’t see it move from 2 to 1 GPUs when I hit ‘queue’. The result is the same, it only used/uses 1 GPU.
The desired value for a job parameter, 2 for Number of GPUs to parallelize in the current case, needs to be specified before one presses the Queue Job button. What happens when you try that?
Sorry for the misunderstanding- the number of GPUs was set to 2 prior to building/ putting in the queue (as it always has been). There was no build ‘lag’ in order to reset to 2 (from 2). Upon hitting queue, the job started with using only 1 GPU. Sorry for the miscommunication.
This may be best under a new thread, but I’ll start off here.
I just added 2x 4Tb SSD memory to the computer as I kept getting errors stating “RuntimeError: SSD cache needs additional 5.53 TiB”. So I now have 12 Tb of SSD memory available and mounted. How do I get cryoSPARC to recognise this new memory? It is mounted, I see it when I do a df, etc. My ‘scratch’ file is currently set to /home/tom/scratch and I’m guessing I need to somehow add these other /dev/xxx to the scratch file but I can’t find the way to do this using the search terms I’m using…
Any help would be very much appreciated.
Forget that last question- I finally found the right combination of keywords to get a fix.
In case anyone was interested- I created a new scratch directory with the 2 new SSD cards using mkfs.btrfs.
Just updated to cryoSPARC 5.0, so maybe that will fix my one GPU problem…
@tom What did you find in v5?
Thanks for the question- I haven’t tried to use 2 GPUs on very many things so far, but yes, at least for a couple of things the upgrade to version 5 has fixed the issue and I can now set jobs to use both GPUs. I’m happy with the new version!
Best regards, Tom