I am struggling to understand where things are going wrong. I have purchased a computer and have installed all/ most of the latest software to run cryoSPARC and have successfully read in the images, run the motion correction and CTF estimations on the movies and am trying to run blob picking. I’ve tried this several times with the same or slightly different parameters and each time it crashes the whole computer at a random point (a random movie/image). And when I say ‘crash’ I mean a hard crash- the only thing left going is the power light, everything else is dead.
I’ve got cryoSPARC 3.3.2, cuda release 11.7, NVIDIA driver 515.65.01.
The new machine has:
An AMD Ryzen 9 5900X cpu
Zotac GeForce RTX 3090 24 Gb gpu
64 Gb of RAM
Gigabyte X570S Aorus motherboard with liquid cooling.
One of the weird or interesting things from my viewpoint is that I’m barely using the cpu/gpu when the blob picking is going on, so it isn’t a question of the gpu or cpu running hot. We’ve done diagnostics on the hardware running the cpu/gpu at 100% overnight without issues (as I suspected that it was a hardware issue initially).
Any help would be very much appreciated!
Best regards, tom
@wtempel I am running into a similar issue with my system as well and would very much appreciate some guidance on this. Below are my system specifications:
CryoSPARC now comes with its own CUDA bundled, so one shouldn’t need to install separately.
In order for me to get my system running, I had to upgrade to Ubuntu 25.04 (tried 22.04 and 24.04), but I am now running the RTX 5090 GPU, so that may not be important for you.
Is there anything else running at the time? What OS? What dataset (mic dimensions, angpix, box size, etc)? What errors are reported in CryoSPARC (check metadata/logs)? What errors are reported in dmesg?
Skylake (the generation which the 7920X belongs to) is quite old and officially only supports 128GB of RAM (I presume the 156GB is a typo and should be 256GB). It may be as simple as the memory controller struggling with too much memory. Do the crashes happen anywhere else on the system (e.g. if you run RELION or cisTEM)?