cryoSPARC requires a minimum of one NVIDIA GPU with compute capability 3.5+ and 4GB+ RAM. This includes the most common cards K40, K80, TITAN Z, TITAN X, GTX 1080, GTX 1070. We have tested on K40, TITAN Z, and TITAN X, but new Titan and GTX cards are faster.
Have you ever tried the NVIDIA 1080 Ti card? I seems to be very similar to the Titan X in some tests and has a larger memory 11GB compared with the 1080 card.
We haven’t tried the 1080Ti ourselves yet. If any users have, please let us know. There have been some reports that the current NVIDIA drivers for the 1080Ti have some bugs, but I can’t confirm.
In principle, the 1080Ti should provide excellent performance (better than Titan X).
I have successfully installed and tested cryoSPARC on our GPU server with dual GTX 1080 Ti GPU cards with 381.22 NVIDIA driver and CUDA 8.0.
I don’t have any comparison to other GPU cards, but creation of ab initio model for this 80S ribosome data took ~1:15 h (significant amount of that time was used for SSD caching), and the refinement to 3.2 A took ~39 minutes, so its probably very similar to Titan X.