Cryosparc on arm64

Install, Configure and Update
1

Hello CryoSPARC developers. You are doing a fantastic job keeping up CryoSPARC fresh.

The last decade has brought a revolution in utilizing GPUs, specifically Nvidia GPUs and CUDA, to accelerate data processing. Nowadays, there is a visible push from Nvidia on their CPUs as well, in the form of DGX Spark and DGX Station, which appear to be absolutely magnificent pieces of hardware for data processing.

So far, I have had no issues getting various CryoEM applications to work on Arm with either conda or apptainer/docker containers. Things just worked with full CUDA acceleration.

Therefore, Arm64 support should not be a far-fetched goal, so I have two questions related to a future CryoSPARC release.

  1. Are you planning to and working to release CryoSPARC on arm64?
  2. If so, does it appear to be a distant future?

I do not expect an exact answer, but specific hardware decisions can be a lot easier.

Thank you so much for the great work!

5 Likes
2

I would like to second @osinskit ‘s request to support CryoSparc on ARM64.

Specifically, I have recently purchased two Dell GB10 ( Dell's version of the DGX Spark fixes pain points - Jeff Geerling , ) mini workstations for my lab at OIST, which are Dell-branded Nvidia DGX Spark with the Grace Blackwell super chip. Grace is an ARM64 CPU. The compact system has 128GB shared memory which is fully accessible to the Blackwell GPU.

The overall performance is very good. The system runs Ubuntu. It ships with CUDA 12. ChimeraX runs on it using their generic ARM64 installer. It would be awesome if you could port CryoSparc to this architecture. I believe that nVidia’s technology is bound to replace the dominance of Intel x86.

Modern LLMs used e.g. in CoPilot allow effortless code adaptation to other languages and platforms, as well as using CUDA, which significantly reduce the barrier for die-hard legacy x86 programmers to enter the future of massively parallel GPU computing. There is simply no way to justify ignoring this technology leap anymore.

I would like to use CryoSparc to reconstruct very large volumes that do not fit into most conventional GPU memory anymore. There is surely also a roadmap for CryoSparc with cryo-ET, which would also benefit from such large volumes. Please consider ARM64 by all means.