cryoSPARC v4.7.1-cuda12+250814 cannot be downloaded/installed

Install, Configure and Update
1

Hey,

We have a new workstation with Nvidia Blackwell GPUs running on Ubuntu 24LTS and CUDA 575.51.03. I initially followed the cryosparc instructions to download and install cryosparc, next I tried to update to the Blackwell-specific version with cryosparcm update --version=v4.7.1-cuda12+250814, which fails:

 cryosparcm update --version=v4.7.1-cuda12+250814

CryoSPARC current version v4.7.1

          update starting on Thu Nov 13 10:40:21 AM CET 2025

 

=============================

Updating to version v4.7.1-cuda12+250814.

=============================

  Removing previous downloads...

  Downloading master update...

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current

                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:--  0:00:02 --:--:--     0

  Downloading worker update...

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current

                                 Dload  Upload   Total   Spent    Left  Speed

  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0

  Done.

 

 Update will now be applied to the master installation,

 followed by worker installations on other nodes.

 

CryoSPARC is not already running.

If you would like to restart, use cryosparcm restart

  Deleting old files...

  Extracting...

 

gzip: stdin: unexpected end of file

tar: Child returned status 1

tar: Error is not recoverable: exiting now

I also tested with only downloading and then applying, which results in the same error.

It appears as the master/worker and is not downloaded correctly. I also tried to follow what was suggested previously by downloading a specific version of the worker and master:

curl -L https://get.cryosparc.com/download/master-v4.7.1-cuda12+250814/$LICENSE_ID -o cryosparc_master.tar.gz

curl -L https://get.cryosparc.com/download/master-v4.7.1-cuda12+250814/$LICENSE_ID -o cryosparc_worker.tar.gz

The two tar files appear to be empty and faulty:

 tar -xf cryosparc_master.tar.gz

tar -xf cryosparc_worker.tar.gz

tar: This does not look like a tar archive

 

gzip: stdin: unexpected end of file

tar: Child returned status 1

tar: Error is not recoverable: exiting now

tar: This does not look like a tar archive

 

gzip: stdin: unexpected end of file

tar: Child returned status 1

tar: Error is not recoverable: exiting now

I would be happy if someone could help me!

Best,

Ole

2

@OleUns Please can you try the sequence

cryosparcm update --version=v4.7.1-cuda12
cryosparcm patch
3

@wtempel Thanks a lot, that worked!

I am now running into the next issue. We have a bunch of workstations that are all serving as worker nodes attached to a single master. This is quite a nice setup since it reduces the idle time of each workstation.

Unfortunately, the new workstation, with the CUDA12 version of cryoSPARC, causes a version mismatch when attached to our master.


Traceback (most recent call last):
  File "cryosparc_master/cryosparc_compute/run.py", line 79, in cryosparc_master.cryosparc_compute.run.main
AssertionError: Version mismatch! Worker version (v4.7.1-cuda12+250814) and Master version (v4.7.1+250814) are not the same. Please update.

My way out of this would be to update all workstations to the CUDA 12 version. Does the CUDA12 version work with Turing, Ampere, and Ada GPUs (in theory it should, the CUDA version is compatible in my understanding), or is there another way around this issue?

Best,

Ole

4

That is the only way out as far as I know. If there are many workers, you could share a single cryosparc_worker/ directory via nfs, but the mount point/absolute path would need to match

  1. between all workers and
  2. with the absolute path of cryosparc_worker/ at the time of installation

Given the various ways in which compatibility could break, hopefully some users of Turing cards can weigh in here. At minimum, v4.7.1-cuda12 requires that GPUs are compatible with driver version 570.26.

5

I went that route, and it worked.

From my very limited testing (Patch motion, Patch CTF, denoising, 2D, 3D, Ab-initio, Nu-ref, Homogenous-ref, hetero-ref, and 3DVA), it appears to work on Turing, Ampere, Ada and of course Blackwell. In case I notice anything weird, I’ll let you know.

Regarding the workstations, they run either Ubuntu 22.04.5 LTS or 24.04.3 LTS, alongside Nvidia driver version 580.95.05 on all workstations.

Since Nvidia will not move backwards and RXT10XX and older GPUs will slowly phase out, I assume this cryoSPARC build will become the default at some point.