Motion correction failed with pixelsize smaller than 0.8A

Motion Correction
1

Hello Everyone, we are running cryosparc 4 on a computer with 4 NVIDIA GeForce RTX 3090 cards. The movie stack is 5760x4092x50 in size and the pixelsize is 0.68Angstrom. The motion correction failed with an error pycuda._driver.LogicError: cuModuleLoadDataEx failed: device kernel image is invalid - error : Binary format for key=‘0’, ident=‘’ is not recognized.

If I change the pixel size to 0.825 Angstrom, the motion correction worked. Does anyone have any idea how to solve the problem?

Thanks.

Liguo

2

Am I assuming correctly that the pixel size was changed by repeating movie import with a modified Raw pixel size (A) parameter?
Please can you post the output of these commands (replacing P99, J199 with appropriate project and job IDs)

csprojectid=P99 # replace with actual project ID
csjobid=J199 # replace with id of the extract job
cryosparcm cli "get_job('$csprojectid', '$csjobid', 'job_type', 'version', 'status',  'params_spec', 'errors_run', 'input_slot_groups', 'instance_information')"
cryosparcm eventlog $csprojectid $csjobid | tail -n 50

for the following four jobs:

  1. movie import with 0.68 Å pixel size
  2. subsequent (failed?) motion correction
  3. movie import with 0.825 Å pixel size
  4. subsequent (successful?) motion correction