NU-refine runs too slow after using particles from local CTF refine

3D Refinement
1

I just run two NU-refine jobs, one is use about 286,806 bin1 particle (600 box size) with all parameters of NU-refine in default, the initial volume just from ab-initio refine, the another job use all 286,806 bin1 particles (600 box size) from local CTF refinement with other inputs and parameters keep the same.

  • the first NU-refien takes about 2h40min to finish. (3.3Å)
  • the second NU-refine takes about 55h40min to finish.(3.27Å)
    Why the consumed time changed so much?
    image
    image1920×686 92.6 KB
2

Do you have Transparent Huge Pages enabled? For me, that was the cause of a lot of jobs taking significantly longer than expected, seemingly randomly. How about CryoSPARC server topology? Were they both run on the same node and GPU type? Was local CTF refinement carried out inside the NU refinement (that will make it slower)?

One of the CryoSPARC team may ask you to share the job logs with them (privately) to help diagnose. :slight_smile:

Welcome to the forum. :smiley: