Prime95 test repeating already completed FFTs, only on one core

[bobjane]

CoreCycler v0.11.0.3
CPU: AMD 5500GT

Config:
Prime95
Auto runtime
All FFT sizes

One particular core never seems to complete Iteration 4 (at least not within 6 hours). All the other cores completed Iteration 4 in around 2.5 hours IIRC.

No errors and no WHEA errors.

I have repeated this test on this core 10+ times at decreasing undervolts and finally with all BIOS settings stock. The behavior is the same - Iteration 4 runs indefinitely.

Looking at the logs, it seems the "The number of FFT sizes already tested" counter stalls due to it repeating FFT sizes that have already been completed. The "All passed FFTs" list grows while the "All unique passed FFTs" list stops growing.

CoreCycler_2026-05-15_15-38-42_PRIME95_SSE.log
Prime95_2026-05-15_15-38-42_SSE_ALL_FFT_4K-32768K.log
config.ini.txt

[sp00n]

I've come across this behavior before when 2 threads are selected, after the first full iteration Prime95 starts to randomize the test order of the FFT sizes, and at some point the two virtual cores seems to deviate a bit too far from each other and so the completed FFT sizes cannot be correctly detected anymore.

You could try to set restartTestProgramForEachCore = 1, which should work around the issue, as this prevents Prime95 from randomizing the FFT test order.

[bobjane]

Thank you I'll give that a try.

Does the fact that 1 core does this while the other 5 cores behave differerntly infer anything?

[sp00n]

I can't really say.

I might just need to improve the detection algorithm. Maybe there could be some clock stretching involved, although normally this should affect both virtual cores.