search

sp00n / corecycler

Script to test single core stability, e.g. for PBO & Curve Optimizer on AMD Ryzen or overclocking/undervolting on Intel processors
Other
663 stars 30 forks source link

Errors only present when more than 1 core selected in config. #45

Closed chrisvanhorn closed 1 month ago

chrisvanhorn commented 1 year ago

Hello,

I have a very strange issue that I've noticed that I can get a core to pass 24+ hours of validation with a single core defined using Y-CRUNCHER, but if I add additional cores (cores 11, 14, 12, 15, 10m each cycle, for example) and run that for 12 hours cycling between all 4 of them, all of them eventually throw "expected power errors" in Y-CRUNCHER within a few hours.

This seems pretty bizzare to me. Any ideas? Making me question pretty much all of my testing.

An additional example, I have tested Core 15 with Y-CRUNCHER for 24 hours straight at -50 with 0 errors, but if I define Core 11, 14, 12, 15 in the config, it produces errors at -50, -48, -46. Both these scenarios have all my cores with their 12+ hour validated curve optimizer offsets set.

sp00n commented 1 year ago

Hm. It may be due to Vdroop or negative overshoot between changing cores where the errors happen. A higher LLC setting might help in this case.

image

Or maybe it's something else. Unfortunately Y-Cruncher doesn't create a log file itself and also doesn't support pulling its output into CoreCycler, so the only way to determine if there has been an error is to check the CPU usage. Which depends on the integrated Windows Performance Counters, which are much more fragile then I expected them to be. You could take a look at the CoreCycler log file, maybe there's a pattern you can identify.

Or maybe you could try to set stopOnError to 1 and wait for an error to happen, the Y-Cruncher window with the message should still be open then.

sp00n commented 1 month ago

Closing due to inactivity