so then not enough vcore...
interesting behavior. Something i wouldn't see on my phii. I wonder why it does that when undervolted.
As stated, it was a theory that low Vcore might contribute to workers falling far behind in completing tests by core.
"mandrake4565" was theorizing that it was caused by low Vcore and I was testing known "enough Vcore" and was not seeing test completed variance in the same time frame that varied by 10 or more completions per core.
"mandrake4565" reported back in "SouthernStyle" s thread after testing his known good P95 Blend testing but doing so with lowered Vcore did in fact fail the P95 test but during the time the test was running before failure, the number of tests completed per core was NOT actually varying. From this "mandrake565" concluded the problem of a few workers completing P95 Blend tests with a great difference in the number of tests per core do not have a huge difference that is caused by low Vcore.
So actual testing has not progressed further. User "SouthernStyle" has not posted any further in his thread where he had workers completing 10 tests and other cores closer to 30 tests during the same time period of testing.
I have come to believe in "SouthernStyle"s case he may have a bad cpu with such huge variance in number of worker tests completed during the same time frame. I have not voiced that in his thread yet. I have about come to the bad cpu conclusion after researching number of tests completed in one time frame varies per core. The number of completions should not vary so greatly, is the conclusion I have come to.
I see most understand that the first core may complete less tests in a single time frame since it maybe having cpu cycles stolen due to other overhead. That makes sense actually. Once getting beyond that consensus, you suddenly get ideas that are all over the place. Most suggest running P95 Blend by itself with all background uses DISabled so that cpu cycles are not stolen from P95 Blend testing. On the surface that would make good sense, but in my particular system, not DISabling other cycle stealers, still does not cause a huge variance in the number of tests each of my 8 cores is able to finish in the same length of time.
In closing I don't have "SouthernStyle"s issues. He has not reported into his thread where he voiced his concerns of the number of test completions varying so greatly by core, over the same period of time. Thus I guess from my perspective my testing of such an issue is over.
RGone...
Dr Mandrake out 
