Long story short... I have four 8086Ks. I get to pick one to keep for delidding and putting in my main rig. Thought picking one would be easy. Turns out I am mistaken in that thought.
Board is a Maximus X Formula. Ram is 2x 16GB G.Skill Trident-Z RGB 3200MHz CAS14 (yes RGB puke move along
). PSU is 700W Seasonic Titanium Prime. Cooler is a Noctua DH15 with both fans on 100% and a fan blowing over the RAM and VRM.
I set up bios using XMP and disabling all the multicore and adaptive voltage crap and set cache to 4.7GHz for all tests. Basically followed der8auer's Maximus IX Formula video. Set vccio and system agent to 1.2375V and left it there. Will try and get those lower once I get into the final config. Using prime95 v29.4 with AVX disabled as when I started this process I didn't have my liquid loop going. I'll go for full AVX stable once delidded and under water. Basically run 1344 in place until its stable for an hour. Then switch to 24 hours of the default prime95 tests, finding blend being the toughest to get 24 hour stable.
Note: voltage recorded are the setting in bios, not actual read by software. In all the testing I didn't see the reported voltage change between them so I didn't bother to record it. And temps are not directly comparable, due to space constraints my test bed kept moving around between tests.
First round: 5.0GHz
CPU #1 - 5.0GHz @ 1.29V (forgot to record temps)
CPU #2 - 5.0GHz @ 1.25V (71 76 83 79 78 78 max temps)
CPU #3 - 5.0GHz @ 1.24V (75 70 75 73 71 72 max temps)
CPU #4 - 5.0GHz @ 1.25V (74 72 76 74 72 72 max temps)
So at first glance CPU #3 is the best chip and I should keep it right? But I wasn't comfortable with how close the results were and the fact I kept moving the computer around so the ambient conditions weren't always the same. But CPU#1 is clearly the worst so I dropped it from round 2.
Second round: 5.1GHz
CPU #2 - 5.1GHz @ 1.32V (87 81 89 85 83 83 max temps)
CPU #3 - 5.1GHz @ 1.31V (85 80 86 84 80 83 max temps)
CPU #4 - 5.1GHz @ 1.29V (79 75 81 79 78 77 max temps)
Ah, so now CPU #4 pulled ahead! At least... I think so? During the 5.0GHz testing the CPU was more prone to hardlocks vs just causing errors in prime95 than CPU #3. Think that means anything?
I have my water loop going again. EVO Supremecy and a big ass 560 rad. Should I try testing @ 5.2GHz before delidding?
What are everyones thoughts on my testing and results. Go with CPU #4? Or the less prone to hard locking #3? Or do some different testing methodology?