Given the refinement that Intel did as well as the changes to the die makeup, their attempt at improving cooling as well as getting these chips to 5.3Ghz gives us a good crack at trying to see if these CPUs can do 5.3 at all-core load. We’ve had no luck pushing our Core i9-1090K to 5.3Ghz but we did find that most of our samples (we had two 10700Ks and a Core i9 10900K) can do 5.1Ghz although pulling that off with the i9 at decent temps was a challenge. For the i7, it was tad bit easier. We’ve delided and overclocked a couple of 9900Ks and the rules was pretty much the same for the Comet Lake CPUs. We had an ROG Ryujin 360mm cooling our CPU and a nice large 10ml syringe of Thermal Grizzly Kryonaut from Paulo Avelino (shoutout!). We’ve changed our loading scenario of instead of running AIDA64 stress test, we decided on Derbauer’s 12K Prime non-AVX load to verify stability and temps. I insisted on AVX stress testing before as I believed that games development would eventually adopt AVX512 but with AVX2 or so still prevalent, a offset of -2 or -3 should be enough if you can’t do stable no AVX offset.
Going back to the 10700K, Our first sample could do 5.2Ghz at 1.42v but that introduced throttling already. We went back to settling with 5.1Ghz and saw one of our samples capable of holding 5.1Ghz 1.3v and our other sample with 1.35v. If this variance in voltage is presence in retail chips, some folks can still settle to 5Ghz. Our sample better sample did 5Ghz at lower voltages but we’re chasing numbers here.