First would like to share what I saw in Turbo V few days ago. CPU Aux 3.3V goes to 3.6V when left on [Auto] when OC'ing.
This has made me now set all voltages manually.
Secondly I'd read recently members sharing that P95 set to use high RAM is better test than RAM tests. Personally for lengthy while I have thought it is best to do round of all tests if wanting high stability. I have seen x program last for y length, then you fire another program and boom the OC profile falls flat quickly. Each test "we" may do, the HW/settings used maybe more or less sensitive to, so doing various tests may help evolve settings better/quicker.
As show before in a recent post I had passes of 3400MHz 4x8GB in 2 versions of P95 greater than 8hrs. I could also roll down SOC based on those tests if did not run for longer length. In that same post I also shared HCI test screens. One instance was a rerun on same POST of system. Why I do that is once I noted on my rig that after say passing HCI for quite decent length (>4hrs), the moment I rerun it on same POST I had errors. I repeatedly saw this on same settings, so now not only do I rerun tests on differing POSTs but also same POST.
Seems as if GSAT was exposing lack of SOC voltage. I tried tweaks of looser timings and or ProcODT/CAD Bus changes plus VDIMM/VTTDDR. I still had fails in GSAT, SOC of 1.081V seems to have fixed setup. Below is video of end of test plus settings used.
Settings used txt:-
View attachment 6301_4.1_3400GT2_v4_setting.txt