Originally Posted by davidm71
New F8 bios out for the Master. Anyone try it out yet?
Tried it out and went back to F8E. Seems like some bugs were probably fixed (I haven't tested the issue I saw with F8G where when you flip on the power supply, the board 'warms up' for 3 seconds then shuts off, also didn't test that F8G issue with F8H either, but F8E did not do that), but I am "slightly" more unstable at my "minimum stable voltage" (same ambients) with F8 compared to F8E (F8H seems to be the same as F8).
Hard to test stuff like this without equipment but: F8E: 5 ghz, 1.265v, LLC Turbo, Realbench 2.56: one WHEA/L0 error after 1 hour 10 minutes. F8: 5 ghz 1.265v, Realbench 2.56: one WHEA/L0 error after 15 minutes (no crashes, just logged in HWinfo64).
Power draw and VR VOUT seemed to be identical; usually stuff like this that goes pear shaped when you are borderline stable can do with RAM subtimings, especially where AVX instructions are concerned, as I didn't notice anything different on 15k small FFT's at all. But I don't have the the time nor patience to spend hours and hours testing stuff like this. Also voltage fluctuations from higher loadline calibration and load changes can cause failures when you are borderline to be very erratic and happen sometimes fast and sometimes it takes forever. Probably more effective to drop the LLC to high or medium (but this requires a much higher bios voltage, but lower LLC=much better transient response than LLC=Turbo) to get the same VR VOUT, then go down 5mv until you start getting errors more predictably. Very torturous stuff and I can't be bothered with that crap now.
When I get my delid kit (ordered one this morning), and I delid and LM the core and relid, hopefully with reduced temps, I can drop the volts then I can see if I can test more precisely. It's really not fun when it takes ages (and heat) to find an error. And I know very little if anything about RAM subtimings or tertiary timings.