Originally Posted by Invisible
Alrighty, just put it at 1.35v. Do I need to do normal Blend, or do 1344, or what type of test?
I'm actually more interested if you will pass prime at the original vcore, (instead of crashing in APB), although testing that would waste your time. As far as blend vs small fft's, what you need to know is if you're getting a little vdroop with some amount of Loadline calibration applied (as long as your LLC reduces vdroop, instead of reversing it (Vrise); you will usually get MORE vdroop with small FFT than blend, since small FFT hits the CPU cache and instruction sets hard (you always get higher temps witih small fft), and that could be enough to trigger a BSOD, if your vcore drops too low from the droop. I'm not sure how this works on digital VRM's, as their LLC compensation might be more exact, so what I said might not apply to you, but on analog vrms, that applies.
If your old prime setting bsod's, then your CPU is degrading, although at your vcore, it's just minor and should settle down. Has happened to many people here, although many others have been more fortunate. Back when I was following the SB stability threads, I saw a number of people who were around 1.3'ish vcore, who had been stable for a few months, and then started getting BSOD's and had to go up to 1.325'ish and the stabiity returned. Because I've degraded two CPU's with repeated prime testing, I don't prime anymore, unless its just for a few minutes, which will tell me if I can at least run programs. High vcores, max loads and high temps will cause more degradation, and sooner. Found that out the hard way. Temps are MUCH more important now than on 45nm, if you pump vcore.
One folder on the [H] used to fold 24/7 at 5 ghz and 1.5v, and now his CPU won't do more than 4.6.
Usually if you're at 1.25-1.3v and you degrade slightly, you just need a small vcore bump and you're fine long term.