Quote Originally Posted by drnick View Post
So I have a scenario with my 2500k where at 4.9ghz and 1.385 vcore the system is linx stable (20 runs, all memory used, SP1 not loaded), OCCT blend stable (I let it run 4 hrs then stopped it), and memtest86+ stable (I let it run for 5 complete passes then stopped it). It fails prime (version 26.4) blend, usually within 1 hour with a stop 124 blue screen. It also crashes with handbrake video encoding with a stop 124 blue screen.

I increased vcore to 1.395 and now it encodes video without errors so far, but still bluescreens with prime blend. It also bluescreens and freezes in games at this increased vcore.

At 1.415 vcore I'm still crashing with prime blend.

So what is the logical next step if I want to keep running at 4.9ghz? Keep increasing vcore until I'm prime blend stable? Or would prime small fft's be more diagnostically useful? My memory (2 x 4gb gskill sticks) in all cases is at 1600mhz with stock timings and voltage.

Attempting to achieve overclocked real world stability with these Sandies is taxing my feeble brain. With previous generation intel chips, it seemed to me that linx and OCCT stability were usually within a couple of vcore notches of prime blend stability, and I don't recall ever having stability issues if I could prime blend for 12+ hours without errors. I wonder if I should still be aiming for prime blend stability as my gold standard of real world stability. I did try loading up SP1 and using linx with avx, but my temps on air cooling (mid 80's) made me uncomfortable and I stopped before achieving the requisite vcore to be linx + avx stable.

I don't want to concede defeat yet at 4.9ghz!!!
FWIW, I'm on a Biostar TP67XE and I was using offset vcore with LLC off and just kept having to add vcore to get prime blend almost stable. Seemed like it would never end.

When I switched to FIXED vcore and LLC on I was able to get prime blend stable at 4.8GHz 1.416V load.

Of course, now the thing randomly blue screens at idle.