Quote Originally Posted by icecold316 View Post
I was able to run 3dMark06 @ 4 Ghz. But Prime would fail in less than 30 mins.... lol. I guess my batch doesnt like 4+ mark.

But my rig is Prime stable at 3.9 on 1.25V Load with QPI : 1.3125.
Sigh, Prime blend stability is tricky on high Nehalem clocks if you are trying to run the memory fast also. You may need to tweak timings, add LOTS of VTT voltage or in some combinations of settings increase Vcore by one notch past what's stable in LinX! I think the cores can show instability once you actually feed them enough data to saturate them. This may be a Vista/Windows-7 64-bit thing only.

I think I was able to get 4.2ghz to pass LinX x 20 on air only because I was running the Uncore at exactly 2x the memory which chokes off performance slightly as shown in my test data. I could bench all I wanted to with 3400mhz and 3600mhz Uncore @ 4.2ghz and I saw gains at both of these increased Uncore clock speeds. It was very hard to get LinX stability @ 4.2ghz on air so I didn't even want to deal with the Prime blend crashes after an hour in again. I will try Prime again once I get my first WC loop operational.