Quote Originally Posted by masterkaj View Post
Been reading around for a solution to this prob.. maybe you guys can help.

I can run Prime 95 "small FFT"/Memtest86+ for 12 hours each no errors. However, once I start up the blend test in p95 it fails everytime at fft length 896k or 50 minutes into the test. So I started messing around with volts/timings and it still failes at that 50 min mark or if I set it to custom for that fft length.

I know its not my cpu... so I figure its either the timings or voltages for NB/cpu PLL/FSB term volt.

Considering I have loosened the timings quite a bit and it still failes at the same point makes me think one of my voltages are off. Here is what I have tried so far.

NB- 1.59 to 1.63v
CPU PLL- 1.58 to 1.66v
FSB term- 1.3 to 1.54v (It used to fail after 1 min when set to 1.54 so I figure this at least helped me reach the 50 min mark.)

I have active cooling on my NB/SB/ddr (4x1 Muskin 8500 ddr 1199 in bios set at 2.28v, which is 2.32-2.34 in everest)

Temps on NB never raise above 46C
Temps on SB never raise above 37C

All my other settings I use were recommended by anandtech guide. Any ideas?
Well first of all I would throw that Memtest86 bootable disk in the trash, as it's pretty much worthless for testing memory OCs. It's really for testing defective memory, as only the most huge errors are detected. I had a set of DDR memory that would pass Memtest86 for 8 hours, but it would fail HCI Windows memtest in 30s. Use this:
http://hcidesign.com/memtest/
It works in Vista as well.
BTW if yo're running that much voltage it's most likely overkill.....also make sure that your NB temps are below 50C when running memory tests, as this is known to cause OC instability. I have a 120mm fan pointed toward that area of my board: