Originally posted by: Icepick
Originally posted by: myocardia
Originally posted by: n7
None.
Memtest is not really the best thing out for testing stability.
It is a good basic indicator though.
I prefer to see results from an OS environment, aka Orthos blend
Agreed, although it's best to run one full pass of Memtest, on a new motherboard and RAM.
I ran both when overclocking and burning in my new e6300. Both tests correctly identified the limiter in my overclock to be bad RAM. However, Orthos blend consistently showed errors in about a minute where memtest86 ran for 10 minutes before detecting errors that were produced by my faulty RAM. Therefore, I'd say that Orthos blend is better for testing memory.
1 pass is plenty. I just use it to see if I still have stable RAM settings as I push the FSB on my CPU up. Actually, I usually just run tests 5, 8 and 11 because they seem most sensitive.
You guys seem to be under the impression that Memtest 86 is supposed to be some overall stability testing suite. It's not. The beauty of Memtest 86 is that it helps you narrow down the problem and takes several factors out of the equation. That's something an infinite number of in-Windows Orthos/Prime95 failures won't tell you. Prime 95 failed - now I know I'm not stable. But you don't know WHY. Essentially - you know next to nothing when Prime 95 fails unless you've already tested your RAM at those settings elsewhere.
That's why I don't run the blend tests - ever. I've already done my RAM stability test outside of windows and I don't need Prime 95 to tell me something I already know. I need Prime 95 to stress my CPU - and only my CPU - because I already know my RAM is fine.
Orthos/Prime 95 isn't the end-all be-all either. If you don't do any 3d apps to stress out the rest of the rig, then you haven't really finished stability testing. I usually run all three 3dmark proggies and spend a few hours in various games before I call it stable.