Prime95 BSOD's at stock speeds on i7 920

I upgraded my machine a few days ago, and it BSOD's while running Prime95 at the stock speeds. I've been unable to read the BSOD as it restarts to fast.

I have:
EVGA E757
6 GB Corsair XMS3 - 1600 mhz
OCZ Powerstream 700 watt
Radeon HD4870
3 hd's, 1 DVD-RW

I had to manually set the timings and voltage on the RAM to the advertised rates - 7-7-7-20 @ 1.65.

I originally had the CPU voltage on auto, and so I manually changed it to 1.3, thinking it may be the fluctuations caused by the auto setting that killed it. It still died. I've upped it to 1.306 and am trying again, but I don't expect it to work, especially considering I've read that other people are OC'ing their 920's at 1.25 volts w/ no problems.

Any idea what the issue could be?

Edit: Should mention that it runs Crysis with no problems for long periods of time, if that helps at all.

Edit again: Also - CPU-Z reports the voltage as 1.247
Everest reports temps (under full prime95 load) as:
CPU - 55-56c
Cores1-4 - 70-72c

Why? Like I said, it was on auto when this started.

Have you run multiple passes of Memtest86+ to test for RAM errors? BSOD's are most commonly caused by RAM problems.

I'm running Memtest right now. Seems to be going fine. It's on 65% w/ no errors. I'll run a couple more passes as well.

The CPU, mobo, and RAM are all new. Sorry, should have been more clear

Good suggestion - will try that.

@Selea - I get what you're saying, and while the performance gain between CAS 7 & 9 may be minimal, I did spend the extra $30 to get the CAS 7, so I'd prefer it to work properly.

Thanks.

Remember I also said Prime95 will detect more memory errors than memtest.

I went through and tested each stick at the bios default speed/latencies. They all worked fine.

Then I happened to read that someone was using these at 1.5 volts (vs. my 1.65, which is what newegg and the sticks themselves say to use), so I set them to 1.5.

Prime95 has now been running blend for about 2 1/2 hours. I think they may have been getting to hot or something at 1.65, and are now staying at a safe temperature. Of course, that's just my guess, but I can't come up with a better.

Anyway, thanks for all the suggestions. I'm going to leave it up at the very least until I go to bed in an hour or two just in case.

What "test" did you perform?

So to recap. We have a PC with a new mobo, cpu, and memory that:

- runs Crysis with 2 RAM sticks without failing but fails under Prime95.
- runs Prime95 successfully with either stick.
- has a graphics card and psu that ran fine with their previous mobo, cpu, and RAM.

Some new questions:

- What did your mobo set RAM timings to when on "auto". (You said you *had to* manually set the timings.)

- Did you do a fresh install of your OS from an OS CD (not a "recovery disk" )?

- How old is your psu and do you have another to try?

I guess I'm liking the psu or mobo as the culprit here. And in that pairing most often its the psu.

Have you ever tried just setting everything back to auto, and letting the mobo do what it finds comfortable?

I understand about the memory timings, and the $30, but it would be nice to get it running, and pin down the fact that this mobo/ram can't run as advertised.

So obviously you know where the problem is.

Why not take your findings to EVGA and Corsair?