Prime95 failing after 45 mins?

[supster]

I came back after leaving it running for about 5 hours and saw this:

[Sat Mar 15 10:03:54 2003]
Self-test 10K passed!
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.


What could my problem be? Its an Athlon XP 2400 @ 166x13 = 2.17ghz

I'm still using defualt vcore (1.64). Should I give it some more juice? Idle temp is 35C and load is about 45C.

I also noticed some skipping of frames in 3DMark after about o/cing to 150mhz, heres a copy and paste from a post I made on another forum that will explain that better:

----------------
Heres another update:

I finished CPU overclocking and here are my results:

11748
13x166

11467
13.5x155

11047
14x150

11122
15x140

I was hoping to break 12k but am 150 shy of it =( After about 150 FSB I srated noticing sime frame skipping in some of the tests, specifically High Poly Count (8 lights), Envoironment Bitmap Mapping, DOT3 Bump Mapping, and Vortex Shader. I'm still running defuatl VCore (1.64), should I up it? Could that be a reason that I'm skipping? This is the exact same thing that happend to me before (read up a couple posts) exceptits not on the nature tests.

Also, should I be locking the AGP speed? If so what should I leave it at?

After a full 3DMark run and about 20 mins of Prime95 I'm running at 46C, I hope thats cool enough

PS: Vid card overclocking comes next =)
-----------------

Any ideas?

I was just running 3DMark 2001 and got an auto reboot in the Dragon game at high quality. Mem timings at 7 3 3 2.5 (kingston ddr333). Asus Geforce4 Ti4200 and an Epox 8RDA+ for the board.

I just ran 3DMark again, no reboot scored 11833

What I really want to know is if I need to up the voltage to anything more and if I should lock the AGP speed, if so then what at?

 

[John]

Yes, raise the core voltage to 1.7v and run prime95 for 6+ hrs w/out erroring. If you still error, bump it to 1.725, 1.75, etc.