Fatal error at 6 hours, 40 minutes.

[jakebrick]

Hey everyone, first I want to thank everyone for giving me the knowledge to build my own computer, it went really awesome! I got an e6750, a 7900gs, ocz platinum revision 2, antec ea430, WD 250 (7200), and an asus p5k-vm.

Let me begin:
I overclocked it to 3.3 Ghz with stock cooling, and it never went above sixty, and that was fine with me because in a few weeks I might get a zalman or a tuniq, and then I'll try for more. The memory was at 5-5-5-15 (I relaxed it), and prime 95 ran for 6 hours, so I thought that was good. Then I put the timings to 4-4-4-12 (2.0 volts), and I put prime95 to work overnight. It passed 6 hours and nothing was wrong, and then at 6 hours and 40 minutes it failed. (FATAL ERROR: Rounding was 0.5, expected less than 0.4). Now I don't know if it was the memory or or the CPU or what. I really thought it was rare to get an error after 6 hours. Does anyone have any suggestions?

 

[Pdawg]

It's not unheard of. I personally think 6 is too short. I've failed at 7.5 hours before. I at least run 8 - most of the time 12 hours to make sure. You need to put down more info too.

On your first run, leave the ram alone - put it to auto so you can eliminate that as a factor.

What was your v-core, multiplier, and FSB set at when you did your first run? Do you have any kind of Speedstep, C1e or other power savings options enabled? If so, disable. The more info on settings you can give the better it is to analyze the problem.

If you can eliminate the RAM as a factor in the first run and it fails then you know that you need to raise the v-core on the CPU - run again and if you can make it to 12 (the most careful say 24hours) then you start messing with the RAM.

 

[gurusan]

What I usually do when I overclock is to first find the max limit of my memory, then FSB.

First I bring the multiplier on the CPU to lowest ....then find the max memory by using dividers
-Then I know what my memory can do

Second I put memory at 1:1 ratio and leave the CPU at its lowest multi and try to max out the FSB.
-If something fails and it's lower than what I know my memory can do then I know I've hit the FSB limit. Time to up mobo voltages and see if you can raise it.

Lastly I bring the CPU multi up and find the max cpu clock.

By doing it this way you usually know what exactly is failing. It's time consuming but it should get you the best overclock.

 

[jakebrick]

Okay, I will put the Ram to auto for now, and then try it again. All the power savings options were off, and the fsb was 413x8 which is roughly 3.3Ghz. It's 1:1 so the RAM is only at 830Mhz ish.

I heard of this program OCCT, it supposedly does a solid test in 30 minutes? Is this program advisable?

 

[Nagasama]

its not as solid as orthos or prime 95. but its good for a quicke test. i wouldnt base cpu stability on it though.

 

[jakebrick]

So it's mostly for the RAM then? I will run that, and then I will run prime95 overnight again.

 

[jakebrick]

I put the timings too 4-4-4-15 (factory timings), and then I ran OCCT and everything was fine, and then I ran Prime95 for 9 hours and everything was fine. So my final overlock (for now) is 3.3GHZ at about 1.31V and my ram at 4-4-4-15 at 830Mhz, 2.0V. Thanks for the help.

 

[anthegreat]

i have the same probs like u too...but it lasted for like 2 hours....sigh...read my thread....

 

[commandercup]

your supposed to run Prime95 for 12 hours+ if your really checking stability, usually a 30 minute run will find any pressing errors that you really need to fix

 

[Nagasama]

9 hours is acceptable.
i've rarely had them fail when they run over 6 hours actually.
i usually try for at least 8. 12 if you are a real stickler.