System stable apart from 3DMark'03 1st test

Dutch Guy · 28 Jun 2006 at 10:22

Bear with me, this might be a bit longer than I wanted.

I ran my computer for several months at ~[email protected] and yesterday I set the CPU to [email protected] and ran Prime95's small FFT's to test the CPU, it ran fine for 30 minutes so I was happy as an initial test and ran 3DMark2001 and got a score of 30400 which was what I would expect as 2.7Ghz gave me 29807

I then ran 3DMark'03 but it hard locked at the first test so after rebooting I increased the fan on my rad to see if it was a cooling problem but it locked again (CPU is watercooled so heat isn't the issue) and the videocard is clocked the same at all times.

I do not get the hard lock at 2.7Ghz.

But that leaves me with the question what causes the lockup :confused:

Possible causes:
- CPU overheating (unlickely as the CPU temp is 38C watercooled)
- Northbridge overheating (could be but during Oblivion the temp goes much higher)
- Memory failing (very unlikely as I was using a divider to run at 200Mhz when at 2.7Ghz the RAM runs at a higher divider at 208Mhz)
- 3DMark'03 bug (anyone else get this?)
- Driver issue (again not likely as at 2.7Ghz it runs fine)
- ????????????

Mikey1280 · 28 Jun 2006 at 10:30

-Increased HyperTransport bus speed is too much for it?
-2.7 is limit of your chip?

Dutch Guy · 28 Jun 2006 at 10:51

Mikey1280 said:
-Increased HyperTransport bus speed is too much for it?
-2.7 is limit of your chip?

HTT multi is at 3X so at 11x255Mhz it is still well under 1000Mhz.

2.7Ghz could be the limit of my chip but wouldn't that mean that Prime95 SmallFFT's torture test would fail straight away when it ran for 30 minutes without error?

Is it possible that the 166Mhz memory divider is unstable when the 183Mhz divider isn't?

Mikey1280 · 28 Jun 2006 at 11:18

Sometimes various mem dividers are unstable, this could well be your problem, try the next one down and see if the problem still occurs.

Dutch Guy · 28 Jun 2006 at 11:43

Mikey1280 said:
Sometimes various mem dividers are unstable, this could well be your problem, try the next one down and see if the problem still occurs.

Yeah, I will try that next and if the 150Mhz divider works fine I can always run 10x280Mhz with the 150Mhz divider, just a shame I have to wait until tonight to test it.

Dutch Guy · 29 Jun 2006 at 08:52

Tried with the 150Mhz (3:4) but it produced a nice BSOD in test 3 so I think I will give up and accept my poxy failed dual core 3700+ can only get to 2.7Ghz

Minstadave · 29 Jun 2006 at 11:15

Have you tried running a good few loops of Memtest86+ test 5? Small FFTs might not be picking up on memory errors, whilst something memory intensive like 3DMark won't like it at all.

It does look somewhat confusing, one thing I would suggest is that you set your RAM timings to:

MAL: 9ns
Read preamble: 8ns

It helps stability with 2Gb kits on dividers

I think I've mentioned this to you before, when you were running Crucial Ballistix.

Yeah here it is:

http://forums.overclockers.co.uk/showthread.php?t=17569587&highlight=MAL

Dutch Guy · 29 Jun 2006 at 11:47

Minstadave said:
It does look somewhat confusing, one thing I would suggest is that you set your RAM timings to:

MAL: 9ns
Read preamble: 8ns

It helps stability with 2Gb kits on dividers I think I've mentioned this to you before, when you were running Crucial Ballistix.

Yeah, those are the settings I changed to run on a divider, It can't be the memory itself as I looped Memtest86+ test 5 and 8 for hours at 208Mhz ([email protected]) and yesterday when I got the BSOD the memory was running at 200Mhz ([email protected])

I am beginning to give up, if it takes a lot more testing it's just not worth it for a 100Mhz more CPU speed.