Troubleshooting with Prime95

tecninja

Distinguished
Feb 2, 2008
60
0
18,630
Short story is I keep getting errors with Prime95 when trying to figure out a new system of mine. I'm not actually overclocking anything yet, but I figured the OC forum would probably be the most knowledgeable on such issues.

Anyhow, the computer seems to be happy with the Small FFT's stress and has run for hours and hours with it with no errors, but on the blend test it gets a failure anywhere between 1 minute and 3 hours. Now I've tried a bunch of different things, and none of them seem to have any effect on the results at all.

ASUS P5E MB, C2Q Q9450 CPU, 4 GB Crucial Balistix RAM (2x 2GB)

1.) I've relaxed the timings on my RAM (down to 5-5-5-18 2T), and also even moved it down to 667 MHZ (even though its 800 MHZ RAM).

2.) I've upped the voltage on the CPU a decent amount (up to 1.35v core)

3.) I've upped the voltage on the North bridge from 1.25v to 1.31v

4.) I've upped the voltage on the RAM from 1.8 to 2.0v

None of these seems to have an effect at all on the stability of the system. It always passes the Small FFT, it always fails the blend. I'm wondering what might be the culprit. I have run Memtest86 for over 48 hours with my RAM set at (4-4-4-12 1T, 400Mhz) and it ran clean. I was thinking possibly dirty power from the PSU, but wondered why it wouldn't also fail the Small FFT test as well? I guess i'm a bit stumped as to what might be causing the rounding errors, and I don't have very many parts to swap out to play the "try it" game.
 

Mondoman

Splendid
Sounds like a memory issue. Memtest86+ isn't as hard on the RAM as PRIME95's blend mode. A number of people have been having problems with Ballistix recently. First, figure out what the spec voltage is for your specific modules (likely 2.0V or higher), then set your RAM voltage to that value. Do NOT increase your CPU voltage -- you'll just mess things up and possibly damage your CPU. Leave the north bridge and other voltages on auto (except the DIMM voltage, as noted above). If the RAM still gives errors in PRIME95, get it replaced under warranty.
 

JDocs

Distinguished
Apr 2, 2008
496
0
18,790
I doubt the memory needs to be higher than 1.8volts for 800mhz as that is the official spec, however for 1066mhz 2.0 to 2.2 is quite common (unless set back to 800mhz).
 

tecninja

Distinguished
Feb 2, 2008
60
0
18,630
I actually set the RAM down to 667mhz for testing, and that didn't help either, so even at 667mhz 5-5-5-18 2T timings at 1.9v (up from 1.8v default) it seems the same.

What I find Odd is, I would expect borderline RAM to come into line if you reduce its specs by that much. I would think just plain old "Bad" RAM, would error like Heck in Memtest86 when pushing it at 4-4-4-12 1T 800mhz like I did for 48 hours.
 

JDocs

Distinguished
Apr 2, 2008
496
0
18,790
Then it sounds like the RAM. I actually had a RAM chip that would pass memtest and run XP with no problems but made Vista as unstable as hell. A short while later it started to flare up in XP as well. Perhaps yours is similar? Starting to fail but not completely there yet?
 

Mondoman

Splendid
As I mentioned, Ballistix seems to have been making a higher-than-normal number of dud DDR2 DIMMs recently. JD, you're correct that all "DDR2" RAM *should* run at 1.8V (at some speed/timings). However, the RAM manufacturers have gotten so caught up in selling factory-sanctioned overclocks that it can be hard to find a module rated at 1.8V these days. All too often, these higher-voltage factory-sanctioned overclocks don't work quite properly at 1.8V.
 

tecninja

Distinguished
Feb 2, 2008
60
0
18,630
OK, took my PC apart and futtzed with the cooler (not that it should affect my Prime95 issue) and put in a much better Power supply (Silencer 610w rather then the 500w Generic one). I was waiting to get the PSU back since I had to RMA it originally. Anyhow i'm rerunning tests with the new PSU in, just in case it was some crazy "dirty power" issue.

Also messing with the cooler lowered my temps by about 7 degrees C, but they weren't that bad to begin with (was 55c max under full load and 25c ambient room)
 

tecninja

Distinguished
Feb 2, 2008
60
0
18,630
Well i'm not sure i've "Fully Tested things" but since I swapped out the Power supply, I've run tests 3 times for 10 hours + and Prime95 has not given me any errors. Previously I was unable to go longer then 3 hours, and a few minutes was more likely to get an error. So while i'm not positive that the PSU was the only problem, it seems like it was at least contributing to my issues.

PSU's really are a pain, there is no real way to test them properly at home. The equipment needed to truly give it an evaluation just isn't available to your average person. So many companies are churning out crap "generic" PSUs too. Guess this is more reason for me to stick with name brands like PC P&C going forward.