Weird Prime95 Results

[capreppy]

I got 4x2048 G Skill (DDR2 1000) in the other day and plopped these into my current rig. I had previously confirmed my previous memory OCZ Plat Rev 2 DDR2 800 (4x1024) were stable @ 3.51.

Silverstone TJ09
Q6600 @ 3.51 (I worked up to this from 2.4 with the new memory, but not sure if I saw the results below).
GA-P35-DS3P
Corsair 620HX

I've been running Prime 95 (small FFT) for just over 8 hours. I checked this morning and I'm seeing very weird results. At least I think they are weird.

Prime95 is still running and has not died.

Results @ 7:32am 12/26/07
Worker Thread #1: Passed 10K Self Test and is on Test #6 using FFT length 12K
Worker Thread #2: Passed 10K Self Test and is on Test #7 using FFT length 12K
Worker Thread #3: Passed 10K Self Test and is on Test #2 using FFT length 12K
Worker Thread #4: Passed 8K Self Test and is on Test #1 using FFT length 10K

Shouldn't the threads be processing at the same speed? I assume since I am getting different Worker Thread progress, they are working at different speeds? Or is the fourth CPU doing other things causing it to make less progress?

Also, I noticed that the worker threads are running different numbers of self tests. WT#1 did 12 self tests @ 8K FFT (completed @ 7:05am)
WT#2 did 3 self tests @ 8K FFT (completed @ 7:03am)
WT#3 did 5 self tests @ 8K FFT (completed @ 7:15am)
WT#4 did 4 self tests @ 8K FFT (completed @ 7:33am)

Is this a memory issue or a CPU issue?
Or is this normal Prime95 behavior and I shouldn't worry about it

 

[capreppy]

I pulled the memory and put the old memory back in the box. Prime95 works fine and windows doesn't take nearly 5 minutes to boot.

I RMA'd the memory and have a new set enroute. Hopefully this next set is a little better.

 

[hazard_one]

i would actually like to know. i have the same results. Prime95 isnt failing but the threads work at different speeds

 

[capreppy]

I think it was some issue with the memory. My wife and I just moved and I can't find any blank CD's to burn MemTest and run it.

My system was doing a lot of other weird things
- it was taking up to 5 minutes to boot into Vista x64 when it only took seconds previously
- Prime95 ran at drastically different speeds. Even with small FFTs some memory is utilized and I have to think the memory had something to do with it. With the old memory in place, Prime95 runs much more consistent in speed.
- Prime95 ran inconsistently with different numbers of tests per iteration. Again, I think bits were getting "lost" the circular file in the heavens. With the old memory in place, this went away.

Depending on how drastic the difference betwen workers, it may be that you have a program or two that is taking CPU cycles away from Prime95.

I switched out the 4 x 2048 G Skill DDR2 1000 (CL5) for 4 x 2048 G Skill DDR2 800 (CL4). I can't hit 500FSB anyway so for $20 more, I was able to get the faster chips. I'll let you know if I run into this issue again. You may want to see if the issue is memory related on your side. I'm pretty confident it was for me.