Confused about prime95 test

[Spike]

Hi guys, I did some looking at other posts about prime95 but did not see a completely relevant post, so here it goes...

According to the stress.txt and most others here, the computer usually fails within the first few min if there is a problem. I ran prime95 today on my 2500+ @2.2ghz and 400fsb and it had a fatal error... but it was after 1 hour and 35min. The first 4 tests ran fine. I have had this machine running at these speeds for a little over 2 weeks and have done extensive gaming with crashes occuring only in postal 2 (according to the game website the bug is a known one). I was curious how serious the problem is if it took that long before it crashed the machine?

Tonight I am going to run memtest only because I have no idea how long it is going to take. If that comes up clean than I am at a loss. Thanks for any suggestions

-Spike

 

[WobbleWobble]

For me to accept my O/C as stable, I'll run Prime95 for at least 12 hours. I've had it fail on me after the 5th hour...

How serious the problem is? I'd rather have 0 CPU related crashes instead of a few more MHz.

 

[ntrights]

Originally posted by: WobbleWobble
For me to accept my O/C as stable, I'll run Prime95 for at least 12 hours. I've had it fail on me after the 5th hour...

How serious the problem is? I'd rather have 0 CPU related crashes instead of a few more MHz.

Dont take prime too serious there are other ways of measuring stability also. I do run Prime also but playing games like speed underground (need for speed) stresses the whole system in a way that prime95 looks like peanuts in comparision

 

[Spike]

hmmm, I have not tried underground, though morrowind running at 1600x1200, postal 2 at 1024x768, and Halo seem to stress the system pretty well since none of them crash (except for that known postal 2 bug) and I have played for as long as 7 hours in a row.

Thanks for the reply's. I will do the memtest and maybe try a few others (3dmark, sissoft) to see how they preform.

-spike

 

[Jojo1971]

Originally posted by: ntrights

Originally posted by: WobbleWobble
For me to accept my O/C as stable, I'll run Prime95 for at least 12 hours. I've had it fail on me after the 5th hour...

How serious the problem is? I'd rather have 0 CPU related crashes instead of a few more MHz.

Dont take prime too serious there are other ways of measuring stability also. I do run Prime also but playing games like speed underground (need for speed) stresses the whole system in a way that prime95 looks like peanuts in comparision

so how stable is your 2.4c at 3.63? can you tell us more about your setup?

2.4c@3.63 302FSB (v1.6 cpu-z)
P4C800EDLX Bios1014 Prescott Ready
2X256 Mushkin L2 2-2-2-6

 

[WobbleWobble]

Originally posted by: ntrights
Dont take prime too serious there are other ways of measuring stability also. I do run Prime also but playing games like speed underground (need for speed) stresses the whole system in a way that prime95 looks like peanuts in comparision

While Prime95 doesn't stress everything, it does stress CPU (and memory) so you can easily isolate errors.

 

[ntrights]

Originally posted by: WobbleWobble

Originally posted by: ntrights
Dont take prime too serious there are other ways of measuring stability also. I do run Prime also but playing games like speed underground (need for speed) stresses the whole system in a way that prime95 looks like peanuts in comparision

While Prime95 doesn't stress everything, it does stress CPU (and memory) so you can easily isolate errors.

True...but you can pass 7hours error free prime but get errors in games such as underground! Thats why i mean if you are going to test system stability prime95 is not enough IMO.

 

[Duvie]

It usually is 2 things heat or power.....BUt in the heat realm it can be heat and/or power at the cpu. ram, or NB chipset....

The fact the gaming doesn't tank isn't surprising since yes the games stress it more thoroughly throughout the system but the fact is most gaming cards are so powerful now they don't offload as much to the cpu. I think prime is loading the cpu more thoroughly...I would rule out power problems as running a game is a better stress on power IMO since card is blazing 3d, sound card blazing, and cpu....I would loo at temps and or just notch it up a .025 of vcore and see if that helps...Sometimes it may be just a brief second where the vcore drops to a certain level that may cause it to error out..

Now lets talk about what kind of error?? If you are getting fatal error or program error I would check this as a possible software issue..remember version 23.7 has known bugs and you should use an older version on version 23.8.....

NOrmal errors of things I am talking about above would be like rounding error, hardware error, and sumout error and they would be listed in the scroll of text and not cause the app to shutdown...

 

[Duvie]

Originally posted by: ntrights

Originally posted by: WobbleWobble

Originally posted by: ntrights
Dont take prime too serious there are other ways of measuring stability also. I do run Prime also but playing games like speed underground (need for speed) stresses the whole system in a way that prime95 looks like peanuts in comparision

While Prime95 doesn't stress everything, it does stress CPU (and memory) so you can easily isolate errors.

True...but you can pass 7hours error free prime but get errors in games such as underground! Thats why i mean if you are going to test system stability prime95 is not enough IMO.

I agree that is why I do encoding apps, rendering apps, and basically a wide array of applicatio to make sure I am testing all different types of stress conditions....

 

[joe2004]

Prime95 is the ultimate stability test for chipset-cpu-memory. And while there might be some other tests or benchmarks that will fail even if Prime95 passes it is most likely that Prime95 will detect instability before any other test.
That is why it is used. The testing environment changes, I had Prime95 failing after 4 hours or so, it can fail after 8-12 hours as well, all signals instability of the setup. That can happen because room temperature changed for say 5 C up or any other reason but you better believe it - Prime95 fails means that your system isn't ultimately stable, as simple as that.

 

[Spike]

Originally posted by: Duvie
It usually is 2 things heat or power.....BUt in the heat realm it can be heat and/or power at the cpu. ram, or NB chipset....

The fact the gaming doesn't tank isn't surprising since yes the games stress it more thoroughly throughout the system but the fact is most gaming cards are so powerful now they don't offload as much to the cpu. I think prime is loading the cpu more thoroughly...I would rule out power problems as running a game is a better stress on power IMO since card is blazing 3d, sound card blazing, and cpu....I would loo at temps and or just notch it up a .025 of vcore and see if that helps...Sometimes it may be just a brief second where the vcore drops to a certain level that may cause it to error out..

Now lets talk about what kind of error?? If you are getting fatal error or program error I would check this as a possible software issue..remember version 23.7 has known bugs and you should use an older version on version 23.8.....

NOrmal errors of things I am talking about above would be like rounding error, hardware error, and sumout error and they would be listed in the scroll of text and not cause the app to shutdown...

my error was a fatal error after 1 hour 35 min. It said something like "result = .499blah and known answer is <.4" then the program stopped it's testing. I just d/l prime last night and am running 23.7, so maybe that is the case. I guess voltage could be a problem as I did not adjust my voltage when I changed the fsb from 333 to 400 on the 2500+. My mobo monitoring utility shows my voltate staying almost exactly 1.65 all the time with very little variance, even when I am playing games. Does that mean anything?

-spike

EDIT* The temps, according to the utility, never went above 47c during the prime test.

 

[Duvie]

Yep that fatal error is a rounding error...I mistakenly thought you were crashing with a fatal program error back to the desktop....This error is not a hard freeze...

This in most instances is related to cpu...so temps look good raise the vcore up a notch and tr again...definitely get the 23.8 version or use an older version 22....

 

[dnuggett]

. It said something like "result = .499blah and known answer is <.4" then the program stopped it's testing. I just d/l prime last night and am running 23.7, so maybe that is the case.

Did you read Duvie's post before yours? He just typed that this error was not due to the program version.

 

[Spike]

sorry about that... I read it but somehow I interpreted it differently. I am going to d/l the other version and give my voltage a small bump. thanks for the advice

-spike

 

[morbie]

I'm having the same sort of rounding error problem when i run the prime95 torture test, it failes after about 13 min. Im running an athlon XP 2000+ at 1666 Mhz (1800Mhz max), not overclcoking at all, and the external clock is running at 133 Mhz. My motherboard is an ASUS A7V333 which reports my cpu temp at 48C idle, MB temp at 32, and Vcore is at 1.792, +12V is at 12.128, +5V is at 4.945 and +3.3V is at 3.344.

 

[screw3d]

Originally posted by: morbie
I'm having the same sort of rounding error problem when i run the prime95 torture test, it failes after about 13 min. Im running an athlon XP 2000+ at 1666 Mhz (1800Mhz max), not overclcoking at all, and the external clock is running at 133 Mhz. My motherboard is an ASUS A7V333 which reports my cpu temp at 48C idle, MB temp at 32, and Vcore is at 1.792, +12V is at 12.128, +5V is at 4.945 and +3.3V is at 3.344.

Could be faulty RAM. Try running Memtest86 to check!

 

[Thor86]

Prime95 is a good utility to test for stability, but should be used with other software such as 3dmark loops and memtest.

I usually follow these simple (albeit time consuming) rules to system stability.

1) Memtest your system at stock settings to eliminate bad ram/incorrect bios settings
2) Memtest your system at oc settings again.
Whenever running Memtest, make sure you select all tests, cache on, all memory for 10 passes minimum without one error.
3) Prime95 test (2 instances if using multi processors/hyperthreading) and torture test for at least 24 straight hours without failure/errors. I've had Prime95 fail on me after 12+ hours.
4) Loop 3DMark for video/cpu/memory stability for at least 2+ hours, or better yet, play your favourite games for multiple hours (all in the name of stability testing of course ).
6) Use whatever application/software you regularily use to further test for specific stability, whether it is video/audio encoding, crunching SETI, etc.

If you can pass all these with flying colours, then I'd say your OC'd system is stable, and your hardware is running on par.

 

[govtcheez75]

I don't think that most people are using Prime95 correctly. Most people don't even know that it requires a "password". You have to enter the password, then run the Large-FFT stress test under priority 10. Running it at priority 10 will render your computer pretty much useless for the duration of the test, but it will detect many more errors than anything else I've used. I just got done running my Athlon64 overclock for 49 hours...which is pretty much when I determine that it's "rock solid stable".