Prime95 + SuperPI = Ultra Stability

[photogearjunkie]

Hi,

I have a question that I hope some of you guru's can answer. I'm trying to test my system stability by running a Prime95 "Torture Test" coupled with a SuperPI calculation of 32M digits. (P4 3.0C @ 3.6GHz, various voltages; ASUS P4P800-D; 1G Corsair 4000PRO; WD Raptors in RAID; etc...) Prime95 can run for 12+ hours, alone or with other apps going. But when I simultaneously run a 32M digit SuperPI calculation, I usually get errors in one or both of the tests.

My question is this: Should I ONLY run Prime 95 to test the system stability? Or should the system be stable if I run both of these apps simultaneously?

Thanks a lot!

 

[Corporate Thug]

12+ hours of prime is good enough in my book

 

[photogearjunkie]

But if the system is TRULY stable, shouldn't it be able to chug away at both programs without errors??? My only thought is that maybe the two programs conflict with eachother in some way??? At least that's what I'm hoping- then I can o/c even further ; )

 

[Shimmishim]

running prime for 12 hours is good enough

some guy once ran prime for 876 hours and it errored in the 876th hour...

now tell me... would you say his system is stable?


you can also try running superpi if you want, but i'd run it afterwards...

but if you're using a p4 with HT then run two instances of prime95 at the same time....

 

[Lyfer]

Trying to determine stability is like trying to determine what is normal.

 

[OCDuron]

I ran 2 copies of Prime95 for 24 hours on my XP-M 2500 @ 2700MHz. That was good enough for me

 

[Sahakiel]

Originally posted by: Shimmishim
running prime for 12 hours is good enough

some guy once ran prime for 876 hours and it errored in the 876th hour...

now tell me... would you say his system is stable?

On a consumer level, that's probably somewhat stable.
On an enterprise level, that's not stable at all.

 

[saechaka]

Originally posted by: photogearjunkie
Hi,

I have a question that I hope some of you guru's can answer. I'm trying to test my system stability by running a Prime95 "Torture Test" coupled with a SuperPI calculation of 32M digits. (P4 3.0C @ 3.6GHz, various voltages; ASUS P4P800-D; 1G Corsair 4000PRO; WD Raptors in RAID; etc...) Prime95 can run for 12+ hours, alone or with other apps going. But when I simultaneously run a 32M digit SuperPI calculation, I usually get errors in one or both of the tests.

My question is this: Should I ONLY run Prime 95 to test the system stability? Or should the system be stable if I run both of these apps simultaneously?

Thanks a lot!

i actually tried both tests on my barton 2500 at 3200 mem at 2.5,4,4,7 and it took 40+to finish 32mb super pi and primed for 5hrs without errors. just wanted you to know

 

[suklee]

Originally posted by: OCDuron
I ran 2 copies of Prime95 for 24 hours on my XP-M 2500 @ 2700MHz. That was good enough for me

holy cajoly :Q I'm jealous!

 

[maluckey]

Lately, I got impatient to stability test, and now I run UT2004 at max resolution and details, with OC'ed vid card, and EAX sound, while running SuperPi and Prime95 Torture Test (blend). If it all holds together for an hour or so, then I quit and call it a day. I was able to run Prime95 overnight without errors before, but when I did this, the setup initailly crashed in 30 minutes. A bit of adjustments, and now it hasn't crashed in the same test at over an hour and a half. I figure that it's all a matter of perspective. I am impatient, and figure this torture must represent a load that in normal operation will never occur.

I also find out the the setup is stable while eating lunch, instead of waiting till morning.

 

[beatle]

Originally posted by: Shimmishim
some guy once ran prime for 876 hours and it errored in the 876th hour...

LOL, that's ridiculous... running for over a month.

 

[photogearjunkie]

Wow, thanks for all the info everyone. I'm going to try a few of your suggestions and see how it goes.

However, I don't think this 3.0C will get to 3.6 with full stability though. And I'm not willing to go over 1.60 vcore with this crappy, overvolting, voltage-fluctuating, lousy-voltage-regulating ASUS board! Do you sense some irritation ; ) I think I'm going to move to an ABIT IC7 Max3, and give it another go.

Thanks a lot!

 

[ectx]

Originally posted by: photogearjunkie
Wow, thanks for all the info everyone. I'm going to try a few of your suggestions and see how it goes.

However, I don't think this 3.0C will get to 3.6 with full stability though. And I'm not willing to go over 1.60 vcore with this crappy, overvolting, voltage-fluctuating, lousy-voltage-regulating ASUS board! Do you sense some irritation ; ) I think I'm going to move to an ABIT IC7 Max3, and give it another go.

Thanks a lot!

Agreed. P4p800-D is not greatg for oc. When you run p95 - voltage dips dramatically. I have antec tc 550 and it did not help.

If you try to run p95 "custom" test; set min fft to 2048 K I bet you have problems in a few hours.

For p4 with HT, I do think you need to run p95 + sonething else to get to any problem quickly.

I also found out, for Barton and p4 (512K cache) - 3dmk2003 CPU test (you need a license or a hack) looping will uncover some problems quicker than any other tests. But 3dmk2003 cpu lopping stablity does not mean you will be stable for p95, though. (so you should not limit yourself to 3dmk2003 test only.)

Eg., I have a 3.0c - at 250x15 p95 + 'heat up" together, it would run for >12 hrs, but I loop 3dmk2003 cpu test- it crashes in a few min. The same chip at 248x15 - 3dmk2003 cpu test loop forever but p95 + hu will crash at about 24 hr. I tested and get similar results on 3 bartons. However, my 1700+ will never fail 3dmk2003 cpu test at any reasonable oc.

What have I learned? P95 small FFT tests are no good at all. Blended tests do not uncover problem quickly. Run custom test with bigh fft's. Run 2 app for HT chips's. Run 3dmk2003 cpu test for about 10 min, it it does not crash, you should move back to p95 = sopme other app and test it patiently.

Or, you could simply "believe" you rig is stable.

To me, the goal is to find a test that can detect instability in a few hours. I am still searching.

 

[ectx]

Originally posted by: Shimmishim
running prime for 12 hours is good enough

some guy once ran prime for 876 hours and it errored in the 876th hour...

now tell me... would you say his system is stable?


you can also try running superpi if you want, but i'd run it afterwards...

but if you're using a p4 with HT then run two instances of prime95 at the same time....

I hope your story is not true. I have a leap of faith - I tested >30 chips and I have never seen a rig that passes p95 tests for more than 40hrs and subsequently fails anything else. (a) I test high fft's only in p95 - low fft does not give you much unless you really purhed your chip (in that case low or high fft would both error out immediated) and (b) I have seen one or two exceptions, but I think it was heat (tested in winter but failed in summer).

But heck, I have never tested any of my rigs for more than 75hrs. Maybe I am not as obsessive as I though I was.

 

[maluckey]

I have...

My old gaming rig (KX7-333) ran Prime95 for four days on blend, but crashed repeatedly during instense 3d gaming, in under an hour. I tweaked the setup a little more and all isues went away. That's when I switched to running multiple stressors at the same time, and ALWAYS including a 3D application as part of the final testing.

FWIW I've found that DivX encoding also can be a good stress tester if done while running something else with it.