Dual Prime 95 stable for 19hrs

[Azsen]

As per the title my system is dual Prime 95 stable (Large FFTs) for 19hrs before I stopped it, but when I run SuperPI it will error in less than 1 second?

What's up with that? Is SuperPI a better indication of stability? I find it hard to believe the Prime95 will go for 19hrs without any problems on the most stressful test, but SuperPI will fail instantly. :Q

 

[imported_Kensai]

SuperPi > Prime95 for testing stability.

 

[bjc112]

Originally posted by: Kensai
SuperPi > Prime95 for testing stability.

Meh.. I don't know about that..

Good when used in conjunction with one another..

SuperPI is more sensitve to memory errors and such, back your memory down, or loosen you timings and you will probably pass with flying colors.

 

[theMan]

run memtest. which prime95 test were your running?

 

[Leper Messiah]

pi is a bit more memory dependent. raise vdimm a couple of tenths and you should be okay.

 

[ChiPCGuy]

The best indication of stability is a combination of tests. First and foremost, run MemTest86 v1.60 for at least five passes. Then, run SuperPI. If you make it that far with no problems, then run SnM Burn at Long Mode with Safe Mode disabled. SnM Burn will give you about a three hour intensive torture test that will raise your CPU temp higher than you have ever seen it go. If you make it past that point, then running Prime95 is almost not worth it--the aforementioned tests are pretty darn hard on the system. If you really want to torture everything, run SuperPI in the background while running 3DMark2001. That will reveal any faults, believe me.

 

[Lithan]

Prime = CPU stab test. It will test mem a little on the right settings but takes HOURS to find what memtest finds in minutes.

 

[ZL1]

P4 with HT ?

you need two prime95s running small FFTs !
how to run second instance "prime95.exe -a1"


D

 

[ZL1]

Originally posted by: Kensai
SuperPi > Prime95 for testing stability.

???

superpi is a bench not a tester


D

 

[ectx]

I read it somewhere that 2 prime95 would not be useful for HT (they share fpu???). You need 1 prime95 and 1 other application.

 

[ZL1]

Originally posted by: ectx
I read it somewhere that 2 prime95 would not be useful for HT (they share fpu???). You need 1 prime95 and 1 other application.

I ran all kinds of tests and I can easily tell you they work just fine

D

 

[Azsen]

Originally posted by: bjc112
SuperPI is more sensitve to memory errors and such, back your memory down, or loosen you timings and you will probably pass with flying colors.

Well you see I was running at 250 FSB and the memory is DDR550 and running at stock timings so still well under stock memory speeds.

I've tried increasing the vcore up from 1.4375v to 1.4875v so far but still getting the superpi error in <10secs.

My friend was telling me that SuperPI doesn't actually have anything to check the numbers against to make sure it's accurate like Prime95 so how could it be good for stability testing?

 

[ChiPCGuy]

Originally posted by: Azsen

Originally posted by: bjc112
SuperPI is more sensitve to memory errors and such, back your memory down, or loosen you timings and you will probably pass with flying colors.

Well you see I was running at 250 FSB and the memory is DDR550 and running at stock timings so still well under stock memory speeds.

I've tried increasing the vcore up from 1.4375v to 1.4875v so far but still getting the superpi error in <10secs.

My friend was telling me that SuperPI doesn't actually have anything to check the numbers against to make sure it's accurate like Prime95 so how could it be good for stability testing?

Uh, you just said that you get a SuperPI error in less than 10 seconds. How is that not checking itself?

True, SuperPI is designed as bench tool, but it is also particularly effective for testing memory stability due to how it operates. If you run MemTest86, you will probably also get errors. Have you tried Memtest86 yet?

http://www.memtest.org

Download the bootable ISO and use a burning software to burn the ISO to CD. Reboot. Test.

SuperPI does check the pi.txt file for accuracy at the completion of the run. Your friend is wrong.

 

[ectx]

Originally posted by: ZL1

Originally posted by: ectx
I read it somewhere that 2 prime95 would not be useful for HT (they share fpu???). You need 1 prime95 and 1 other application.

I ran all kinds of tests and I can easily tell you they work just fine

D

Of course it will run, but you actually do not push the chip to the limit by running 2 copies of prime 95.

 

[ZL1]

Originally posted by: ectx

Originally posted by: ZL1

Originally posted by: ectx
I read it somewhere that 2 prime95 would not be useful for HT (they share fpu???). You need 1 prime95 and 1 other application.

I ran all kinds of tests and I can easily tell you they work just fine

D

Of course it will run, but you actually do not push the chip to the limit by running 2 copies of prime 95.

I meant I tested the machine in every possible way, Im very picky about stability, and dual prime did the best job.

D

 

[morkman100]

I've tried increasing the vcore up from 1.4375v to 1.4875v so far but still getting the superpi error in <10secs.

He said to raise the VDIMM, not VCORE. VDIMM is memory voltage. VCORE is CPU voltage. Raising vcore will only help if your CPU is causing the error.

good luck.

morkman

 

[AthlonAlien]

Have you tried Prime95 blend mode? It tests a lot more RAM.

 

[ZL1]

Originally posted by: AthlonAlien
Have you tried Prime95 blend mode? It tests a lot more RAM.

on a HT CPU you need to run dual prime, dual prime in blend will fight for ram and wont use the cpu at 100%

to test the ram he needs to run memtest

D

 

[Azsen]

Originally posted by: morkman100

I've tried increasing the vcore up from 1.4375v to 1.4875v so far but still getting the superpi error in <10secs.

He said to raise the VDIMM, not VCORE. VDIMM is memory voltage. VCORE is CPU voltage. Raising vcore will only help if your CPU is causing the error.

It did provide a few more seconds stability before erroring on SuperPI again. I've been all up the vcore ranges and tried maximum vdimm, so the only thing left is to lower the clock speed?

 

[ZL1]

Originally posted by: Azsen

Originally posted by: morkman100

I've tried increasing the vcore up from 1.4375v to 1.4875v so far but still getting the superpi error in <10secs.

He said to raise the VDIMM, not VCORE. VDIMM is memory voltage. VCORE is CPU voltage. Raising vcore will only help if your CPU is causing the error.

It did provide a few more seconds stability before erroring on SuperPI again. I've been all up the vcore ranges and tried maximum vdimm, so the only thing left is to lower the clock speed?

what are the full system specs and what temps do you get ?

D

 

[imported_Kensai]

Originally posted by: ZL1

Originally posted by: Kensai
SuperPi > Prime95 for testing stability.

???

superpi is a bench not a tester


D

Failing SuperPi = Non-stable system.
Failing StressPrime = Non-stable system.

 

[ZL1]

Originally posted by: Kensai

Originally posted by: ZL1

Originally posted by: Kensai
SuperPi > Prime95 for testing stability.

???

superpi is a bench not a tester


D

Failing SuperPi = Non-stable system.
Failing StressPrime = Non-stable system.

granted, but you said spi is better than prime at stress testing and that isnt so
first hand experience: my p4 will run spi at 3.4, but will fail single prime at that speed

D

 

[Azsen]

Originally posted by: ZL1
what are the full system specs and what temps do you get ?
D

Full system specs in sig. Just got a new replacement motherboard so I'm trying for 4Ghz stable again.

Temps are like -12C idle, 5C load.

 

[ZL1]

Originally posted by: Azsen

Originally posted by: ZL1
what are the full system specs and what temps do you get ?
D

Full system specs in sig. Just got a new replacement motherboard so I'm trying for 4Ghz stable again.

Temps are like -12C idle, 5C load.

what stepping is the cpu ?

btw why did you go from lga back to s478 ? lga can handle more power and prescotts sure like power


D

 

[Azsen]

Originally posted by: ZL1
what stepping is the cpu ?

btw why did you go from lga back to s478 ? lga can handle more power and prescotts sure like power

D

E0 stepping.

I did have a LGA 3.2 Ghz but I bought a Vapochill PE and the LGA775 socket kit + better overclocking motherboard to run LGA775 was going to be another $500 or something, so I just decided to go with S478.