prime95 settings

[paradisiac]

ok, never really used prime95 as i dont really understand what the best way to set it up is so always stuck with OCCT as that pretty simple.
have had over 24hours stable with it, but i keep reading prime95 or intelburn are better, so, how do you setup prime95 to get the most acurate stabilty result?

 

[Ginger-Menace]

just use the blend test for cpu testing thats what i use

 

[paradisiac]

the default option?
does this just go on and on forever?

 

[Sin_Chase]

Small FFT shows up instability fastest for me, seems to push temps more than Blend will also.

 

[WingZero30]

just use the blend test for cpu testing thats what i use

Use the small FFTs which is specifically for cpu stress testing. Blend is for overall system stability (cpu + ram + northbridge), so it won't test the cpu as hard as small FFTs.

 

[paradisiac]

blend did 3 hours without error and small FFTs has been going for an hour now and no problems

 

[Cold_And_Miserable]

Small FFTs is a waste of time. It could pass that for 12 hours and BSOD in Age of Empires 3 in 40 minutes, fail blend in 30 minutes or fail Intel burn test in 3 minutes.

 

[paradisiac]

Small FFTs is a waste of time. It could pass that for 12 hours and BSOD in Age of Empires 3 in 40 minutes, fail blend in 30 minutes or fail Intel burn test in 3 minutes.

so what are you saying is the way to test stability?

 

[Awahwah]

I use a mixture.

IBT for 5 runs while upping the feq then Prime blend over night to make sure it's 100% stable.

 

[Max3]

There is no one test that you can rely on - some guys just don't want to spend all that time stress testing.

Personally I like to linx max mem 100 runs, prime with small fft/large fft/blend. I've had 12 hour small fft / blend pass easily that fail with large fft.

I've stopped using IBT since linx came out since IBT seems easier to pass for some reason.

 

[Liquid]

ever since day one I have used the prime95 blend test, running overnight to confirm stability/overclock and it has always stood me well. Blend imo is better for overall stability on a system. Each to their own though.

 

[zoomee]

agreed, blend test for 8hours should ensure no problems with stability

 

[Orcvader]

Personally 8 hours of blend works for me as it also stresses the RAM.

 

[paradisiac]

does blend go on forever or does it do a certain amount of tests and stops?

 

[Liquid]

does blend go on forever or does it do a certain amount of tests and stops?

It will go on indefinately or until a core fails and even then I think it goes on.

 

[paradisiac]

ok kool. 4GHz on dual core fails after 5mins or so (BSOD). dual core at 3.9GHz has been running for 45mins and temps are 54, tri core though although it seemed stable, after 20mins it was on 65-69, which i thought was taking too big a risk, so will make sure 3.9GHz is stable on dual core, and then get a H70 or something before trying to get the tri core up and running.

 

[Sayso]

Bad company 2 seems to stress everything lots including GPU, ive had prime and Linx stable fail on BC2.

 

[setter]

Prime blend, or LinX/IBT. I could pass small fft till the cows came home, whereas blend, LinX/IBT would fail pretty quickly. Took a bit more vcore and changing ram command rate and slackening the timings to achieve stability with those apps.

 

[chipps101]

i do 20 runs of ibt then 4 hours of prime blend and then if i have time at least of couple of passes of memtest (always handy to have a ubuntu install disk handy for that reason)

 

[C64]

I have used IBT for ages don't bother with prime really and I haven't had any problems 50 passes of IBT as many G/flops a second as possible pass.
As long as you run it correctly as sometimes choosing the presets can give lower G/flops per second and doesn't stress test properly.

I'll give linx a look which is what IBT is using anyway ? IBT/Linx for me rather check stability in an hour than 8-12.