Sign in with
Sign up | Sign in
Your question

Does Prime 95 test need to be 24hours?

Last response: in Overclocking
Share
a b K Overclocking
March 24, 2011 4:34:42 AM

Hi guys i recently did a Prime 95 torture test on my new system and i only did it for about 5hours. My i5 2500k is clocked at 4.2ghz using 1.2volts and my prime 95 test was at 50c avg. I know i can go higher but i really like my temps so i think im going to keep it at 4.2ghz at 1.2volts. Is 5 hours a good enough test for prime 95 or do i need to run it for 24hours?

More about : prime test 24hours

March 24, 2011 4:44:48 AM

5hrs is more then good enough
March 24, 2011 4:51:55 AM

Politely disagree. Have had times when system was stable for up to 8-10 hours but failed shortly thereafter. Was surprised when it happened. Depends on if you will be running system at flat out 100% for that long a period of time, for instance if you are Folding, and if your programs absolutely need perfect computations. I backed down a bit on the overclocks and got stable overclocks in Prime for 24 hours. Worth it IMHO to decrease performance by 1-2% in return for rock solid stability performance. But to each his or her own.
Related resources
a c 138 à CPUs
a c 239 K Overclocking
March 24, 2011 6:32:44 AM

the longer you do the stability test will be more stable your system.
a c 172 à CPUs
a c 197 K Overclocking
March 24, 2011 8:07:54 AM

It's a personal decision. I'm sort of fanatical about stability. I turn off SpeedStep, develop the OC settings, final test with Prime95 small fft's (CPU) and blend (memory) each for 24 hours. Then I turn on SpeedStep and retest with small fft's for another 24 hours.

I used to think that 8 - 12 hours were enough, but I was once testing overnight, came back the next morning, and discovered that P95 had crashed at around 15 hours. Next time it crashed at 18 hours.

After that, I always test for 24 hours.
a c 101 à CPUs
a b K Overclocking
March 24, 2011 9:52:51 AM

Its totally down to you if your mainly gaming then yu can be happy with under 5 hours if you are calculating space missions for NASA then a week may not be enough.
a b à CPUs
a b K Overclocking
March 24, 2011 10:34:43 AM

From the prime95 readme.txt

"...If you have overclocked your machine, I recommend running the torture
test for a couple of days. The longer you run the torture test
the greater the chance that you will uncover an error caused by
overheating or overstressed memory..."



"Twitch his own"
a c 80 à CPUs
a c 206 K Overclocking
March 24, 2011 10:42:09 AM

The amount of time you run Prime95 is directly related to the type overclock.

For instance FSB overclocking a Q9550 on a 775 platform, where you have multiple voltages to tune in, Vcore, PLL, VTT, GTL, NB, NBGTL, Etc. and get to stability, then longer P95 runs are necessary, because you're literally overclocking everything, your entire system has to be rebalanced.

However with your 2500K if you're only raising the CPU multiplier the rest of your system is already within specifications, once you find the voltage to run a certain multiplier, long term P95 runs are just not necessary.

That is, if, you, shutdown any Intel CPU features that cause voltage fluctuations, my 2500K is at 4500mhz @ 1.328v 55c load.
March 24, 2011 1:43:47 PM

Honestly I feel that Intel Burn Test is more harsh/thorough, and multiple runs of it should serve the same purpose (10-20). There are others out there, but I like IBT.

The temps will be higher than Prime95, so you may want to monitor the first 3-4 runs to make sure nothing gets out of hand.

I've been 5 hr Prime95 stable before and still failed IBT. Go to www.google.com and check it out. It's a great way to quickly test your new OC and make tweaks on the fly. You could tweak, run 10-20 passes of IBT, and then do 24 hrs Prime95 if you like.
March 24, 2011 3:52:32 PM

I prefer IBT as well. Its just alot quicker and will uncover just about any instability issues AFAIK.....
!