How long for Prime 95?

D

DangerIsGo

2[H]4U
Joined
Apr 16, 2005
Messages
3,000
Question...may be stupid may be not but I have to ask it. In the FAQ it says Prime 95 should run for about 24 hours to be considered stable. According to [H]ardfolding, "This doesn't usually cause problems in games or Windows, but in precise scientific applications like Folding@Home, this can cause problems." Right now I am @ 275x10 stable. If I bump it up anymore, even 1 MHz, Prime fails. My question is, if im just mostly gaming and sort, how long should I run Prime for?
 
i usually make sure that prime torture test runs all the way up to 4096 K FFT size without crashing, which took me about 10 hrs with my previous settings, and call it good.

i do not leave my computer running over night or when i am away for an extended period of time, though.
 
wadec22 said:
I accept nothing less than 24hrs prime95 stable.
Click to expand...

QFT. Not only do all my rigs Fold, like D3v01D, but it's just good practice - if you're Prime95 stable for 24+ hours it reduces the likelyhood of your OC flaking out in weird ways.
 
ElektronikSeraph said:
QFT. Not only do all my rigs Fold, like D3v01D, but it's just good practice - if you're Prime95 stable for 24+ hours it reduces the likelyhood of your OC flaking out in weird ways.
Click to expand...
i fully agree with the two of you.

i don't do it myself because past history and self preservation dictate that i never leave my computer running unattended.

i admit that a roughly 10 hour run of prime 95 is a lesser standard, however i can honestly say that with 10 hours of prime 95 stability, i can almost always find a reproducable software/driver glitch when i do get a (seemingly) random crash.

additionally, i set the thread priority to high and run with few other things turned on in the background when i do a run with prime, to make the time spent as demanding as possible.
 
The longest I believe I have let it run for was like 9 hours and change. I dont think Ive ever hit more than that. Is that good?
 
Big Fat Duck said:
i do mine for 15 minutes max
Click to expand...

Wow.

Personally, I wouldn't credit any overclock tested for only 15 minutes. I've had overclocks run Prime95 fine until finally failing at 14 hours. Fifteen minutes of Prime testing isn't even worth the clicks it took to start it.
 
ElektronikSeraph said:
Wow.

Personally, I wouldn't credit any overclock tested for only 15 minutes. I've had overclocks run Prime95 fine until finally failing at 14 hours. Fifteen minutes of Prime testing isn't even worth the clicks it took to start it.
Click to expand...
agreed.
 
DangerIsGo said:
What do u mean by that?
Click to expand...
i should think that it would be pretty evident if you look at what prime95 displays as it runs.

look at the last few words of each line.

with default settings, the test runs from 8 K FFT length up to 4096 K FFT length. it alternates between medium and small FFT lengths for the first portion of the test, getting smaller and smaller, until the values converge, then it moves to large FFTs and works its way up to 4096 K FFT length.

i let the program run until it has finished the pattern and has sucessfull completed the 4096 K self test.
 
I have done prime95 for 24 hours if not more a couple times. So I agree, 24 hours!
 
My processor will never be at full load for more than 10 hours. It comes close during the occaisional video encoding. Take that into account when deciding how long to prime. I stop it at 10 hours and it's good enough for me. It's also important to be able to pass SuperPI 32M. Some unstable system my pass 10 hours of Prime, but fail SuperPI. I'm not happy until I pass 10 hours of prime, and SuperPI 32M.
 
I just see
prime.jpg


Where do u see 4096K FFT length?
 
It doesn't look like you've hit the 4096k FFT the previous poster was talking about..How long was it running in that shot?

Oline - I understand it's stable for you, apparently you don't use F@H, UD, etc..I can understand you only Priming for as long as an encode goes - I'd still test it for 24 hours though personally, even if I didn't have a 24/7 machine.. If it's not 24 hr stable, what's to say that while encoding, it doesn't spit out a random error somewhere during your 10 hrs - or if you wanted to do a batch encode or something?..

I (and many others here at [H]) obviously do have a need for 24 hour Prime stable machines - as evidenced by the [H]ardFolding test stats in their respective DC projects. Personally, I do run UD, about 99% of the time (stats below, actually almost 550k total points tho), but with a) crappy downtimes there recently,, and b) Opteron 165 waiting @ FedEx warehouse since Friday, that'll change for what's under 1/2 of the IHS - since UD refuses to get with the times and officially support 64bit and dual core/SMP machines :).

 
i've let an oc run prime95 for 12hrs b4, then had it crash in the middle of a guild wars mission, 24hrs required from there on. :)
 
DangerIsGo said:
I just see

Where do u see 4096K FFT length?
Click to expand...
do you have the torture test set to "blend"?

methinks that what you were running in that screen shot was the small FFT test.

blend is nice because it takes turns hammering on your memory, and then hammering on your FPU, keeping processor temps good and high the whole time. the last little bit of the test, which you complete around the 10 hr mark, is one of the meanest things that you can throw at your memory. it also draws a tonne of power and so stresses your HDD a good bit, especially if you do something to spin up all of your drives while it runs.
 
I have the older version of Prime 28.3 I just upgraded to 24.14. Now I see what your talking about. I just finished 24 hours on the older version of prime and its stable. yay. Would I notice a difference in ram speed between 275 and 290 MHz? same timings 3-3-3-8. Im running a 1:1 divider. Is it best to run a 1:1 divider or it doesnt matter?
 
DangerIsGo said:
I have the older version of Prime 28.3 I just upgraded to 24.14. Now I see what your talking about. I just finished 24 hours on the older version of prime and its stable. yay. Would I notice a difference in ram speed between 275 and 290 MHz? same timings 3-3-3-8. Im running a 1:1 divider. Is it best to run a 1:1 divider or it doesnt matter?
Click to expand...
higher core helps you more than running 1:1 does. if putting your memory on a reduced divider allows you to push your core higher, then do it.
 
15 hours. Even then I still have had overclocks fail. Prime just isn't all that great IMO.
 
Prime is great, I usualy let it run for 2 days 48 hours at the end of the setup.

I keep adjusting untill it will run for 2 days...

I have had it fail after 30 hours before, the second day the ambient temp was higher than the first day...

Basiucaly the longer you an let it run the more sure you can be that your 100% stable.

Realalisiticaly, I doubt I would be running a 100% load for more than a few hours at a time though...

So ther eis no set limit, but beore I call my system 100% stable I let it run for days.. Course with windows there is no true 100% stable :p
 
Another question/issue. I have my CPU @ 290x9.5 runnign a 1:1 so shouldnt my RAM frequency also be 290? In cpuz and clockgen, it reads it as 275
 
Big Fat Duck said:
well i did it for 15 minutes and it hasnt crashed so far after weeks of BF2, im good here.. and i didnt waste days priming
Click to expand...


Then please dont waste anyones time Folding either. :)
24 hours NO less here as well.
 
Big Fat Duck said:
well i did it for 15 minutes and it hasnt crashed so far after weeks of BF2, im good here.. and i didnt waste days priming
Click to expand...

The whole point in priming is to check stability, you cannot do this with only 15 minutes. It takes at least 8+hrs of prime to ensure your system is stable (some prefer the 24hr method), you can have an unstable overclock that will prime for 1hr then fail, so your 15 minutes is worthless.
 
You must log in or register to reply here.
Back
Top