Why Prime95?

[veri745]

Why do I see so many posts by people giving Prime95 as their only/primary tool for stability testing?

There are several stress tests that seem to be way harder on your CPU: OCCT and Linpack come to my mind.

Personally, I use OCCT, since I find just 1h of OCCT is a pretty good predictor of stability. I'll run Prime95, too, but I've never had a system that was 4+ hours OCCT stable and not 24/7 Prime95 stable.

I haven't had as much experience running Linpack, but I've heard similar comments about it from others.

Am I missing something?

 

[yh125d]

I don't see any need using anything more than p95 for 95+% of users. Nothing I ever do would be more intensive than p95, so as long as its stable enough for that, I'm good. I don't need to put my computer through even more stress just to prove thats its stable at something much less stressful

 

[WildW]

Originally posted by: yh125d
I don't see any need using anything more than p95 for 95+% of users. Nothing I ever do would be more intensive than p95, so as long as its stable enough for that, I'm good. I don't need to put my computer through even more stress just to prove thats its stable at something much less stressful

I didn't think that was the point of stress testing? My mental image of the whole process is that an error can occur at any time on an overclocked-beyond-stability processor, but may only happen on one instruction in, say, a billion. . .and that something like Prime95 is a quick way to do those billion operations, and which will flag up any single error when it happens.

 

[yh125d]

Not necessarily. You can have a setup that will be stable under low loads forever, but crash the first time it gets stress tested. You can also hav e setups that you can game on all day long without a hitch, but fail 5 mins into OCCT. IMO, if it's stable enough to game on, I don't care if it fails OCCT (I have had a situation like that)

 

[Trevelyan]

Hmm... well I plan on O/Cing my new i7 and definitively want to do as hard of a test as possible to see if its stable...

 

[faxon]

i have seen it the opposite way around a few times, where my e5200 passed 18 hours of P95 but kept crashing a few minutes into playing quake wars when set to 3.4ghz. on the other hand, OCCT runs the same algorithm as prime 95, it just finds errors faster and it admittedly does put a bit more stress on the system at the same time. i generally dont bother testing with linpack unless i plan on running the system under heavy loads (like F@H or SETI) 24/7