Is 16 hours on prime95 enough?

D

Dutt1113

[H]ard|Gawd
Joined
Jun 30, 2005
Messages
1,623
I was wondering if you guys think 16 hours on prime95 (v25.4), the one with the green icons that automatically does all the threads for you, in my case 4 for my q6600, is enough?

I have my q6600 at 3.6ghz at 1.54v on water and the temps are 59 59 54 54 as of right now max load.
 
Absolutely not. I've had Orthos fail on me after a 25-hour bench session, so 24 is the bare minimum.
 
I think it is enough personally. But if it's already been running for that long, may as well wait 8 more hours and let it run it's course.
 
honestly what you should do is run it one whole year straight. This way you'll account for the variances of temperature with the time of year as well as humidity fluctuations. One day in the winter is not the same as one day in the summer.
 
More then enough time.

OTOH...

My Q6600 has never, I repeat never crashed on me since I put it together almost a year ago under normal usage (x264 encoding/gaming/PS...etc), yet it will blue screen after 10 minutes of one of these "stress test" apps. I don't put much stock in them for that fact...
 
I second the 24 hour minimum. Personally, my stuff hasn't failed me but I have seen it many times where someone will run their stress test for 20+ hours and it fails.
 
At some point it becomes random. If the system is really unstable, it'll fail relatively quickly, in the first hour or so. If it runs for 20 hours and then fails at 21 hours, it is just as likely to run 100 hours and fail at 101 hours, in my opinion. All that means is that there is some level of instability in the system that Prime will eventually find (although no other app probably ever will). These questions always get asked and everyone comes out saying how 12 hours or 24 hours or 48 hours is the right answer, but really there isn't any right answer. It is what ever you feel comfortable with.
 
18 hours is probably enough but as others have said, no real reason not to go to 24. Also, if you're not running any type of Distributed Computing program or other calculations which require precise results and build off of those results you should be fine.

Also, is that voltage before or after vdroop as it seems rather high. Two of my Q6600s run 3.6 and the voltage on them is around 1.42 after vdroop.

 
Prime is good for finding out whether the mem speeds/ settigs as well as the northbridge voltage settings are stable. Imo I don't think its as good at finding out if your cpu is stable at the tested speed/voltages but thats where IBt comes in.
 
I think 16 hours is enough. Remember that a stress test does just that... you want to run it long enough to figure out if your OC is stable, not till you make your hardware fail because it's been under load for so long. I well understand that some people want to do that, but IMHO, it's not good for your gear.
 
I usually wait a fair amount of hours and let it do a few loops. 12 is Fine but when I do it i usually do it over 9000
 
It depends what you use the CPU for. I mainly play games, I have almost 0 faith in these torture tests, and no game will stress a CPU like Prime95, let alone IBT. That being said, screw IBT. It cooks the shit out of your CPU for no good reason IMHO. I just acquired a Q6600, it took 1.4v to get it stable at 3.4ghz for 10 max stress passes on IBT. Temps hit 74 in a pretty warm room. Fired up Prime 95...temps were mid/high 50's. Played GTA4 for awhile, temps were low 50's.

Im currently running it at 3.6ghz at 1.5v, and temps dont go above high 50's in a very warm room while playing GTA4.

Is 16 hours enough? Maybe. Do what you would normally do with the computer...if you have crashes and instability...then your OC is unstable.

On my previous e6700 OCed to 3.6ghz...never had it crash, yet it would fail Orthos with in 30 min.
 
i say if you build your PC just for prime testing, that is not enough but for gaming? or other stuff I suggest you account for that more rather than worrying about prime, I know prime is good tool for stability but I'd be more on the real world testing...
 
With a good mob and ram prime95 24hr stability should be a walk in the park for any moderate oc. Otherwise whats the point of running a system at a certain overclock 24/7 when it can't even pass 30 min of Prime.
 
As stated above I have failed tests past the 20 hour mark before.

24 is good for me. And a thorough testing of the memory with memtest
 
I've got Prime running since yesterday @ 3.861Ghz and now 31 hours later its still going and stable, should I stop it or let prime run out to 48 hours mark?
 
DeadSkull said:
I've got Prime running since yesterday @ 3.861Ghz and now 31 hours later its still going and stable, should I stop it or let prime run out to 48 hours mark?
Click to expand...

Nah, just let it run all the time, that way you know it is stable, or at least can pinpoint the moment it becomes unstable.

Seriously though, 30 hours is more than enough.
 
To each his own. For my desktop OC's ~2hr or intel burn or prime is fine. I know the things i do on it wont push the cpu nowhere near what few hours of stress testing will do. so my e8400 is running at 4250 8.5x500 and is solid as hell. I had it crash in prime once after few hours but paid no attention to it. Not a single crash in games/apps for the month it has been running. Also i don't think stress tests on high voltages are healthy to the motherboard i seen quite of them fry just because of extended burn tests.
 
I usually only let mine run for about 10 hours. If you want to break the cpu in though by all means let it run as long as you like.
 
I test twice, for at least 24 hours.

After the first test, I cripple the cooling system in a manner that allows it to run 5-10 degrees hotter for the second test. The idea is to test for stability under worse conditions than it will see in actual operation. I usually do this by reducing the fan speed by one method or another.

For example. A given overclock passes a 24 hour burn in test, during which time the temp peaks at 45 C. For the second test, I'll make it run about 55 C. Then when it passes that, I'll consider it stable.

My line of thinking is that I don't just want to know it's stable now. I want to know that my system is not right there at the cliff's edge, where the heatsink builds up a little dust, the temp goes up 5 degrees, and that give me a BSOD, wrecking whatever I was working on in the process.

My process is a little different, but I don't think I'm the normal overclocker. I go for a modest overclock, to get a better value out of my cpu. I think the 3GHz I'm getting from my 18.GHz E2160 is fantastic, even though I could push it harder if I wanted to. I'm more concerned with being certain it is stable, than I am with pushing for the max limit. I know this will be sacrilege to some of you, but so be it.
 
I've had it fail at the 23 hour mark before. So I've run it past 24 and hit even 48 hours.

The fact of the matter is, if one core fails, it doesn't take out the whole system, BUT if you are doing anything that requires a precision calculation, you don't want to take a chance that a random error gets introduced.

In the long term too, what if that little bit of system instability grows? You can be 1 month 6 months or a year into using your "stable" system, then one day its unstable and you assume its a program problem or something else until your mission critical data goes corrupt. It's not just the CPU you should worry about, when overclocking you are pushing other components too, so you have to keep in mind that not all of them may handle the overclock perfectly in the long term and that small instability could be the first sign.
 
You must log in or register to reply here.
Back
Top