• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

A note on Prime 95

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Jimmy53

Member
Joined
Oct 13, 2003
Location
Virginia
I know a lot of people use P95 to stress test. I also know that a lot of people here have a set number of hours that they need to pass before they count themselves stable.

I just wanted to point out that I failed P95 after 19 hours. If you prime for less than 24 hours and you think your system is stable, you might want to check again. Not flaming or trying to cause ruckus. Enjoy.
 
Jimmy53 said:
I know a lot of people use P95 to stress test. I also know that a lot of people here have a set number of hours that they need to pass before they count themselves stable.

I just wanted to point out that I failed P95 after 19 hours. If you prime for less than 24 hours and you think your system is stable, you might want to check again. Not flaming or trying to cause ruckus. Enjoy.

i myself try and run prime for 24 hours after reading mulitple recomendations when i first started visiting the ocforums. reason people shoot for the 24hour mark is to let the system run through a typical thermal cycle throughout thier house/apartmen/dorm etc etc. if you failed on the 19th hour was it around noon or one of the "hotter" times of the day? this can be a issue on air cooling especially.
 
Well the PC is on a bench right now and I was only overclocking the memory. It failed at around 6 in the morning. I'm not griping or anything, I just see a lot of people say: "My system is rock solid, I can prime for 12 hours straight!" Apearently some computers have failed P95 after literally weeks of testing.
 
It's true you can't be certain your computer is completely stable unless you run P95 for a long time, but realistically it may not matter much. For example, if you just overclock your computer to play games one error in 24 hours on a stress test may not be a concern. Most games don't stress your CPU that much, so you're looking at an error happening something like once every couple of weeks or months, and then it will probably be a small error that has no impact on anything. You'll probably crash more due to program bugs than due to instability.

Having an unstable computer is a pain in the *** and I don't recommend it, but after a point you might consider a computer "stable enough". Being able to run P95 for a couple of weeks without errors only really matters if you're really going to be doing something where absolute stability is a must like encoding DivX video or 3D animation.
 
Nagorak said:
It's true you can't be certain your computer is completely stable unless you run P95 for a long time, but realistically it may not matter much. For example, if you just overclock your computer to play games one error in 24 hours on a stress test may not be a concern. Most games don't stress your CPU that much, so you're looking at an error happening something like once every couple of weeks or months, and then it will probably be a small error that has no impact on anything. You'll probably crash more due to program bugs than due to instability.

Having an unstable computer is a pain in the *** and I don't recommend it, but after a point you might consider a computer "stable enough". Being able to run P95 for a couple of weeks without errors only really matters if you're really going to be doing something where absolute stability is a must like encoding DivX video or 3D animation.

i think that is a very reasonable approach. i run prime usually for a couple hrs,(3-4) then try out the computer run games etc. if it stays stable im happy, now of course if i get instability i run it longer.
 
Nagorak said:
It's true you can't be certain your computer is completely stable unless you run P95 for a long time, but realistically it may not matter much. For example, if you just overclock your computer to play games one error in 24 hours on a stress test may not be a concern. Most games don't stress your CPU that much, so you're looking at an error happening something like once every couple of weeks or months, and then it will probably be a small error that has no impact on anything. You'll probably crash more due to program bugs than due to instability.

Having an unstable computer is a pain in the *** and I don't recommend it, but after a point you might consider a computer "stable enough". Being able to run P95 for a couple of weeks without errors only really matters if you're really going to be doing something where absolute stability is a must like encoding DivX video or 3D animation.


Well said bro!!!
 
Having an unstable computer is a pain in the *** and I don't recommend it, but after a point you might consider a computer "stable enough". Being able to run P95 for a couple of weeks without errors only really matters if you're really going to be doing something where absolute stability is a must like encoding DivX video or 3D animation.
Exactly what I think. Way to sum up my thoughts, and those of the two posters above me. Four hours or so is great in my opinion. As long as games aren't crashing or anything and Prime95 gets some time, I'm good.
 
All very valid points and well spoken too. The way that I see it is that if your computer fails prime somewhere between 12 and 24 hours, it's no different than your computer failing prime somewhere between 1 and 12 hours. Computers should never generate errors, ever. If you fail out of prime after a week it could translate to a crash out of D3 / CAD / VB / Word / whatever in 5 seconds. That's an extreme example I know, however I do see a lot of people saying I stress this way and that way and so I know my computer is stable but why I am having random reboots / blue screens / etc. The more you stress, the more problems rear their ugly head.

Computers should never generate errors, ever. That's why they're computers. Software crashes and errors are ultimately human error; hardware working within it's specification should never generate errors, ever. That's the way I see it.
 
All computers, even computers at stock, will generate an error SOMETIME.

I'm a little confused...why aren't your running ECC memory, etc, if no errors are so important?
 
ECC is pricey and not fast enough. All computers, even running at stock, MAY generate an error sometime. They are not gurenteed to though. If I build 10 stock computers today, hook each one up to a UPS, and let them run P95 for 10 years I am sure at least one of them would come up with an error eventually. However, I am also sure that at least of them would pass every test and never generate an error, ever. Thats what computers are designed to do, theyre stupid machines but 1+1 always equals 2.

Us overclockers are of a different breed. We are constantly tracing a thin line between stability and blue screens. Each one of us has a different idea of what constitutes stable. I'm sure there are people out there that don't mind if they desktop out of games after a couple hours, as long as they can get every possible FPS.

The way I see it I want to run 2 outlook windows, 8 IE pages, 6 spread sheets, 10 word docs, 6 adobe docs, 5 visio diagrams, and a pleathora of other misc programs. If I crash from doing all of that I don't want to have to guess: well could that be my memory, cpu, cooling etc. I want to be as sure as possible that my system is as stable as possible. Like I said, we each have our own definition of stable.

Honestly different definitions of stable had never occured to me until I started this thread. If your computer is having unaceptable stability issues that don't show up under stress testing, it might just mean you have to stress more in order for that unstability to show itself.
 
when i first start to see what a cpu/memory/mobo is capable of ill begin by just running small prime runs usualy a hour or so each then when i leave for work or know ill be away from my comp for a extended period of time ill try and squeeze in a longer run.

the majority of pc users i know on a firsthand basis reboot thier pc at least once or twice a day i do however know a couple people who run file servers and whatnot and they try and keep there comp up for weeks at a time. these are the people who need to have that 100% reliability.

myself i tend to agree with the majority of other posters in this thread if your comp can go a day or 2 or so without a error your ok

games crashing after an hour or 2 in my book means its time to turn the clock down a bit. again this is just my opinion but i play games and compete on competitive ladders and would be awfully upset to loose a match cause my comp crashed when i needed it most. in all honesty what goods an extra 10fps if you cant count on it.
 
ECC is pricey and not fast enough. All computers, even running at stock, MAY generate an error sometime. They are not gurenteed to though. If I build 10 stock computers today, hook each one up to a UPS, and let them run P95 for 10 years I am sure at least one of them would come up with an error eventually. However, I am also sure that at least of them would pass every test and never generate an error, ever. Thats what computers are designed to do, theyre stupid machines but 1+1 always equals 2.
Well...what I was really getting at is this. If total stability is so important to you, what do you do to ensure that YOUR computer never generates an error? See what I mean? Since your definition of stable is no errors, I find it odd that your computer is almost certainly "unstable."

I ran Prime95 on a server at work the other day, and of the course the server is totally at stock settings. It stores backups for the rest of the store, etc. Prime didn't last a minute. Made me realize that the IT guy they have there completely sucks.
 
I completely understand that. Unfortunately I don't have the luxury stressing 24/7 to find stability issues. I still want to USE my computer you know? What I do to give myself a wider margin of stability is to run P95 for 24 hours instead of 12. It's really not much I know, and I cannot gurentee myself a stable system just by doing that but I really do feel like I have a much wider margin of stability by priming for 24 hours instead of 2 or 6 or 12. I feel this way because ever since I started OCing I have used P95 and I almost always get my errors within the first 5 minutes, or between the 12th and 24th hour.
 
Mine fails after 6 minutes :(
Even at non overclocked speeds it fails at 6 minutes, I'm still working on figuring out the problem but I just cant seem to find anything.
 
All very valid points and well spoken too. The way that I see it is that if your computer fails prime somewhere between 12 and 24 hours, it's no different than your computer failing prime somewhere between 1 and 12 hours. Computers should never generate errors, ever. If you fail out of prime after a week it could translate to a crash out of D3 / CAD / VB / Word / whatever in 5 seconds.
What I do to give myself a wider margin of stability is to run P95 for 24 hours instead of 12. It's really not much I know, and I cannot gurentee myself a stable system just by doing that but I really do feel like I have a much wider margin of stability by priming for 24 hours instead of 2 or 6 or 12.
So...? Still confused. :confused:
 
Well we can all agree that a highly unstable overclock will usually fail after only a few minutes. But thats only for highly unstable overclocks (or poor souls like Knivez).

A moderately stable (but still shaky) overclock will usually fail after a few hours. However, that same moderately stable overclock might not fail until 8 or 19 hours. If I run 100 P95 tests and each one fails before 8 hours, and then on test 101 it runs for 16 hours before failing does that mean my computer is any more stable than it was before? No, the computer is still coming up with 1+1=3 and just because it took the computer longer to figure that out doesn't mean it's any less prone to doing the same thing again.

A OCer can spot an unstable overclock. If a computer fails prime in 3 minutes is it unstable? I think most likely. It even says in the P95 manual that flaky computers fail within minutes. But failing after 3 minutes is in the degree of highly unstable. Some people here speak of "stable enough" which is much different from highly unstable. I don't think there is "stable enough", but I do think there are degrees of stability. I would seperate stability into three categories. Highly unstable (computers that crash all the time), moderately unstable (computers that crash every once in a while, or might not even show signs of instability on the outside), and stable (computers that don't crash). Irregardless of degrees of stability, a computer is either stable or not. Whether it fails at 3 minutes, 3 hours, or 3 days, if it fails it is unstable.

If somebody ALWAYS tests for stability by priming for 10 hours, and then one day they prime for 18 hours and fail should they ignore it? I don't think so.
 
Back