How important is Prime 24/7 stable, really?

[ShawnD1]

I believe it is important(though I usually only test for 3-5 hours) Couple years back I overclocked an Athlon X2 5000 BE to 3.2Ghz, I believed it was stable like that for months and the occasional program issue was just Vistas fault(probably didn't even crash a program weekly). The games I was playing at the time worked fine too, until I got Left 4 Dead. It took me days of trouble shooting, but no matter what I did Left 4 Dead wouldn't run for more than 30-45 minutes at a time, finally discovered it was my overclock. Decided to run Prime95 for the first time since I assumed it was stable, I wasn't able to pass more than an hour even after upping the voltage. Dropped my cpu to 3.1Ghz and never saw a problem again and was prime95 stable as well.

My story is similar. My Athlon 2500+ was overclocked to whatever speed the 3200+ was and it was stable enough to run windows and all of my games. I was running F@H at the time which put the CPU at 100% 24/7, so the computer would randomly crash every few days. I would sometimes wake up at night because the computer would beep every time it reset. Eventually Windows XP was so corrupt that it wouldn't even boot, so I had to reinstall it every few months.

Eventually someone got me to run Prime95 and my system couldn't even run it for 1 minute before having errors. Now I understand the significance of stability and data corruption, so I'm not happy unless the system can run any stress test for any period of time without failing.

That's also why I don't overclock my video cards... ever. "driver has stopped responding" seems to show up at the most inappropriate times. It doesn't matter how many times I test with Furmark or OCCT, there never seems to be any assurance that the video card is stable, so I just leave it at stock.

 

[EnzoLT]

if you have been overclocking for 20 years, why even ask this? im not being a douche or anything but i gotta question your understanding of the way computers work. just because your computer is running hot doesnt mean its unstable. these are byproducts of the cpu running. people stress test to make sure that 100% of the cpu is working correctly.

I hear you, but Prime puts a much greater stress on the CPU in terms of heat than it will actually ever likely see, hence making it more likely to produce an error then under ordinary conditions.

lolwut? since when does heat lead to the production of an error? it can make components fail if there is too much heat but not in the sense of logic/arithmetic failure or corrupt data.

Take the car analogy. Is it necessary to run your Taurus at 140 mph for 5 days straight in order to say that it's ok to pass at 80mph for 1 minute?

badd but im not going to dwell in that.

 

[ShawnD1]

Take the car analogy. Is it necessary to run your Taurus at 140 mph for 5 days straight in order to say that it's ok to pass at 80mph for 1 minute?

Actually a better car analogy would be doing crash tests on car. Most of our cars will never be involved in an accident, but we test them anyway. If we all agreed that testing was worthless because we don't buy our cars and crash them right away, we would still be driving cars with no seatbelts

 

[quadomatic]

I guess it's important if it avoids crashes and corrupted data. I feel like if your processor goes without errors for a couple hours on Prime95 that might be stable enough, since you're not likely to use your computer at 100% for more than a few hours at a time.

Then again though, you're avoiding corruption and crashes, so yeah. I'm no expert though.

 

[tweakboy]

Not too important. Your CPU will never reach 100 percent full load.

At most if you render it will take 60 or 80 percent.

Prime will take the whole core, soo it might fail at 100 load. But gaming and using apps can work fine. Since they don't take that load. Benchmarks and tests are pointless. The real test is in playing the game for hours.. Or composing music on my DAW ,, thx

 

[Markfw]

Not too important. Your CPU will never reach 100 percent full load.

At most if you render it will take 60 or 80 percent.

Prime will take the whole core, soo it might fail at 100 load. But gaming and using apps can work fine. Since they don't take that load. Benchmarks and tests are pointless. The real test is in playing the game for hours.. Or composing music on my DAW ,, thx

I would say thats an overstatement. If you do any distributes coputing, (like me) you will be using 100% 24/7. There may be other things that are similar, but thats one common example.

 

[LoneNinja]

Not too important. Your CPU will never reach 100 percent full load.

At most if you render it will take 60 or 80 percent.

Prime will take the whole core, soo it might fail at 100 load. But gaming and using apps can work fine. Since they don't take that load. Benchmarks and tests are pointless. The real test is in playing the game for hours.. Or composing music on my DAW ,, thx

What type of rendering are you talking about? If you render in a program like 3ds max it brings the cpu to a constant 100% usage, even on a quad core. You can also have instability and program crashes if your cpu isn't 100% stable without using 100% of your cpu power.

 

[TJCS]

Take the car analogy. Is it necessary to run your Taurus at 140 mph for 5 days straight in order to say that it's ok to pass at 80mph for 1 minute?

Lolz at this, I totally missed this fallible analogy. Put it this way, if you know that there is a possibility that your steering wheels and brakes might hard-lock at 100mph, wouldn't you get a mechanic to test that shit out for you before you get behind the wheel?

After reading your posts, it almost seems like you are asking that if you can avoid throttling CPU to 100% then there is no point to running these benchmarks. You simply can not control CPU usage all the time.
​

 

[JAG87]

LOL at all these 24h stress testers. You do realize that 24h is a span of time that WE invented, and that if something is 24h stable it does NOT mean it's 24/7 stable.

Prime or linpack hammers your hardware harder than anything else you will ever run, there is no need to run a 24h test. If the test runs for a couple of hours without errors you can rest assured you will not see any instability running anything else. 24h is just a number that makes you feel good because it equals a day, and you equate this with 24/7 stability, which in fact is not. 2h of small prime or linpack puts more stress on the cpu and the power delivery than 2 years of folding/gaming/rendering/encoding... whatever your trade is.

 

[TJCS]

I thought 24/7 prime stable meant you are able to maintain stability any time you choose to run Prime95 (not running it 24 hours a day and 7 days a week). No test programs out there guarantee 100% absolute stability.

 

[F1N3ST]

I don't know if my CPU is prime stable, but I am assuming it is, since I haven't ran into a crash ever, for like 4 months no crash at this speed. 3.4 Q9550

 

[schneiderguy]

lolwut? since when does heat lead to the production of an error? it can make components fail if there is too much heat but not in the sense of logic/arithmetic failure or corrupt data.

Excessive heat causes errors. Look at graphics cards when they get too hot, the artifacts that are produced are from errors.

 

[The Sauce]

if you have been overclocking for 20 years, why even ask this? im not being a douche or anything but i gotta question your understanding of the way computers work. just because your computer is running hot doesnt mean its unstable. these are byproducts of the cpu running. people stress test to make sure that 100% of the cpu is working correctly.

Well, doing this for 20 years is exactly why I am asking this. I think its important sometimes to question dogma. 24h Prime testing has become overclocking dogma. This thread proves it. People are getting defensive in here over suggesting that Prime testing may not be an accurate reflection of stability in real-world use.

Also you are saying that heat does not cause errors? Then why use water cooling? Phase-change? How are people hitting 5 GHz+ with liquid nitrogen? What is the point of that if not to decrease heat and reduce errors? If you stand by your words then please, take off your CPU heatsink and run without it... let us all know how it works out.

 

[The Sauce]

LOL at all these 24h stress testers. You do realize that 24h is a span of time that WE invented, and that if something is 24h stable it does NOT mean it's 24/7 stable.

Prime or linpack hammers your hardware harder than anything else you will ever run, there is no need to run a 24h test. If the test runs for a couple of hours without errors you can rest assured you will not see any instability running anything else. 24h is just a number that makes you feel good because it equals a day, and you equate this with 24/7 stability, which in fact is not. 2h of small prime or linpack puts more stress on the cpu and the power delivery than 2 years of folding/gaming/rendering/encoding... whatever your trade is.

That is EXACTLY what I am getting at. 24 hours is completely arbitrary. Why not 10 minutes, or 2 months? Who studied this and came to the conclusion that 24 hours or Prime small FFTs means your computer is stable? Maybe it does and I just never heard how we, the overclocking community, decided that. But I am guessing it was never studied and it just somehow came to be accepted.

 

[Gillbot]

Stability is a long disputed term really. To me, if my system will perform my everyday tasks without crashing, it's stable to me. I really couldn't care less if Linx, Prime95 or OCCT will run for 1 or 400 minutes without crashing since I don't need Prime95 to display my email.

 

[The Sauce]

Lolz at this, I totally missed this fallible analogy. Put it this way, if you know that there is a possibility that your steering wheels and brakes might hard-lock at 100mph, wouldn't you get a mechanic to test that shit out for you before you get behind the wheel?

After reading your posts, it almost seems like you are asking that if you can avoid throttling CPU to 100% then there is no point to running these benchmarks. You simply can not control CPU usage all the time.
​

Ok...the car analogy was bad. I was being provocative.

I am not saying that one should avoid running their cpu at full load. I am just saying that I don't think it really happens that often - especially full-on, 4-core, 100% loads that pile heat and stress on the CPU. I may be wrong about that. Several people have corrected me there regarding distributed computing stuff. I'm guessing most people probably won't use their computers that way. That is a good argument for Prime testing, though. If I'm wrong about that then my argument loses validity, for sure. The question then becomes, how long do you need to run it for to declare stability?

 

[ShawnD1]

I run stuff off and on for random periods of time until I feel comfortable with the results. After building a computer for my mom, I started running the OCCT large data set at background priority just to make sure the heatsink was on right, the memory works good, etc. I forgot that it was running, and it was still running when I came back over a week later. I'd say that was a good run, and it didn't slow her work at all since it was a background task.

The thing about stress testing is that it needs to be done often, not just once. My friend's Athlon started having blue screen problems, so we ran Prime on it, and it failed almost immediately. That computer worked for years without a problem but now it was broken even at stock speeds and voltages. Frequent testing is important if the computer is used to store important information.

 

[Alienwho]

I have an X3 720. Last night I unlocked the 4th core and passed a 100pass test with Intel Burn In test. I then proceeded to play TF2 and blue screen hard crashed in about 20 minutes.

 

[VirtualLarry]

That is EXACTLY what I am getting at. 24 hours is completely arbitrary. Why not 10 minutes, or 2 months? Who studied this and came to the conclusion that 24 hours or Prime small FFTs means your computer is stable? Maybe it does and I just never heard how we, the overclocking community, decided that. But I am guessing it was never studied and it just somehow came to be accepted.

Where's IDC when you need him. As far as I know, the reasoning behind 24hr testing is statistical. If you test stable for 24hr, it's far more likely that you will be stable for 48hr, etc. The longer your test period, the more statistically likely you are to be stable. 24hr just became a convenient cutoff point for testing, after which more testing doesn't seem to prove much.

But if you only test for 2hr, all you can really say is that you are likely to be stable for that period or possibily twice that, but you have no statistical evidence that you are likely to be stable for 24/7. In fact, I think it's quite the opposite, if you don't test long enough.

 

[VirtualLarry]

The thing about stress testing is that it needs to be done often, not just once. My friend's Athlon started having blue screen problems, so we ran Prime on it, and it failed almost immediately. That computer worked for years without a problem but now it was broken even at stock speeds and voltages. Frequent testing is important if the computer is used to store important information.

That's an often overlooked point.

My friend's rig which I built for him, is an E5200 on an IP35-E. I got it to 3.75Ghz (300FSB), prime stable, at 1.425v BIOS for him when I built it.

But about a year or so later, when the system started to beep indicating overheat alarm, we blew the dust out of it, and I retested with prime95, and it started failing. I had to lower the overclock to 3.65Ghz instead. Whether it was heat or voltage or the mobo, I don't know. But I know that we had to slightly reduce the overclock to be stable.

Granted, that machine wasn't crashing, even with marathon sessions of Warhammer Online, so it was "stable" enough for gaming, even though it wasn't prime-stable. But I like to make sure, so I lowered the overclock anyways.

(Chip would do 4Ghz, but not prime-stable, on that voltage.)