Prime95 errors

Frosteh

2[H]4U
Joined
Nov 30, 2004
Messages
3,664
Just wondering how seriously I should take the Prime95 errors when doing torture tests?

I've got my 64 4000+ at 2.7Ghz stable with Prime95, not kicking up errors even when leaving it on overnight for about 8 hours. But when I overclock to 2.75Ghz Prime95 kicks up an error after about 1-2 mins.

The thing is the PC runs games fine, have been playing the CPU heavy TES:Oblivion recently and all the other latest games for extended periods of time, have done other benchmarks such as the 3dmark CPU tests, have used superpi to calculate pi to 32million decimal places.

Not had any lock ups, crashes, BSOD, or errors of any kind that I can possibly spot while using the PC, so for ACTUAL day to day use it's been pefectly stable, in fact stable for everything EXCEPT prime95

So whats going on? I don't get it, from what I hear if Prime95 gives errors then your CPU or RAM is not stable and thats simply the end of the story.

my Vcore is at 1.55v at the moment with temps of about 37 degrees for idle (or relativly idle), my temps when running Super Pi are about 50 degrees

So I guess what im trying to ask here, is how does the Prime95 tests actually help, it doesn't appear to be any kind of indication of real world stability, or is it simply my CPU is returning incorrect calculations but not enough to crash the PC (so for instance my calculation of Pi might be wrong)

I wan't a stable PC, but currently it's what I'd called stable for my needs, it's never crashed ever or given my any unexpected problems, and It's genreally running all the time except for restarts for driver updates etc.

Another thing to note, I have a few friends who say Prime95 gives an error after about ~5 hours with stock everything on their 64 3800 CPU, got really low temps and everything else in their PC is new.
 
Well, the better question is, how important is that last 50MHz if it's failing prime number calculations? You're not going to see any difference between 2.7GHz and 2.75GHz. I myself wouldn't want corruption in my OS and other stuff that is associated with incorrect calculations (think 1+1=3). But, it's all in what you find "stable". But I seem to see a close association with users who don't care about prime and users who also format their machine every 4 months. I've had my RAID array for about 2 years and wiped it recently as Windows had just accumulated tons of crap in various hidden temp folders and stuff, and it just needed a complete cleaning.

 
Well my system is setup for pretty good recovery if anything goes wrong, it's also easy to reinstall windows on.

I have a RAID 0 setup for my primary drive, with 2x Raptor 10,000RPM drives, not the best for data security, nothing of any importance is stored there, all my media is saved to folders on a 2ndry drive which is much larger and slower, purely for backup. It also stores all my drivers and program install for fast reinstallation of windows, this makes the perfect setup for anything bad happening to my install of windows because the most I loose is a few hours of my time reinstalling.

The extra 50Mhz is not NEEDED in the strict sense of the word, I could live with out it, but if the only repercussion is that I have to rebuild windows every 4 months instead of every 6-8 months instead (which I do just to keep my system fresh and clean) then I don't see a reason not to have it.

I suppose the difference is that Prime95 is measuring accuracy directly, not the instability which is derived from that inaccuracy.

And 50Mhz isnt all that small of an increase really, the way I see it I'm getting a 350Mhz increase over my stock speeds rather than a 300Mhz Increase, and that increase leads to a real world increase in what I can do with my PC especially with games.
 
If you fail prime, it indicates your cpu is returning false information, which in certain cases can go unnoticed for quite awhile, and in others result in corruption and loss of data. I take it very serious, why would you want a cpu that is malfunctioning?
 
Something can malfunction and still be perfectly useable if the problems it's having isn't effecting the programs running or the user.

So far I've had no problems, If I ever get any serious problems that can be directly related to CPU overclocks then I will of course back down the clock speed, but so far I've had none.

I guess my view on this is kinda like the HardOCP view on benchmarking, use the applications you're actually going to use with the computer to test the frame rates and/or quality. I mean the only thing so far thats had problems is Prime95 and since I dont use that in my day to day use then it should be ok.

I'm ok with taking that risk though because of the redundancy I have setup, others understadably might not want to take such risks.
 
That's one hell of a risk you're taking.

Those errors will show up eventually, and if it's only in your system files, then you're lucky. They may turn up in your work, your backups, and files that you can't replace. Your system will not be as stable. Remember, it's a 2% increase in performance.

I hate to be blunt, but running a machine that you know is unstable is STUPID
 
Why did yiu run prime in the first place. If it doesnt matter than why run the test. The amount of reduncy you have wont help if your cpu is pruducing false calculations. You'll just have it saved mor than once.

It could be a sign of something else as well. Strangley, I couldnt get my 4000+ to get past 2739 stable until I changed power supplies. Now it's running at 2821stable in prime at 1.45v. I wouldnt run it if it wasnt stable, regardless if ti ran my games fine. Your playing with a time bomb.
 
Id have to agree. Saying "it hasnt happened yet" doesnt mean its not going to. If its failing prime, then at some point, in some certain scenario, the CPU WILL crash and possibly screw up your data. Im not sure why anyone would run an unstable cpu. Try doing some tasks that are CPU intensive such as encoding, photoshop, or rendering. Work on a project for a few hours then watch the computer crash and lose your data. I dont see the point in taking that risk. I think most people will agree. Theres a reason people prime95 for 10+ hours. :p The program is not useless. :)

Also you should know that even prime95 stable, does NOT mean the cpu is even completely stable. Thus why people make the effort to make sure its AT LEAST prime95 stable. So a system that is knowingly prime95 unstable, is just that much more risk added to the equation.
 
Retsam said:
Id have to agree. Saying "it hasnt happened yet" doesnt mean its not going to. If its failing prime, then at some point, in some certain scenario, the CPU WILL crash and possibly screw up your data. Im not sure why anyone would run an unstable cpu. Try doing some tasks that are CPU intensive such as encoding, photoshop, or rendering. Work on a project for a few hours then watch the computer crash and lose your data. I dont see the point in taking that risk. I think most people will agree. Theres a reason people prime95 for 10+ hours. :p The program is not useless. :)

Also you should know that even prime95 stable, does NOT mean the cpu is even completely stable. Thus why people make the effort to make sure its AT LEAST prime95 stable. So a system that is knowingly prime95 unstable, is just that much more risk added to the equation.

I agree, and besides it's only 2%, what that equates is maby .5FPS difference in system heavy games, IMO not worth risks, I'm just running my system at 2.6Ghz cause its prime stable and onlys like 4C higher then the stock 2.0Ghz, when its idleing or playing single core games thiers no temp difference though.
 
The fact that you don't care if you lose all of the data on your RAID array for 50MHz just astuonds me. I have a fileserver to store all of my media, but I still dual prime my rig to make sure it's operating as it should be. It doesn't take much skill to overclock, it takes skill to overclock correctly.

 
Back
Top