Prime95 BS :(

[dug777]

Used 3marks, superpi, and intensive Doom 3/Far cry playing to test the stability of my overclocks b4 now, but thought i'd try Prime95 today...get some stupid 'illegal sumout' and drivel about a possible hardware failure...instantly. It obviously restarts, then flashes the same error 100 times, then halts.

I could accept that my machine wasn't stable if it at least ran the test for more than a nano-second, it just doesnt seem right Anyone had the same problem, or some constructive advice as to what the problem is?

My rig is absolutely rock-solid in all the apps i mentioned at the start, and i can run the sandra burn-in test (just the cpu tests/HDD test/memory bandwidth/cache) all day long without a problem

 

[WobbleWobble]

It's up to you how stable you want your machine to be. If you feel everything is fine, great! Ignore Prime95.

I personally do care because then I know that my CPU make incorrect calculations. They call it a Torture Test for a reason

 

[dug777]

I'm perfectly happy with the stability of my rig, it just shiats me that i can't run prime95

It just doesn't seem to make sense that it errors immediately...

 

[Sunner]

Prime95 is what I always use to test stability since it seems to be incredibly sensitive to high overclocks.
Most of the time I get errors fairly quick, not right away like you did, but after anything from 10 seconds to maybe a minute or two.

Maybe you could try downloading Knoppix and run it from there, a good test since you'll eliminate the possibility of something with your Windows/Linux/whatever install screwing things up.

 

[JavaMomma]

I use Prime95 to test my system. I always make sure it can run it without error for 24hrs.
I figure if it can not run Prime95 without problems, well that is atleast 1 program my computer can not run properly. I want my computer to be able to run everything.

I remember I used to have the attidude that if it is stable in what i use it for, good enough. Then I started getting corrupt zip files sometimes. And then I relized they werent really corrupt it was my CPU, and it only happened maybe 10-20% of the time and only on the big ones like 200MB +.

Since then I have been a fan of a 100% stable computer. I don't want weird errors popping up.

 

[sharkeeper]

Does it err out when you're not overclocking?

If not, then you know the answer to your question. The system is NOT stable at that overclock. If it plays games fine, then so be it.

Just don't depend on that system to output anything important as it cannot be trusted.

Cheers!

 

[dug777]

i'll give that a shot and let u know- i had thought of it but my rig's always on and doing loads of stuff at the same time-so it's kinda annoying doing a restart

Is it strange that Superpi is fine, but Prime isn't?

EDIT: I tried clocking back to stock (xp2000+, usually runs as a 2600+), and sure enuff, i had no probs with prime95

But i then tried a d3 bench (timedemo demo1), and had to scrape my jaw off the floor as i stuggled to a average frame rate of 27 I usually get 38-40 Fps (at medium 1024) at 2600+, so blow that for a joke It was pretty much unplayable (i'd hate 2 think what min fps it was hitting!), and the situation was similar for Far CRy (in fact, even worse).

So Prime95 stability can go r00t itself as far as i am concerned-i've never had ne other problems o'cd in any other apps, my games are rock-solid (and i can actually play them ), and clocking back down made me realise how much of a boost i got from the overclock

Thanks for the feedback folks

 

[dug777]

'bump'

 

[Sunner]

Sounds strange that you'd get such a disproportionate performance drop.
2000+ would be...1.66 GHz right? And 2600+ would be 2.13
So an overclock of roughly 25% gives you a performance increase of over 35%, supralinear scaling is pretty much unheard of, especially in gaming situations.

Or maybe my math just sucks, hey I did it in my head, and I'm tired, so get off my back, ok?

Oh and in any case, I'd follow Sharkeepers recommendations, don't do anything important on a box that's not 100% stable(well, as close to that as you'll get with a PC).
Though if you're only gaming, I guess the worst thing that could happen is you losing a bit of playing time due to a crash.

 

[dug777]

2083mhz at 2600+ i think- at least that's what mine's running at.

I checked the benches again, same settings and got almost exactly (those two words really don't belong together do they ) the same results, remember that at 1667mhz i am running with a 133FSB, but at 2083mhz i am running at 166FSB- thats the other thing i thought would give my scores such a beating, whaddya reckon?

As i keep mindlessly squwaking my rig is one of the most stable machines i have ever owned, i run for up to 5 days between restarts (tho i do hibernate her most nights), and that is fairly intensive gaming, watching movies (divx and dvds), and the ususal word/email/IE/acrobat stuff, and i really can't think of or remember the last stability issue i had

 

[VirtualLarry]

dug777, Your system isn't Prime95 stable, and you use Hibernation? Wow, you're daring.
Btw, is there any particular reason why you don't just keep your CPU at a stable speed, and clock your FSB up to 166Mhz?

Sunner, from my understanding, while games should scale linearly or sub-linearly up until they are no longer CPU-limited, Athlon systems in general scale much more with FSB speed increases than other platforms, so combined, the two could be significant, to the point of appearing supra-linear.

 

[boshuter]

It sounds like your system isn't even close to being stable at the speed you normaly run.......... obviously you are happy with it and that's what counts. Just play games and don't count on it for anything important and it's not a problem.

 

[Sunner]

Originally posted by: VirtualLarry
[
Sunner, from my understanding, while games should scale linearly or sub-linearly up until they are no longer CPU-limited, Athlon systems in general scale much more with FSB speed increases than other platforms, so combined, the two could be significant, to the point of appearing supra-linear.

Actually, from what I remember, the Athlons have much less to gain compared to the P4's when increasing the FSB.
The P4 for example made a fairly impressive jump when going from 533 -> 800 MHz FSB, while the jump from 333 -> 400 Mhz for the Athlons made little difference.

Of course, right now I'm both quite tired and a tad drunk, so I'm too lazy to look for some reviews, so if you any to prove me wrong, I'll stand corrected

 

[THUGSROOK]

Originally posted by: dug777
I'm perfectly happy with the stability of my rig, it just shiats me that i can't run prime95

It just doesn't seem to make sense that it errors immediately...

denial....

 

[Goi]

If it doesn't run Prim95, it isn't stable. It's as simple as that.

You might be comfortable with a lower level of stability, but you won't be able to say that your system is 100% stable.

 

[dug777]

OK-its not 100% stable

But it's exactly stable enuff for my purposes if i never have any problems unless i try running prime95 The link between prime95 stability and real world stability seems pretty tenuous in my situation

I have never had a single problem with Hibernate on this rig, (out of interest whats the worst that could happen 2 me using it?). I didn't know that my rig wasnt '100% stable' until yesterday when i ran prime, and that was just because it was getting a bit of an airing on the forum and i was interested, having never used it b4.

I realise that a claim of 100% stability is inappropriate, but i never made that claim Are there any other programs that would highlight my 'system instability' that prime has revealed?

Cheers for the feedback

 

[Goi]

Prime95's one of the best. It stresses the FPU a lot. I'm sure there are others that stress other components of the system such as the integer pipelines or the memory hierarchy. If it's stable enough for your purposes, good for you. Just know that if you have a crash in the future, it could very well be due to instability that you didn't acknowledge.

 

[imported_Phil]

Originally posted by: dug777
OK-its not 100% stable

But it's exactly stable enuff for my purposes if i never have any problems unless i try running prime95 The link between prime95 stability and real world stability seems pretty tenuous in my situation

I have never had a single problem with Hibernate on this rig, (out of interest whats the worst that could happen 2 me using it?). I didn't know that my rig wasnt '100% stable' until yesterday when i ran prime, and that was just because it was getting a bit of an airing on the forum and i was interested, having never used it b4.

I realise that a claim of 100% stability is inappropriate, but i never made that claim Are there any other programs that would highlight my 'system instability' that prime has revealed?

Cheers for the feedback

If your system is not 100% stable, please do not run Distributed Computing (SETI, Folding@Home, etc), as the results generated may not be accurate.

 

[dug777]

[q

If your system is not 100% stable, please do not run Distributed Computing (SETI, Folding@Home, etc), as the results generated may not be accurate.[/quote]

No worries - i had read the little readme that came with prime on the matter. I ususually have plenty of stuff going on at the same time, and i am too lazy anyway, so it was never that likely that i'd get involved in any of that on this rig.

Going back to my question above is there any other bench/app that would replicate the instability i see with prime that neone knows about?

 

[Budman]

last time i seen an illegal sumout error it turned out to be the ram,try relaxing your ram timings & test again.

And like others said... " if it's not prime95 stable it isn't stable "



And btw you might want to run memtest also . http://www.memtest.org/

 

[flexy]

yes, prime95, rthdribl and sometimes 3dmark are the rude awakening for some "self crowned overclocking masters"
Eg. "I just cranked up my CPU and STILL am able to boot ----> means: stable overclock ---> quick, lets post everwhere and report my overclock success"

I am sure that 75% or more of people claiming surprisingly high overclocks dont even know how to test their systems right etc....and i take every claim of CPU/GPU wonder-clocks with a grain of salt anyway.

If your CPU has problems calculating (eg. in P95, Seti etc.) because of the OC - well, then you HAVE a problem. Or would you want to buy a $320 CPU and then use it and crank up OC so your CPU calculates fantasy-numbers and can only work in 90% of apps/games right - with a 10% chance that you will have problems with other apps ? Not worth imho....then better a few mhz less and Prime stable and THOUROUGHLY tested

 

[sharkeeper]

If your system is not 100% stable, please do not run Distributed Computing (SETI, Folding@Home, etc), as the results generated may not be accurate.

I believe these software should perform a system check and refuse to upload completed work units if the processor is overclocked more than 1.5% over stock.

Cheers!

 

[imported_Phil]

Originally posted by: dug777

If your system is not 100% stable, please do not run Distributed Computing (SETI, Folding@Home, etc), as the results generated may not be accurate.

No worries - i had read the little readme that came with prime on the matter. I ususually have plenty of stuff going on at the same time, and i am too lazy anyway, so it was never that likely that i'd get involved in any of that on this rig.

Going back to my question above is there any other bench/app that would replicate the instability i see with prime that neone knows about?

SiSoft Sandra should do a pretty good job with its Burn-in test.

 

[dug777]

Originally posted by: Dopefiend

Originally posted by: dug777

If your system is not 100% stable, please do not run Distributed Computing (SETI, Folding@Home, etc), as the results generated may not be accurate.

No worries - i had read the little readme that came with prime on the matter. I ususually have plenty of stuff going on at the same time, and i am too lazy anyway, so it was never that likely that i'd get involved in any of that on this rig.

Going back to my question above is there any other bench/app that would replicate the instability i see with prime that neone knows about?

SiSoft Sandra should do a pretty good job with its Burn-in test.

Thats part of my irritation with my prime problems, i can run the sandra burn-in-test all day long (and have quite recently ) I only run the 2 cpu tests/mem bandwidth/cache/filesystem tests, but no probs there. I ran a quick (not) 100 run cycle of the two cpu tests last nite w/out a problem.

q]Originally posted by: flexy
yes, prime95, rthdribl and sometimes 3dmark are the rude awakening for some "self crowned overclocking masters"
Eg. "I just cranked up my CPU and STILL am able to boot ----> means: stable overclock ---> quick, lets post everwhere and report my overclock success"

I am sure that 75% or more of people claiming surprisingly high overclocks dont even know how to test their systems right etc....and i take every claim of CPU/GPU wonder-clocks with a grain of salt anyway.

If your CPU has problems calculating (eg. in P95, Seti etc.) because of the OC - well, then you HAVE a problem. Or would you want to buy a $320 CPU and then use it and crank up OC so your CPU calculates fantasy-numbers and can only work in 90% of apps/games right - with a 10% chance that you will have problems with other apps ? Not worth imho....then better a few mhz less and Prime stable and THOUROUGHLY tested [/quote]

But that doesnt apply here, i may not be prime stable BUT THERE ARE NO OTHER PROBLEMS AFAIK
I know (now) that i have a prime problem, but atm thats all i have, and thats been the case for the last six-months or so. I'll be quite prepared to accept the importance of prime stability if someone can point out other benchs/programs that will replicate the instability, but going on my experience i don't see the point

 

[Somniferum]

Have you tried Budman's suggestion of relaxing memory timings and testing again? I'm curious to see the result.

Also as he suggested, you should run Memtest86 overnight and see how that does. You might just need to tweak your memory settings.

One other thing -- you said Prime95 runs fine at stock but errors out in a nanosecond at your max overclock -- have you tried overclocking incrementally and running Prime95? I'd be interested to see your max stable overclock that runs Prime95, compared to what you are running now where Prime95 won't run at all. If the diffence isn't too great it may be worth clocking down by a few MHz just for the peace of mind. Oftentimes in my experience, the difference between Prime95 stable and not can be something tiny like 20MHz. (Of course all this assumes your mobo supports incremental overclocking.)