overclocking, prime95 fails in about 3 seconds, works otherwise..

[pamf]

im running my athlon 1.2 at 1.4ghz.. ive upped the core voltage to 1.85 (the highest my kt7a's bios allows) .. prime95 fails however in about 3 seconds I've been running it this way for about two hours.. played some quake, ran some benchmarks in sandra, etc.. everything seems fine.. no crashes or lockups yet, but prime95 suggests a fairly big problem .. since i cant put the voltage up any more, does anyone have any suggestions what else i could do? does the io voltage affect this in any way? its still at default.

the cpu's a 1200c axia .. you can see the core here in all its chipped glory if the rest of the numbers and such on it make any difference.. the boards a kt7a rev 1.1, with the 64 bios .. ive got 512mb micron pc133 (256 x 2)... it's running at 56c with about 40% load.. im not too worried about the temp though since ive had this cpu up around 70 back in my storied 'orb days' ..

 

[Jman13]

Run SETI for a while and see how fast your machine locks up. When I OCed too far, it ran Prime95 fine, but after 2-3 hours of SETI, it would lock hard. 2 hours is not a good test period. You'll likely find that your machine starts to exhibit very strange behavior at that speed. I'd recommend bumping it down a notch.

 

[pamf]

well, the problem isn't whether or not its stable yet. . supposedly if prime95 has a problem, somethings not working right and its getting bad results from the cpu .. so running seti wouldn't do much good .. i dont want to return bogus results ..

after i get prime95 saying its all ok then ill go for the torture test and handle that when i get to it (bigger heatsink etc )

so you dont know what else i can do after the voltage is up at its max?

maybe i could edit the bios or something to change the max voltage setting.. but before id try that id want to know what the max safe voltage i can use without frying my cpu is ..

 

[Pecos]

Load up one of the Serious Sam games and let it loop the demo. I'd be interested to see how long that runs. Not very long I bet. Sam seems doesn't seem to be real forgiving on some oc'd systems.

 

[pamf]

whether or not things crash right now isnt my main concern.. right now what im concerned about is how prime95 gets an incorrect result after about 3 seconds. i want to fix this before i start stress testing. since my voltage is already maxed, what ELSE can i do? what does the io voltage do? anything would be helpful.

 

[BigJ]

Heh. Try running Prime95 + 3dMark2001SE looping, + SETI with rendering software in the background. If your computer doesn't lock up/have errors, then congratulations, your OC is stable.

 

[Plester]

how does prime95 do w/ the chip running at default speeds? sounds like you just ain't got a stellar chip there.

 

[Garet Jax]

pamf,

Prime 95 is very sensitive to overclocking your CPU. From experience, when I was very close to being stable, I would run Prime 95 and down grade the OC until the torture test would run to completion without problems.

I was not confident that my OC was stable until Prime 95 would run with no problems.

My guess is that your OC is pushing the CPU a little too far. I would back down the OC a little and keep trying Prime 95 to see if it improves.

Having said all that, a lot of people feel that as long as your system runs everything you want it to without problem, then it is stable (regardless of the prime 95 results).

 

[pamf]

when i ran at 1200, prime95 ran for about 4 hours with no problems uintil i finally quit it. at 1333, it ran for a few hours with no errors until i quit it, but i had to up the voltage to 1.8 .. now at 1400 it gets an error a few seconds in, but the voltage only goes to 1.85.. aside from this ive had NO noticable problems.. no lockups, crashes, etc. the only difference is some higher benchmarks (smoking p4 1.6's according to sandra ) and about 6c on my cpu temps.. i played quake, rtcw and a bunch of other games for a while with no problems at all.. im content to leave it at this ,even if it occasionally results in a lockup, but id like to fix it if i can.. i was planning on buying a bigger heatsink (the pal8045) to bring my temps down and run at 1.4, but if heat isnt the problem then im not going to spend a hundred bucks on a heatsink and fan..

so, what DOES io voltage do? would changing it help the situation? make it worse?

and does anyone have any suggestions which DONT involve stress testing? right now im JUST trying to get prime95 running. i can think of a thousand things on my own that would molest my system.

 

[rogue1979]

You cpu simply is overclocked to high, Prime 95 catches any error instantly. Since your board doesn't allow higher voltage settings that leaves you with two choices. Lower the overclock, that will work for sure, or perhaps better cooling will get you stable at 1400Hz. Try the original Millenium Glaciator for $23 shipped from their website, excellent cooling, no delta. If your computer runs everything else stable, then you could pursue 1400MHz as is, but I would advise against it. A glitch or two will certainly show up sooner or later, and there is the definate possibility you will slowly corrupt you operating system and end up having to reformat.

 

[pamf]

ok .. so can i get an answer on what io voltage does? would changing it possibly help? would it fry my system? should i leave it where it is?

 

[Plester]

rogue1979 - how long you been running that chip at 2.09v air-cooled? i ran a 1gig AXIA @ 1.55 at 2.1v water-cooled for a few months, it was stable and cool (high 30s to low 40s), but it up and died completely one day - never booted again. i would be very careful running that kind of voltage - especially aircooled.

 

[dunkster]

The two items that I'd check out are heat and cpu voltage.

If you're seeing cpu temp of 56C at 40% load, your cpu temp will climb rapidly at 100% load. I'd invest in a better cooler and look also at how well your case is exchanging air.

Regarding 1.85V VCore - is it really necessary to reach 1400MHz? Any increase in VCore beyond what's needed for stability only increases the cpu heat problem and decreases cpu stability.

Same for memory voltage - any voltage higher than necessary may contribute to instability.

You could play with reducing dram timings for higher FSB stability - but that is usually a losing tactic with regard to overall system efficiency.

Hope this helps!

 

[pamf]

dunkster, i tried prime95's benchmark and it locked up seconds later ;/ i tried booting at 1400 and 1.75v, but it locked up at the splash screen.. 1.8v got me this far.. prime95 still gives an error though .. -sigh-.. would cooling actually help this? because i dont really want to spend money on a better heatsink when the one i have now works fine at 1333 if i cant hit 1400 anyhow..

im at 55c right now, 41% load..

 

[Plester]

sadly your chip is running a bit too hot, get it under 50c at full load and you will see more stability. you do not have to spend a fortune on the 'best' cooler - go over to overclockers.com and have a look at their cooler tests - they are the most comprehensive and accurate. you should be able to find very good cooling in the $25 range.

ps. run prime95 w/ an open case on a cold day and you will find that it probably doesn't crap out in 3 seconds.

pps. you indicated that even at stock speeds prime95 kicks out after 4 hours or so, that should NOT happen at stock speed.

 

[pamf]

no, i said after 4 hours i gave up and closed it. it didnt give an error.

 

[Def]

Simple, you're overclocking too high(is there an echo in here?). I'd say you'd need a pretty good heatsink to get it completely stable at that speed(maybe high 40's FULL load, which it looks like you're almost to 60 at FULL load).

Is it really worth it for that extra 66Mhz? Can you actually tell a difference? Just run it at 1.33Ghz and be happy you got something for free.

If you are going to spend money, save up for an XP 1700+ and overclock that. No point in spending mad money on a heatsink to get 66Mhz, when you could save up a bit and get a better chip that will clock alot higher.

 

[pamf]

if i was going to buy an xp, which i fully plan to when i can, i'll need a new board since for some #@%@Iing reason kt7a's less than v1.3 (aka mine) dont support xps. so theres another $200 right there. and i doubt my $20 thermaltake fan would do very well on an xp, so id need a new heatsink anyhow. how much i overclock this cpu on a new heatsink isnt really the issue since id be using this one on my next cpu anyhow. meanwhile i'd get a higher clock and some extra bragging rights. truth be told i cant tell the difference between 1200 and 1333 either but that doesnt stop me from running it.

anyhow, thanks for the replies.

 

[Jeff7]

I/O voltage is for the RAM. Only raise that if you are overclocking the FSB.

 

[rogue1979]

Plester, you are making me paranoid. My full load temps under Prime95 are 46c, I have been running like this only a week. Before I was running 2.01v at 1458MHz. This cpu is about a year old, most of that time it was running 1400MHz at 1.88v. Do you think the high core voltage killed your cpu, or is it possible that it was "just one of those things". My family has been through about 30 different cpu's in the last 2 years, most of them highly overclocked, only one has ever failed and I chalked that up to the law of averages. I am gonna post a thread and see if anyone else believes high core voltage has prematurely terminated the life of a good cpu. I am somewhat concerned, I have owned this cpu much longer than any other, and this system is running extremely fast on a 167MHz fsb.

 

[Plester]

rogue - i am fully convinced that the voltage smoked my chip. i would throttle back to 2v, but then again you never know - chips are like fingerprints.

 

[Duvie]

I agree the heat seems pretty high right out of the gate....

However my experience has shown that if prime95 fails fast it is usually a voltage issue or ram issue...heat takes a bit longer then 3 seconds to build up...you may not see highest temps for several minutes into test...

That is just my opinion from my past experiences...

My p4 will run all night with 1.8ghz@2.4ghz w/ 1.57v, memory and pci locked...but if I take it up to 2.5ghz it fails almost immediately and my temps hadn't even crested above 48c at that point...I need another voltage boost but the msi board wont go higher. I actually have it set to 1.65v but it only delivers 1.57v...many report same issue on all msi boards.

 

[rogue1979]

OK, I chickened out. Backed off to 1467MHz at 2.01v. Actually I tried to get it running 1403 at 1.92v, but I had a hard time getting it stable. Running rock solid now, my full load temps dropped to 43c, so this is where it stays for now.