Re: testing for CPU faults
orthos and prime95, all cpu tests are designed to root out instabilities
Re: testing for CPU faults
Well you could do a load test such as prime 95 with all stock settings.. that will show if the PC itself has any issues and focusses mainly on the CPU, though the instability could be the RAM or Mobo (afaik, never used prime myself) if there is one.
Chances are there will be nothing wrong with it asumming it actually works at all, I've never seen a dead chip from anything other than what could be considered technical abuse by an absolute moron and a little overclocking doesn't do very much unles the chip gets cooked.
Re: testing for CPU faults
do a load test using prime95, that'll show you any instability
edit: i'm way too slow :(
Re: testing for CPU faults
Just thought that I would add my own comments about P95.
I really don't see the point of this, as my pc cant do 5 mins of CPU test without failing :D, yet it is normally perfictally stable:).
http://i330.photobucket.com/albums/l...e_2008/p96.jpg
http://i330.photobucket.com/albums/l...e_2008/p95.jpg
Re: testing for CPU faults
Does it normally work at 100% load..?
Re: testing for CPU faults
yer, it had stopped a while ago...
lol, I just got the parts for my new one, will build on my b-day (when i offically get it,21st july:D)
It has been like this for about a year (ever since i first ran orthos or p95:)
Re: testing for CPU faults
Another way to test your CPU's stability at 100% load is Super-Pi. Pick the highest option and see if your CPU finishes or not. One CPU/overclocking guide I read on Coursair's website swears up and down that if a CPU can successfully do Super-Pi on its highest setting, its likely about 99% reliable; essentially giving the same end result as Prime85... My overclocked CPU did it in about 15 minutes and i saw my CPU warm up nearly 20c from its usual temps. O_o I must say the guy was right though; I've never had a problem with my overclocked CPU, and I never bothered running Prime95 for several hours.
Re: testing for CPU faults
Oh I'm all for lazy options.. I think I might try that next time :p
Re: testing for CPU faults
I tend not to test (the computer cost me £60 - so not a great loss).
But when I get my new pc - core2quad Q6600, Asus P5Q pro, 4GB DRR2 800 corsair memory, antec 900, sapphire HD 4850 320GB WD, I think that I will test it. Super pi sounds gr8............supprisingly my current pc cant do anything near the highest output:S
Re: testing for CPU faults
Just tried Super PI on my stock P4 3.4ghz .. took 35mins to complete
- temps went up about +12c
- 50% cpu usage throughout
- about 264MB of memory useage!
I'll give this a go when i get the Quad up and running.. then maybe Prime95 too
cheers
Re: testing for CPU faults
When I ran Super-Pi on my Core 2 Duo E6750, one of the cores was always 100% while the other did very little, but the working core would alternate between the two. I didn't try running two Super-PIs at once because firstly, I never expect to max both cores at once for an extended period of time, and secondly, I was afraid it'd be too much for my poor little dual-core... lol :)
Still, my overclock of 800MHz has proved very stable.
Re: testing for CPU faults
SuperPi is not a very good CPU stablity test. True, really flaky systems will not pass a longer SuperPI run, but it is perfectly possible to have a system that is unstable to the point of doing harm to your data that can still do 32M SuperPi without incident.
For CPU stability testing, I recommend 24 hours of Prime95 small FTTs, to be sure. You'll want to get Prime95 25.6 and run as many threads as you have CPUs/cores. Older versions will work as well, though you may have to run multiple instances to use all of your CPU's cores.
Orthos and OCCT are based off of Prime95. As long as you run enough threads, they should all be effectively the same.
Quote:
Originally Posted by
narz
Just tried Super PI on my stock P4 3.4ghz .. took 35mins to complete
- temps went up about +12c
- 50% cpu usage throughout
- about 264MB of memory useage!
I'll give this a go when i get the Quad up and running.. then maybe Prime95 too
cheers
If your CPU usage is not 100% while running it, you need to run more threads/instances.
The blend test, the default one that uses the most memory, is the worst CPU test. Prime95 is not an ideal memory stability test, even on blend, so there is very little point in using the blend setting.
Quote:
Originally Posted by
Will:)
I really don't see the point of this, as my pc cant do 5 mins of CPU test without failing :D, yet it is normally perfictally stable:).
I wouldn't consider your system stable at all. You could very well be silently corrupting data with that setup. You just haven't encountered any obvious problems...yet.
Re: testing for CPU faults
Quote:
Originally Posted by
oralpain
I wouldn't consider your system stable at all. You could very well be silently corrupting data with that setup. You just haven't encountered any obvious problems...yet.
Yep, If his data involves accurate calculations they'd all be wrong :mrgreen:
From his perspective my PC is stable at 4GHz :) But if it were folding I'm sure it'd be sending bad data back.