Prime still best?
- Thread starter gcaster
- Start date
I generally do 20 runs of IBT when I am trying to find out if its stable after each change. When I believe I have found my limit, I do 12hrs of Prime95. If 20 IBT runs adn 12hrs of Prime95 dont show up any errors, then I presume its stable. I test throughout the next few days on games and real world (but unimportant tasks) and if then its still strong I believe it to be stable. That hasnt gone wrong for me yet.
Soldato
- Joined
- 14 Sep 2010
- Posts
- 2,910
- Location
- Port Talbot - Wales
I use SuperPI for a quick check (if it fails that, it'll definately fail others so it's back to the drawing board...) then if it passes SuperPI, I use IBT, Prime95, LinX and wPrime 
Nothing like being sure! LOL
Note: not all at the same time
Nothing like being sure! LOL
Note: not all at the same time
Associate
- Joined
- 12 May 2005
- Posts
- 1,782
You've got an extremely unstable overclock there.
If it fails Linx or Prime95 "Blend" after even hours of testing it will come back to bite you when you least expect it. BSOD's, random system file corruption and strange behavior in Windows are the joys of running such an unstable overclock.
Even Prime95 "blend" for 12 hours isn't enough. I prefer a custom blend from 1024-4096k. 1280k and 2048k seem to fail the most.
Last edited:
Associate
- Joined
- 13 May 2009
- Posts
- 107
As above, I still find Prime reveals errors after 8 hours or more, that IBT and LinX don't show up.
I use IBT to test a basic level of stability first though.
I use IBT to test a basic level of stability first though.
Associate
- Joined
- 24 Jan 2010
- Posts
- 2,141
I use SuperPI for a quick check (if it fails that, it'll definately fail others so it's back to the drawing board...) then if it passes SuperPI, I use IBT, Prime95, LinX and wPrime
Nothing like being sure! LOL
Note: not all at the same time
Nothing forces errors quicker than running multiple stress tests, all at the same time.
Prime95 for decent stability testing.
One thing I love about prime is the choice Small FFT, Large FFT and Blend testing to stress different parts of your CPU. Blend especially used to be good FSB/NB/Memory stability testing alongside the CPU.
One thing I love about prime is the choice Small FFT, Large FFT and Blend testing to stress different parts of your CPU. Blend especially used to be good FSB/NB/Memory stability testing alongside the CPU.
Soldato
- Joined
- 14 Sep 2010
- Posts
- 2,910
- Location
- Port Talbot - Wales
Yeah P95 is always the last one I run, if it manages 20 passes of IBT and an hour on LinX, then I'll whack P95 on overnight and see what I come down to in the morning
Usually just a clean desktop which means it died lol but I have been lucky with my current setup - even if the vcore is a little higher than recommended for these chips - hey ho, it's all good
If it does die, there's always a the MM for a new C2D or C2Q if people start asking £20 for them lol
Usually just a clean desktop which means it died lol but I have been lucky with my current setup - even if the vcore is a little higher than recommended for these chips - hey ho, it's all good
If it does die, there's always a the MM for a new C2D
or C2Q if people start asking £20 for them lol
This. 5 runs of IBT just to see what temps are like and if it's "pretty much there", and then prime to make sure it's long-term stable. If I'm honest, I only run prime for a couple of hours.
Yep, still got a bit of testing to do, strangely though ive never had any issues with windows.You've got an extremely unstable overclock there.
If it fails Linx or Prime95 "Blend" after even hours of testing it will come back to bite you when you least expect it. BSOD's, random system file corruption and strange behavior in Windows are the joys of running such an unstable overclock.
Even Prime95 "blend" for 12 hours isn't enough. I prefer a custom blend from 1024-4096k. 1280k and 2048k seem to fail the most.
I use non of the above, a weeks worth of F@H for me, if i get no lock ups, crashes, BSOD's in that time, then the clock is 100% stable.
Or should i say i used to as iv stoped folding for a bit. As it was said earlier, custom blend is by far the best bet, and will catch any tiny errors, but even then iv passed that and still had errors in folding, so not even prime will find every last error.
Take each stress test with a pinch of salt, non will show u 100% stability, you have to also test the clocks with what you intend to use them with, gaming, encoding, distributed computing, whatever it may be.
Or should i say i used to as iv stoped folding for a bit. As it was said earlier, custom blend is by far the best bet, and will catch any tiny errors, but even then iv passed that and still had errors in folding, so not even prime will find every last error.
Take each stress test with a pinch of salt, non will show u 100% stability, you have to also test the clocks with what you intend to use them with, gaming, encoding, distributed computing, whatever it may be.
I'm a bit confused by IBT.
When I set my CPU clock higher the tests runs take longer but the speed increases. Why is this?
At 4400MHz my 2500k took ~70s to complete each test at ~62GFlops. Now at 4600MHz each test takes ~80s but run at ~64GFlops. Surely as it gets faster it should take less time to run the tests?
Overall IBT seems pretty good though, gives me temps a couple of degrees higher than Prime manages to hit. Thanks for the recommendation all.
When I set my CPU clock higher the tests runs take longer but the speed increases. Why is this?
At 4400MHz my 2500k took ~70s to complete each test at ~62GFlops. Now at 4600MHz each test takes ~80s but run at ~64GFlops. Surely as it gets faster it should take less time to run the tests?
Overall IBT seems pretty good though, gives me temps a couple of degrees higher than Prime manages to hit. Thanks for the recommendation all.