Which stress test picks up errors the fastest?

[Cheex]

Title & Summary says it all.

 

[n7]

Orthos.

You can also run iTAT to stress the CPU to see ridiculously high temps or if the system will crash under 100% load.

I also like OCCT for overall testing, though OCCT will pass while Orthos still crashes/errors sometimes.

 

[Cheex]

What is OCCT?

 

[RonAKA]

I find Intel TAT will crash the CPU the fastest if it is not stable - and tell you if temps are a problem. Perhaps it is too severe for realistic temps. The other tough test I've seen is the second CPU test in 3DMark06. The only problem is that you can't select that test only with the free version, so you have to sit through the other tests.

 

[Shimmishim]

http://www.ocbase.com/perestroika_en/

I like OCCT as well but prefer to run Orthos for at least 10+ hours.

Another good, fast, dirty way, is to try dual 32M superpi or dual 1M superpi.

If your board is way out of wack, 1M will error out quick. 32M will error out eventually but only takes 20 minutes to run.

 

[DrMrLordX]

Actually, I've found that the app that was most crash-prone on the last system I set up was Cinebench (the multi-cpu rendering test). It would crash very quickly on an unstable system, and it would crash in instances when Orthos would run for hours on end without difficulty.

 

[gobucks]

i like to use superpi 1M to get in the ballpark, since it only takes like 15 seconds and errors out nice and fast. But for final stability testing i use the other long tests

 

[aka1nas]

Large FFT in particular tends to error out quickly in Orthos when I do stress testing.

 

[Brunnis]

Originally posted by: aka1nas
Large FFT in particular tends to error out quickly in Orthos when I do stress testing.

Same here. I've seen people report Small FFTs or Blend, but in my experience it's Large FFTs that is the most sensitive test in regards to instability.

 

[Thor86]

Originally posted by: aka1nas
Large FFT in particular tends to error out quickly in Orthos when I do stress testing.

Agreed. And after testing large fft tests, then blend tests for 48 hours would deem things 99.99% stable.

 

[Zim]

Microsoft Vista

 

[Blurry]

Microsoft Vista

ROFL

 

[Shimmishim]

Originally posted by: Zim
Microsoft Vista

hahaha. good one.

 

[Cheex]

Originally posted by: Shimmishim

Originally posted by: Zim
Microsoft Vista

hahaha. good one.

Ditto.

 

[nyker96]

OCCT fails faster than prime for my last few builds.

 

[Cheex]

When you say "faster", how long does it take to error out?

 

[n7]

OCCT's default stresstest time is 30 minutes, so likely he means sometime within then.

 

[Cheex]

Ok thanks. I'll give it a try later.