PDA

View Full Version : Which prime95 settings for new system torture test..


nippyjun
06-17-06, 12:59 AM
Small FTT, Large, blend?

Which one is good for stress testing a new system build?

watever
06-18-06, 07:05 AM
I suggest using small FTT to test the cpu.. and memtest for the ram..

nippyjun
06-18-06, 10:06 AM
I suggest using small FTT to test the cpu.. and memtest for the ram..


Which memtest do you recommend, the one that you can run through windows or the boot up version?

Which one of these is best:


http://hcidesign.com/memtest/

http://home.earthlink.net/~alegr/download/memtest.htm

http://www.memtest86.com/

Thanks.

Riptide
06-18-06, 12:31 PM
memtest86+ FTW

einstein_314
06-18-06, 12:32 PM
Use the bootable version. That's the one I use. Actually, my DFI mobo has memtest built into it...so I just use that. But it's the bootable version.

And I agree with watever, use small FFT's in Prime95 for testing the CPU.

nippyjun
06-18-06, 06:57 PM
Thanks for the info..

For memtest86 how many passes of all tests is sufficient? Is 1 pass enough?

Soylent
06-19-06, 10:29 PM
Thanks for the info..

For memtest86 how many passes of all tests is sufficient? Is 1 pass enough?

No. I've found the first errors after 10's of passes while overclocking/using lower timings than the memory is supposed to support sometimes.

Even if you see no errors for a long time you'll never be guaranteed to not have any errors. You'll just know that they're very rare. I'd actually overclock the FSB ~10 MHz or so more than I intend to use the system at(even if I don't intend to use the system overclocked). If that doesn't turn up any errors for hours, you can be sure that the probability of memory errors at normal speeds is really, really low.

nippyjun
06-24-06, 12:27 PM
No. I've found the first errors after 10's of passes while overclocking/using lower timings than the memory is supposed to support sometimes.

Even if you see no errors for a long time you'll never be guaranteed to not have any errors. You'll just know that they're very rare. I'd actually overclock the FSB ~10 MHz or so more than I intend to use the system at(even if I don't intend to use the system overclocked). If that doesn't turn up any errors for hours, you can be sure that the probability of memory errors at normal speeds is really, really low.


Is 10 passes enough to ensure good stability then? or should it run for even more?

Thanks.

einstein_314
06-24-06, 12:39 PM
I'd say 10 passes is more than enough. Well, when I'm overclocking trying to get my max speed, I let it run through 3 times. Then go back to the bios and up settings. When I determine what I think is the highest it will go, I typically let it run all day while I'm at work. If it's still running when I get home, it's good. If not, back to the drawing board so to speak.