Prime 95 Strange Issues (OC 9550@4ghz) - Overclock.net

koji · 3 Weeks Ago

Hey folks, I'm new here and offcourse I start off firing questions right away.

I've been running into some strange stuff with prime 95 when running blend tests. It's like this, when I boot, and fire up prime 95 run a blend test, I'll get rounding errors on core 0 or core 1 on my Q9550 (pretty random which one they're on).

Now here's the strange thing, when I run a small FFT test first and let it run a while and then start a blend test AFTER running a small FFT test the blend test will run without rounding errors for hours.

Same goes for starting with a blend test, having it error and restarting one a while later.

Any idea what this could be or should I just ignore this, seems like my system needs to "warm up" a bit first... strange. Could it be that my system has to load completely before starting blend tests? I dunno, I'm a bit clueless about it...

I'll give you guys some info.

Q9550 @ 4.0ghz 471x8.5 (vcore voltage on auto, reports 1.37 on load in CPUz) - on a p5q-pro

corsair dominators running at 942mhz - 5/5/5/15 - 2.1V

NB at 1.40V

all the other stuff is mostly on auto, besides LLC which is enabled

DEVAST8 · 3 Weeks Ago

I'm no expert, but rounding errors usually indicate instability with your overclock. You may need to adjust some voltages or it could be your memory. You could also try another program like LinX to see if it fails. I usually run small ffts, large ffts, and then finish it off with 10 hours of blend tests to check for stability. If you are getting errors on any of the tests, it isn't stable.

koji · 3 Weeks Ago

Thx for your reply, trying to figure out if this is a software issue of some kind with prime or my OC, thing is it does blend tests for hours, but only when I don't launch them right after booting up, makes no sense...

If I run blend right after booting it'll give round errors in a matter of minutes, if I run whatever other test first I'm fine.

My OC still needs a lot of finetuning though so it could be my hardware settings, but it's so strange that one time blend tests runs for hours but on a cold boot not...

The_Parrot · 3 Weeks Ago

well firsat off id get rid of those auto settings and get started on manuel settings.
they never work right and could be over volting somewere that you dont know about
and then work from there

The_Parrot · 3 Weeks Ago

and what are your temps like at 4ghz with air that must be getting a little toasty

Quantum Reality · 3 Weeks Ago

The Small FFTs setting tests the CPU almost exclusively. Blend tests the CPU and the RAM, so if one passes but not the other you can isolate your issues. Also keep in mind memtest86+ can pass memory, but you can still get errors under Windows. Had this happen to me where backing my RAM off a notch fixed the problem.

Shrimpykins · 3 Weeks Ago

It's the auto setting. Manually set the voltage and find a stability point from there. Auto voltages should be set to a manual setting whenever overclocking, especially when overclocking heavily, like you are attempting.

NoGuru · 3 Weeks Ago

Like DEVAST said, run each test for 12 hours and start with small ftt's for the cpu.
Large ftt's for memory. And then blend.
If one fails you have a unstable OC and it needs tweaking.
Oh and Auto on vcore is a no no.

Wishmaker · 3 Weeks Ago

Prime is not good. Linx or Intel Burn. They yield results in minutes about the o/c. The same results you;ll have from Prime after a few hours!

NoGuru · 3 Weeks Ago

Quote:

Originally Posted by Wishmaker

Prime is not good. Linx or Intel Burn. They yield results in minutes about the o/c. The same results you;ll have from Prime after a few hours!

Linx and Iburn is good as well as prime for 24/7 OC's. I suggest using all of them for stability.

		Overclock.net - Overclocking.net > Benchmarks > Benchmarking Software and Discussion
Prime 95 Strange Issues (OC 9550@4ghz)

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)