Low GFlops with overclocking

MooMoo444 · 29 Nov 2010 at 20:13

hi folks,

wondered if anyone would be able to shed any light on my issue atm.

Basically, to sum it up with a picture:

It's passing the tests as "stable" but for some reason the GFlops output is very low compared to others running same clockspeed. It sorta suggests that the CPU isnt running as it 'should'?

I can't for the life of me fathom why.

Any ideas?

Big.Wayne · 29 Nov 2010 at 20:36

Hey MooMoo444

Is 39-42 GFlops considered low for a Intel® Core™i7 @ 4.0GHz? . . . what have you seen other people getting?

Frozennova · 29 Nov 2010 at 20:40

Im not particularly familiar with Intel however that would suggest thermal throttling of some kind if others are seeing Higher GFlops than that.

kitfit1 · 29 Nov 2010 at 20:43

It is low for that clock, yes Big.Wayne.
@MooMoo444, run Linx again, but this time select ALL available ram to use.
In your screenie you are showing a "Problem Size" of 11530, if you have 6gb of ram, the problem size should be over 25000. That should then increase the GFlop figure.

Squakingcow · 29 Nov 2010 at 20:44

Try lowering your clocks a bit, if the performance increases then the chances are that the overclock isn't actually 100% stable.

setter · 29 Nov 2010 at 20:49

From a run i done a few months ago.

Quick blast just now.

All ram selected.

WingZero30 · 29 Nov 2010 at 20:50

make sure your EIST and C1E are disabled in the bios if you have such option.
Your theoretical maximum GFlops: 16 x 4 = 64GFlops. But this is only ideal value. Imo acceptable values should be around within 70% of the theoretical maximum so you should be getting around 45Gflops minimum.

You should also turn off antivirus when running the test. It also depends on the quality of mobo, chipset, components etc.

Check out my thread about running such stress tests in which I also quoted an excellent article on ensuring you get higher Gflops

:

http://forums.overclockers.co.uk/showthread.php?t=18206940

MooMoo444 · 29 Nov 2010 at 21:54

Seems a tiny bit better with higher RAM usage but somethings definitely a bit off. Temps arent going above 80 so I doubt it's thermal throttling.

I'll try it every day for a few days see how it changes. If no improvement I'll either reformat windows or take PC apart and build it again, just to take those out of the equation <_<

WingZero30 · 29 Nov 2010 at 22:20

When I run Intel Burn test sometimes I get lower Gflops. Then I stop the test and restart it again and get higher values. So you may have to keep stopping and starting the test several times to see if that makes any improvements.

setter · 29 Nov 2010 at 22:21

With certain earlier versions of IBT, there was a bug with thread detection if left on auto iirc.

JonJ678 · 30 Nov 2010 at 00:37

I had a problem with a version of intel burn test in that it was passing but then unstable in use. The gflops value was lower than Setter's iirc. Changing to a newer version increased the gflops figure to a more standard one then crashed.

So I think changing to a different version of linx is the way to go.

It could also be turbo doing something inconvenient, like deciding the chip is drawing too high a current so it'll only use the higher multiplier some of the time.

MooMoo444 · 30 Nov 2010 at 08:19

WingZero30 said:
When I run Intel Burn test sometimes I get lower Gflops. Then I stop the test and restart it again and get higher values. So you may have to keep stopping and starting the test several times to see if that makes any improvements.

setter said:
With certain earlier versions of IBT, there was a bug with thread detection if left on auto iirc.

But I'm using linx, not burntest :confused:

Any idea what version of linx is a good one then? I just googled the latest yesterday.

WingZero30 · 30 Nov 2010 at 09:06

MooMoo444 said:
But I'm using linx, not burntest

Any idea what version of linx is a good one then? I just googled the latest yesterday.

Afaik Linx is similar to Intel burn test and both are simpler versions of linpack.

MooMoo444 · 30 Nov 2010 at 09:10

Found an interesting comparison thread while googling;

http://www.overclock.net/intel-cpus/869018-intel-amd-gflops-data-thread.html

For an i7 above the 4ghz barrier, 51-60gflops seems like the normal values.

My CPU is handicapped and I can't work out why

game4it · 30 Nov 2010 at 09:57

MooMoo444 said:
Found an interesting comparison thread while googling;

For an i7 above the 4ghz barrier, 51-60gflops seems like the normal values.

My CPU is handicapped and I can't work out why

Are your sure the OC is fully stable?
Have you passed at least 20 runs at "all" memory?

Have you tried adding a little more Vcore?
I find the the cpu will give lower gflops if it still needs a little more vcore.

If you close down some programs that are not needed while testing you should be able to use 25k problem size which will also give higher flops.

MooMoo444 · 1 Dec 2010 at 17:20

Didn't give up on this!

I updated the bios to the most recent one, and I also reinstalled windows incase I'd mucked up chipset drivers or something.

Seems to have fixed the problem, so it was all worth it!!

Looking much better :cool: