2500k Overclocking help and advice

StYleY · 16 Oct 2011 at 11:04

Hi guys, just got myself the following components;

ASUS P67 Sabertooth
Intel 2500k with a ThermalRight Ultra-120 Extreme
2x4GB Corsair Vengeance 1600MHz @ 9-9-9-24

I've had a bit of trouble, first motherboard I got was nearly DOA, could POST sometimes, most of the time it didnt post at all. Finally it didnt do anything anymore, totally dead except for the green light on the motherboard. Got myself a replacement and that has been working great.

So to the overclocking part, let me show you a screenshot:

Fiddled around with some different settings for an hour or 2 and finally got it 'stable'. Ran 3 IBTs running 1.3v @ 4500.

Something I noticed is that the GFLops speed reported by IBT is almost half of the speeds I see some of you guys achieving, is something wrong here?

Any recommendations or advice on something which I could change? How should I proceed? I want to achieve around 4.5-4.7 24/7 with a safe voltage.

Thanks in advance,
StYleY

pgi947 · 16 Oct 2011 at 11:49

You on Win7 SP1?

Voltages and temps look fine, depending on your chip you 'might' get it lower than 1.3v @ 4.5ghz. As for safe voltage, thats up for debate, some say 1.38v, some say 1.52v, personally I like to keep it below 1.45v.

AlfBaxter · 16 Oct 2011 at 12:38

Set IBT to extreme rather than high, and you should get the same speed rating as others with similar systems. It will also stress your CPU a lot more giving you a better stability test.

StYleY · 16 Oct 2011 at 13:39

pgi947 said:
You on Win7 SP1?

Voltages and temps look fine, depending on your chip you 'might' get it lower than 1.3v @ 4.5ghz. As for safe voltage, thats up for debate, some say 1.38v, some say 1.52v, personally I like to keep it below 1.45v.

Yes I'm on Win7 SP1.
Really? that high? Alright then, seems like I have quite a bit of room to play with then!

AlfBaxter said:
Set IBT to extreme rather than high, and you should get the same speed rating as others with similar systems. It will also stress your CPU a lot more giving you a better stability test.

Thanks for your input, I am running IBT with the custom mode using full ram at the moment.

Meanwhile I have been running LinX with the 'All ram'-option, it did a 10-times-loop fine in 1 hour, started it again and crashed after 9 minutes.

StYleY · 16 Oct 2011 at 14:09

Update:
IBT with full ram ran fine without errors, 5 loops took almost 30 minutes.
(Still reporting 62-63GFlops though.... should I be worried about that?)

Is there something like an error-quota or does 1 error after xx time equal unstableness?

AlfBaxter · 16 Oct 2011 at 14:18

StYleY said:
Update:
IBT with full ram ran fine without errors, 5 loops took almost 30 minutes.
(Still reporting 62-63GFlops though.... should I be worried about that?)

Is there something like an error-quota or does 1 error after xx time equal unstableness?

I would say something isn't quite right. My system is very similar to yours (4.4 OC rather than 4.5) and I get around double your GFlops with IBT set to extreme. It may not be stressing your CPU enough to test stability. Have you tried Prime95 instead? If that runs for 8 hours with no errors then you should be fine.

pgi947 · 16 Oct 2011 at 14:40

There should in theory be no such thing as IBT not stressing enough. If in doubt, get the latest version of LinX and run that. Low IBT scores are usually associated with a version number below 2.52 and lack of SP1 as AVX wouldn't be implemented.

StYleY · 16 Oct 2011 at 14:49

LinX also reported 62GFlops, will run Prime95 now and see what happens, expect a report in a couple of hours.

AlfBaxter · 16 Oct 2011 at 16:46

pgi947 said:
There should in theory be no such thing as IBT not stressing enough. If in doubt, get the latest version of LinX and run that. Low IBT scores are usually associated with a version number below 2.52 and lack of SP1 as AVX wouldn't be implemented.

Surely there is, otherwise what do the various settings for the stress level do? I thought that IBT provides maximum stress when set to extreme, because this corresponds to the highest memory usage. In any case my system and the OP's should give similar speeds and yet mine is reporting about 120GFlops. Doesn't that suggest that the OP has IBT set to a more relaxed setting and so isn't fully stressing the CPU?

StYleY · 16 Oct 2011 at 19:22

Did a Prime95, core3 got an error and stopped working after a few seconds, core0 got an error and stopped working after 9 minutes, the other cores worked perfect for 4h27mins.

Still getting 62GFlops in LinX and IBT with max settings, I'm going to format 2 of my HD's and put them in RAID0, do a fresh install on those and report back here to see if it fixed my GFlops speed.

Also, I will be increasing voltages from 1.3v to 1.31 to see if it fixes all my errors.

Another question: Should I change the LLC and PLL settings from Auto to something else?

AlfBaxter · 16 Oct 2011 at 20:25

StYleY said:
Did a Prime95, core3 got an error and stopped working after a few seconds, core0 got an error and stopped working after 9 minutes, the other cores worked perfect for 4h27mins.

Still getting 62GFlops in LinX and IBT with max settings, I'm going to format 2 of my HD's and put them in RAID0, do a fresh install on those and report back here to see if it fixed my GFlops speed.

Also, I will be increasing voltages from 1.3v to 1.31 to see if it fixes all my errors.

Another question: Should I change the LLC and PLL settings from Auto to something else?

Have you checked to see whether IBT is using all four cores? Have a look whether they are all at 100% in CoreTemp while IBT is running. I would also test for stability at stock speed first, then OC. IBT at stock on my system gives about 92 GFlops IIRC. See if you can run Prime95 without errors, then gradually increase the multiplier and VCore.

Don't know much about PLL, but sometimes increasing LLC can improve stability. I don't know the BIOS of that board, so not sure what other settings you have that you might need to change. Disable spread spectrum though, most say that's good for stability.

I just ran IBT 2.51 at 'high' for comparison, got around 114.5 GFlops in around 23s.

StYleY · 17 Oct 2011 at 10:16

AlfBaxter said:
Have you checked to see whether IBT is using all four cores? Have a look whether they are all at 100% in CoreTemp while IBT is running. I would also test for stability at stock speed first, then OC. IBT at stock on my system gives about 92 GFlops IIRC. See if you can run Prime95 without errors, then gradually increase the multiplier and VCore.

Don't know much about PLL, but sometimes increasing LLC can improve stability. I don't know the BIOS of that board, so not sure what other settings you have that you might need to change. Disable spread spectrum though, most say that's good for stability.

I just ran IBT 2.51 at 'high' for comparison, got around 114.5 GFlops in around 23s.

IBT is using all memory and cores @ 100%, still getting 63 GFlops, will try stock speeds and see what happens.

Anybody else who has advice on BIOS-settings?

JamesM · 17 Oct 2011 at 10:36

That does seem very low. With my 2500k at 4.6GHz I get 120GFlops so even at 4.5GHz you should be getting very close to that.

Did you reinstall windows with the new hardware?

You could try setting the 'Threads' to something other than auto, say 4 for example and see whether that makes a difference?

Seems more like a Win 7 SP1 issue though

Have a read here - http://www.overclock.net/intel-cpus/1106558-intel-i5-2500k-very-low-gflop.html

StYleY · 18 Oct 2011 at 08:40

I see, I will be reinstalling Windows on a new hard drive then update it totally, see if that fixes the problem.

Is it advised to change LLC or PLL?