Help to Overclock i5-2500k on MSI P67A-G45

mikeo · 17 Jun 2011 at 22:51

IdiotCoder said:
Seems to be Intel C-State under CPU Features, at first this kept somehow being disabled automagically in the BIOS.
Enable it and you should notice the multiplier lowering down to 16 when idle.
I also have EIST enabled.
.

I 2nd that.

AgedIdiot · 18 Jun 2011 at 08:16

IdiotCoder said:
If you just want 4.2GHz at less than 1.3V, try out what mikeo did.
Post #14

As mentioned above did try this but it did not take.
Probably the case that I have not turned off / left something on in the Bios which I should have altered .
Will hopefully find out a little later today !K

AgedIdiot · 18 Jun 2011 at 19:51

It was Intel C-State which the board had disabled - now enabled and working as it should. My thanks for the excellent advise :cool:

Next - clocking here we come

bulldog147 · 18 Jun 2011 at 20:13

Speed said:
I'd also recommend a few hours of LinX (All Memory), I tend to use short bursts of Prime95 and LinX to see if it is stable. LinX first, 30 minutes then Prime95 blend 1 hour. If it passes both go for longer stretches. What I've found is certain overclocks fail on LinX quickly or Prime95 quickly, but they seem to pick up different issues. Hence why short bursts of one after another is a good indicator that the clock is stable. No point doing 8 hours of Prime95 if LinX makes it fail after 30 minutes.

I'd suggest being quite careful leaving things like vCore on Auto, you need to keep an eye on it at higher clocks as it tends to push things up to silly levels.

There is also no real need to change much more than Muliplier, Vcore and CPU I/O (to bring the memory speed up). Turbo, EIST, C1E can all be left to their default settings for regular overclocks (sub 4.8GHz) in my experience.

well said speed. I have to agree with him. I never use ALL in Auto. I prefer manual overclock setting. I think the safer maximum vcore on side side is no more 1.375v (most do 1.40v but that's probably will shortern lifespan or could degraded the speed itself) So, far, I had stable 4.6GHz at 1.35vcore at bios but 1.37v in windows but I think I cannot risk going for 4.8 because it might end up pushing more vcore (too risky)

Tyrone77 · 19 Jun 2011 at 12:48

Hi,

I've overclocked my i5 on the same motherboard to 4.4Ghz, 1.27V. Ran prime for 3hrs and ok so far.

Have Hyper 212 + cooler

The strange thing is that one core 0 has temps 15 - 18 degrees cooler than the other three!?!

Anyone know if that's normal?

Phenomenologica · 19 Jun 2011 at 12:59

Tyrone77 said:
The strange thing is that one core 0 has temps 15 - 18 degrees cooler than the other three! Anyone know if that's normal?

That's a bit of a steep difference. Are you sure the cooler is mounted properly? Might be worth taking it off, cleaning it, and reapplying the thermal paste.

Tyrone77 · 19 Jun 2011 at 13:55

Phenomenologica said:
That's a bit of a steep difference. Are you sure the cooler is mounted properly? Might be worth taking it off, cleaning it, and reapplying the thermal paste.

Yeah I should probably do that.

I'm hitting 62degrees on three cores and core 0 is at around 45 degrees... and thats on the overclock on prime.

Speed · 19 Jun 2011 at 16:48

More likely a faulty or stuck core temp sensor. You could try checking the underside of the CPU and the motherboard pins for damage or thermal grease. The only way to be certain would be to try the CPU in a different board.

That sort of difference is unlikely to be a mounting problem, although it never hurts to remount and see if that sorts it.

NickM · 19 Jun 2011 at 18:17

mikeo said:
Right....

Easiest way to OC on this board (mines a P67A-GD65) if you are not looking for the absolute MAX but just a decent stable OC.

Please note that this is only my personal opinion but I do after all have a similar MSI board.

Forget about changing the CPU ratio and fiddling with the voltages (or anything else come to that).

Load default safe settings, making sure you set anything that you definately need (like AHCI for the SATA ports).

Go to the OC setting menu, scroll down to the bottom and select CPU Features. Then change the turbo settings on the 4 cores from 37/36/35/34 (assuming here they will be set as per my board) to say 42/42/42/42. Leave voltages and EVERYTHING ELSE unchanged. Save your changes and re-boot.

JOB DONE.

What you will have now is a system that throttles back the volts/ multiplier when not under load but ramps up to 4.2 GHz under load.

Obviously run IBT and Prime95 to test stability.

I found exactly what you did. If you change the CPU ratio in the top level OC menu. It applies this OC "all the time" and the CPU voltage is around 1.368v. A bit on the high side for this sort of OC .

Doing it as I suggest. Gives me a stable system with a decent OC. Interestingly, CPU voltages never go above 1.288V (much better) and temps in Prime95 after 6 hours are only around the 60C level. Obviously you need a half decent cooler. Mines the Gelid in my sig. (pretty popular at the moment).

In my opinion, I don't think you need to do it the "traditional way" (IE. fiddle with loads of different settings) unless you want to hit really high OC's.

Only my opinion folks. But it works for me.

Interesting way of doing it. I tried using OC Genie but my voltages went up to 1.36-1.38V @ 4.2GHz which seems kinda of high (never OC'd before, voltages before were 1.00-1.23V). Think I'll give your method a try and see what happens. Out of curiosity, why does turbo boost the cores to different levels by default?

mikeo · 19 Jun 2011 at 18:42

NickM said:
Interesting way of doing it. I tried using OC Genie but my voltages went up to 1.36-1.38V @ 4.2GHz which seems kinda of high (never OC'd before, voltages before were 1.00-1.23V). Think I'll give your method a try and see what happens. Out of curiosity, why does turbo boost the cores to different levels by default?

Indeed it is. And works well for me. 8 hours prime95 (blend) and average temp across the 4 cores was 58C.

My experience with OC Genie was identical to you re-volts (IE. 1.36-1.38v). Way too high for 4.2GHz in my opinion. Strangely though (as I commented earlier) if I just change the turbo boost ratios all to 42. Max voltage is now a much better 1.288v max.

I just can't see the point myself in fiddling with all the "usual" setting for OCing if you only want a half decent OC like this.

Only my opinion of course and each to their own. But I'm always for the easy life myself.

Good luck.

Re- the staggered standard turbo ratios. Only guessing here but I think it might be something like... If single core being stressed, it turbo boosts core 1 to 3.7GHz, if two being stressed, it boosts them both to 3.6GHz etc.
I suspect Intel just playing it safe. Crazy when you consider what these processors seem capable of!

NickM · 19 Jun 2011 at 19:07

Well it boots

(and reduces the multiplier to x16 on idle!), going to run Prime95 for an hour and see what happens. If it doesn't cause problems that's definitely a very easy way to OC with reasonable settings.
VCORE so far hasn't gone above 1.288V as you said, CPU core temps are already at 50-56C though

mikeo · 19 Jun 2011 at 19:20

NickM said:
Well it boots (and reduces the multiplier to x16 on idle!), going to run Prime95 for an hour and see what happens. If it doesn't cause problems that's definitely a very easy way to OC with reasonable settings.
VCORE so far hasn't gone above 1.288V as you said, CPU core temps are already at 50-56C though

Good show. Anything below 60C is pretty good from what I can gather.

Out of interest is it "blend" you are running? and what CPU cooler have you got?

NickM · 19 Jun 2011 at 19:31

Yea it is blend.
I've got the Akasa AK-CC017V2 Freedom Tower. At stock, core temps were around 32-40C on light use (watching a video), I think the cooler needs a bit of a dust, going to wait till my dust filters arrive then clean everything. I use PWM (target 45C, min speed 37.5% - its quite noisy at 100%!) so it would probably be cooler if I had the CPU fan running at 100% all the time.

edit: Just finished prime95 60min test. No errors/warnings, max VCORE 1.288V, max core temps 54/58/59/57. Now its stopped the VCORE has dropped back to 1.00-1.05V and x16 multiplier, 34-38C core temp. Great method, thanks mikeo!
2 of the cores did less tests in prime95 though, is that normal? (core0 - 77 / core1 - 74 / core2 - 67 / core 3 - 77 tests)

bulldog147 · 19 Jun 2011 at 20:14

Mine is @ 1.27vcore for 4.2GHz and temp read out 52/56/57/53 with Corsair H50 (2 x vipers fan as a push pull but fans never spin more than 1250rpm. Prime blend for 8 hours. Much better than bloomfield i7 920 @ 4.2GHz. Slightly improvement on sandybridge. Cinebench R11.5 score 8.08 points for i7 2600k while 6.94 points for i7 920. Big different

Thermal Paste is Thermalright Chill Factor II in the middle of sandybridge cpu with a pea sized.

Today after prime blend stable after 8 hours of 4.60GHz @ 1.36vcore read out core temp of 57/61/62/58 pretty good indeed. I am now settle with 4.6 rather than 4.2. Cinebench R11.5 score 9.27 points.

mikeo · 19 Jun 2011 at 21:44

NickM said:
2 of the cores did less tests in prime95 though, is that normal? (core0 - 77 / core1 - 74 / core2 - 67 / core 3 - 77 tests)

I'm just making an educated guess here. But the multiplier does appear to fluctuate on each core during testing (IE. it's not always x42 across the board). So I have a suspicion that could explain it. Unless anyone else has a better theory. Nothing to worry about I suspect.

mikeo · 19 Jun 2011 at 21:49

bulldog147 said:
Mine is @ 1.27vcore for 4.2GHz and temp read out 52/56/57/53 with Corsair H50 (2 x vipers fan as a push pull but fans never spin more than 1250rpm. Prime blend for 8 hours. Much better than bloomfield i7 920 @ 4.2GHz. Slightly improvement on sandybridge. Cinebench R11.5 score 8.08 points for i7 2600k while 6.94 points for i7 920. Big different Thermal Paste is Thermalright Chill Factor II in the middle of sandybridge cpu with a pea sized.

Today after prime blend stable after 8 hours of 4.60GHz @ 1.36vcore read out core temp of 57/61/62/58 pretty good indeed. I am now settle with 4.6 rather than 4.2. Cinebench R11.5 score 9.27 points.

Looking good.

Out of interest, as I see you are running XMS3 memory. What speed and voltage are you running it at? You will know doubt be aware that there is a lot of discussion about memory voltage on Sandy Bridge.

NickM · 19 Jun 2011 at 22:13

mikeo said:
Looking good.

Out of interest, as I see you are running XMS3 memory. What speed and voltage are you running it at? You will know doubt be aware that there is a lot of discussion about memory voltage on Sandy Bridge.

I'm running 8GB of XMS3, I just enabled XMP for mine so I think its running at 1600MHz/1.65V.

mikeo · 19 Jun 2011 at 22:23

NickM said:
I'm running 8GB of XMS3, I just enabled XMP for mine so I think its running at 1600MHz/1.65V.

Did it set the correct timings? The XMP on my board / BIOS choose 9/10/10/25, rather than what it should have been (IE. 9/9/9/24). Even though CPU-z shows the correct timings for the XMP profile. Strange it is! So I just set the speed / timings manually. By all accounts XMS3 is capable of running at 1600 MHz at 1.5v, or even lower if you are lucky. Mine set at 1.521 passes an hour of 4 occurance of Memtest running. Though a lot of folks say memory running at 1.65v should be fine. It's just the word "fine" that got me thinking. Hence why I tried it at a lower voltage.

NickM · 19 Jun 2011 at 22:34

mikeo said:
Did it set the correct timings? The XMP on my board / BIOS choose 9/10/10/25, rather than what it should have been (IE. 9/9/9/24). Even though CPU-z shows the correct timings for the XMP profile. Strange it is! So I just set the speed / timings manually. By all accounts XMS3 is capable of running at 1600 MHz at 1.5v, or even lower if you are lucky. Mine set at 1.521 passes an hour of 4 occurance of Memtest running. Though a lot of folks say memory running at 1.65v should be fine. It's just the word "fine" that got me thinking. Hence why I tried it at a lower voltage.

Well according to the Memory tab of CPUz (not the SPD profile tab) its running at 789.2MHz / 1:6 / 9/9/9/24 / 2T (MSI GD65 bios v 1.11 20/04/2011). So it seems to be on the correct timings, haven't checked in the actual BIOS though.

For the voltages, the Corsair website spec says its rated at 1.65V so even if that is to much voltage and the memory dies surely it will be covered by warranty?

mikeo · 19 Jun 2011 at 22:59

Looks like you are running on a later BIOS to me. I'm still on the version it came with (IE 1.9). Did your board come with that BIOS, or have you flashed it to the latest? I see it mentions improved memory comparability as one of the changes. So maybe that's why your memory on XMP is detected correct. Not like I'm worried about it, just interested.

Question. If you did flash your BIOS. Which method did you choose?

It's not the voltage to the memory that's the "possible" issue. It's that with the memory controller on the CPU now, it's technically only rated at up to a max. of 1.5v (or so I've read) for memory voltage. So it would be the CPU that that could be damaged. Hence my query about the memory voltage.

But as I said, loads of folks say it "should be OK" at 1.65v. Probably not a real issue.