LTTP: 2500k overclocking

[Hornet1990]

Hi all, long time lurker, first time poster

As the title suggests I've (finally) gotten round to overclocking my 2500k to get some more life out of it before upgrading, but I'm not sure if I'm being limited by my motherboard or not. The spec is:

i5 2500k (one of the later production as Ivybridge was out and I picked this up cheap)
MSI Z68-GD65-G3 (latest beta BIOS for SB chips - N88)
2x8GB Kingston HyperX Beast 2400 (but the board only supports 2133)
MSI 760GTX
Samsung 850 pro 256GB
3TB WD Red
TRUE black (lapped) with a single Noctua fan
Corsair AX760 PSU

I've gotten it to 4.4GHz stable (8 hours RealBench with max memory, 4+ hours IBT with max memory) but to do that I've had to set the vcore to 1.360V in the BIOS which from my research seems pretty high with many getting 4.5 on 1.32V or less. And I've had to disable the C-states, enable them at any multiplier and I get instant instability.

All settings are as the guide here https://forums.overclockers.co.uk/showthread.php?t=18340310 other than vcore manually set (auto isn't completely stable at any multiplier over 40) and C states disabled.

Temps wise the hottest core hits 61degC in Realbench, 70degC in IBT, and there's ~10degC spread across the cores - this is consistent across all applications of paste so either I'm consistently poor at applying it or this is just how the chip is, and I did find a pic online of someone elses with exactly the same distribution of temps so not too worried about that.

Now I have found when warm (i.e. been running a few hours) I can set the multiplier to 45 (highest tried) and at the same vcore it will happily stress test for hours, but will not do 4.5GHz from cold boot no matter how much I up the vcore - 1.38V is the max I've tried. Also, when testing from cold it would usually be unstable around the 18-20 minutes mark in real bench if it was going to be which I found the consistency to be very odd, but if you got past that it would in most cases complete successfully. Even stranger after a failure at 20mins I could immediately rerun realbench and it'd run for hours successfully.

All this makes me think the board is a bit duff, and is possibly struggling to supply a stable voltage, particularly from cold. CPUz and HW_Info both report that the chip vcore is 1.352V, but at 1.355V in the BIOS I get the odd rare crash, at 1.360V I've not seen one yet.

I know it's a silicon lottery, but is there anything I could try to either reduce the vcore, get more MHz for the vcore, and/or be able to enable the C-states? My intention is to upgrade to a HEDT 6/8 core at some point in the future, but as I'm not being limited yet I'm not in any rush. When I do the 2500k will become a hyper-v/development server (back to stock clock) so I don't want to burn it out either.

I have found a Gigabyte GA-Z77X-UD5H available (albeit £150, but effectively free to me), would it be worth exchanging the board for this and seeing if its the CPU or the board? The Gigabyte also seems to have a lot more overclocking options in the BIOS than the MSI does which seems pretty limited.

Thanks

 

[TwsT]

1.36 is fine on sandybridge, some got better some got worse. aslong as you're not in the high 80's for temps when under full load you won't have any problems.



and yes Z77 boards are much better than z68 if for nothing else than native usb 3.0 still no guarantee your chip will do 4.5 but if memory serves me well the GD 65 series z68 wasn't a great clocker in general.

MSI's bios back then was just awful too.

 

[Hornet1990]

Thanks TwsT, that confirms what I thought and could find about MSI boards of the time - they may be better now but I'll probably give them a miss from now on.

 

[Disco_P]

I quite liked the bios on my Z77 MPower.

I'd enable PLL overvoltage and try upping the cpu vdroop control such that when you start a stress test your vcore doesn't plummet under load. OCCT in AVX Linpack is very good a showing vdroop once it has filled up your RAM.

 

[wam7]

If you want get the most of your 2500k then get an Asus Z68 or Z77. I used all 3 makes MSI, Gigabyte and Asus which had by far the best bios and results when it came to overclocking.

 

[Jumper118]

I wouldn't upgrade the motherboard to get a higher overclock, the cost/performance gain will not be worth it. If you do decide to upgrade the board then a z77 ud5h is a great board. I currently have 4 1155 motherboards and the only one which has failed was my z77 mower. My z77 ud3h, asrock z77 oc formula and maximus iv extreme have all been solid and both the z77 ocf and mive have been used for sub zero and have held up great.

 

[Disco_P]

I wouldn't upgrade the motherboard to get a higher overclock, the cost/performance gain will not be worth it. If you do decide to upgrade the board then a z77 ud5h is a great board. I currently have 4 1155 motherboards and the only one which has failed was my z77 mower. My z77 ud3h, asrock z77 oc formula and maximus iv extreme have all been solid and both the z77 ocf and mive have been used for sub zero and have held up great.

I've had 2 MPowers fail also. Now on a Sabertooth Z77 painted white.

Jealous of the Asrock OC Formula! Gorgeous motherboard.