Overclocking issue?

[Vaelic]

So I overclocked my i5 2500k from 4.5 GHz (which it has been at for years) to 4.6GHz. I also overclocked my EVGA SC GTX 980 using precision X using the following settings:

power target - 124%
GPU temp target - 84%
GPU clock offset - 133%
Mem clock offset - +0

I then ran 3DMARK Firestrike which went through perfectly and there was no artifacting.

The problem is, every now and again, a game will go to an orange screen and hard lock the PC. I reverted all overclock settings and the games work fine. My question is, do you believe it to be more of a GPU overclock issue or the CPU OC? If it's the GPU overclock, any recommendations on what I should lower the above settings to?

 

[nerp]

The only way to rule out one or the other is to deploy typical troubleshooting technique. Overclock just one and try to replicate the error. If no, remove the overclock and overclock the other. If neither allows you to replicate, then it's the combination of both somehow.

 

[BonzaiDuck]

Amazing. There are still Sandy Bridge processor owners squeezing extra life and usage out of a 6-year-old chip.

There may be a slight shortfall of i7-2500K performance over the HT-capable 2600 and 2700K's -- such as mine. I'd have greater insight into your CPU OC'ing experience if I knew what motherboard you used.

Generally, if OC'ing to 4.6, I would expect to see a peak "Turbo" voltage measured somewhere between 1.36 and 1.37V. The loaded Turbo voltage should be closer to 1.35, by a difference determined by your use of a load-line calibration or LLC feature. You should probably not adjust LLC to completely eliminate any vDroop, and I'd found that leaving a vDroop of a 20mv difference between loaded and unloaded Turbo is just about right.

I don't think I updated my sig to show my 2700K system. I have two GTX 970 cards (MSI Gaming) in SLI. After inserting the second of these configured to default and SLI, I was compelled to re-run all my tests for the CPU overclock. I don't think it required any further adjustment.

And I probably re-ran those tests after overclocking the SLI configuration itself. My 970s are set to offsets of 150Mhz core and 250Mhz memory. I was wondering why you haven't boosted the memory speed in tandem with the core. It's a back-and-forth adjust-and-test process, edging up the core first and then increasing the memory speed. At full-bore, the 970's are running at ~ 1480Mhz core and 7500Mhz memory. To get there, I had to push the voltage up by 20 mV -- or 10 mV beyond what proved to be stable.

You say your crashes are characterized by a lock-up or freeze. It would be helpful to see instead a "blue-screen" with a stop-code. Only based on my 970 experience and aware that I'm not familiar with the 980, I'd suggest backing down the power target to 110% and finding stable offsets from there. While I use the same temperature target as do you, my stress tests of the SLI-configured 970s don't peak the temperatures beyond 78C for the hotter card. The other one maxes out around 71C, if I remember correctly.