Help Overclocking 6950x and DDR4 3200MHz

[CryWolf]

Hello,

I'm a very happy owner of:
CPU: i7-6950x

Memory: 32g DDR4 3200MHz Corsair Vengeance (4x8g)
Water Cooler: Corsair H115i
MOBO: ASUS Rog Strix x99 Gaming LGA 2011-v3
Power Supply: XFX 1250W PRO Series
GPU: GTX 1080
HD: SSHD (2x2T) and 480g SSD Kingston UV400

And I never did an overclock in my life, I think it's good to point this out.

 

So, I'm trying to overclock my memory to 3200MHz, because the MOBO runs it at 2400MHz at default, my processor to 4,2GHz and cache ratio to 3,2GHz. Here some pictures from my BIOS "AI Tweaker". The temperature is fine, around 75 degree celsius.

The problem is that I can't get it to stabilize. I'm trying to run Prime95 stress test and after 10 minutes or so my computer freezes. Could you guys please give me some advices where I can adjust to make it perfect? Could it be memory? Could it be too much or too low voltage?

Thank you.

[ir_cow]

Well it looks like from the BIOS pictures your XMP profile for the memory is enabled and running at 3200 so you are set on that. I haven't played around with the i7-6950x so without some reading I won't be able to help you on the overclocking side of things. I don't even know the safe voltages offhand but it is something you should know before getting crazy with it.

[xenkw0n]

Yea I would try going down to 41 on the multi with those voltages. Remember this is a 10-core chip and scaling gets pretty steep past a certain point. The rest of your overclock looks alright, it is cranking up your System Agent voltage and VCCIO but they're still within spec. Maybe try to set CPU Cache around 1.05v as well.

[CryWolf]

Hello guys. After some heavy studying, I think I made it!

 

Right now I'm running this configs with this temperatures:

*Overclock @ 4.0GHz and 4.1GHz(core 6 and 10) (1.20V)

Prime95 v28.10: Small FFTs: 91C - 25 minutes test

RealBench v2.43: 67C - 1 hour test

AIDA64 v5.7.38: 69C - 15 minutes test

OCCT v4.4.3 (Large Data Set): 77C -15 minutes test

OCCT v4.4.3 (Small Data Set): 82C - 7 minutes test - test stopped because a error was detected. Could this be an instability problem? Should I worry about this?

*** This seems to be the perfect one for me. Temperature at Prime95 is a bit high but I'm probably not going to stress the CPU as Prime95 does.

 

I'm concerned about the OCCT Small Data Set test that failed. I don't know why. What do you guys think?

Here is everything I took note so the next person with the same issue might take benefit:
 

Video that used as reference to start:

 

BEST CPUS: 6 and 10

 

Room temperature: 24C

 

*Overclock @ 4.0GHz (1.20V) - AVXII downclock enabled to 3.8GHz 

Prime95 v28.10: Small FFTs: 86C (3,8GHz)

RealBench v2.43: 63C (4,0GHz)

AIDA64 v5.7.38: 66C (3,8GHz)

OCCT v4.4.3 (Large Data Set): 69C (3,8GHz)

OCCT v4.4.3 (Small Data Set): 76C (3,8GHz)

*** Here we get very nice temperatures, even for Prime95 v28.10 Small FFTs.

 

*Overclock @ 4.0GHz (1.20V)

Prime95 v28.10: Small FFTs: 90C - 5 minutes test

RealBench v2.43: (4.0GHz): 69C - 3 hours and a half test. 

AIDA64 v5.7.38: 70C - 30 minutes test

OCCT v4.4.3 (Large Data Set): 77C - 30 minutes test

OCCT v4.4.3 (Small Data Set): 82C - 30 minutes test

*** Here we have a good overclock and temperatures are ok. I failed RealBech after 3 hours and a half but I think is a software problem. The system didn't BSOD or anything else. Do you guys think that the Realbench test failure could be something? Could it be really an instability?

 

*Overclock @ 4.0GHz and 4.1GHz(core 6 and 10) (1.20V)

Prime95 v28.10: Small FFTs: 91C - 25 minutes test

RealBench v2.43: 67C - 1 hour test

AIDA64 v5.7.38: 69C - 15 minutes test

OCCT v4.4.3 (Large Data Set): 77C -15 minutes test

OCCT v4.4.3 (Small Data Set): 82C - 7 minutes test - test stopped because a error was detected. Could this be an instability problem? Should I worry about this?

*** This seems to be the perfect one for me. Temperature at Prime95 is a bit high but I'm probably not going to stress the CPU as Prime95 does. I'm concerned about the OCCT Small Data Set test that failed. I don't know why. What do you guys think?

 

*Overclock @4.1GHz (1.225V)

Prime95 v28.10: Small FFTs: 92C (4.1GHz) - 5 minutes test

RealBench v2.43: 63C (4.1GHz) - 15 minutes test

AIDA64 v5.7.38: 69C (4.1GHz) - 10 minutes test

OCCT v4.4.3 (Large Data Set): 76C (4.1GHz) - 10 minutes test

OCCT v4.4.3 (Small Data Set): 83C (4.1GHz) - Stopped after 4 min saying "Error Detected"

*** Here the temperatures for Prime95 v28.10 Small FFTs was a little bit high and the voltage was not yet stable as we can see on OCCT Small Data Set. Probably we could get it stable at 1.23V ~ 1.25V, but the temperatures would rise even more. Noticed that the temperatures are not that accurate since I didn't spend too much time in each test. 

 

*Overclock @4.2GHz worked at 1.3V.

At 1.28V some of the processors would drop down usage.

Prime95 v28.10: Small FFTs: 100C (at 1.28V or 1.3V)

*** Not good for my setup. Probably with a custom liquid system we could get it to work.

Edited May 19, 2017 by CryWolf