Showing results for 
Search instead for 
Did you mean: 

High CPU Voltage Question

Level 7
I am running a stable overclock of 4.6GHz but need voltage @ 1.41v to do it. As you can see from by build specs I am running a custom water cooling setup. I have re-applied thermal paste a few times but no improvement.
CPU temps are low 30's at idle and mid to high 70's under full load. I've tried reducing voltage but then it is unstable. To reduce voltage I have to reduce the overclock.

A couple of questions:
Am I killing my CPU with this voltage (even though temps are fine)?
Is there a way to reduce the voltage but keep the overclock that I might not have tried?
Will delidding help? Yes, I realize it would void the warrantee - don't really want to do that.

I'd appreciate any feedback!

My build:
ASUS Maximus VI Formula
Intel Core i7-477k 3.5GHz @ 4.6GHz stable
G.Skill Ripjaws X 16GB (4X4) DDR3, 2400
Samsung 840 SSD 500GB
EVGA SuperNOVA 1000W P2 Platinum PSU
Custom water cooling loop:
XSPC components (mostly)
cooling Northbridge, CPU, GPU

Level 16
I think that the CPU is running very close to the border with the unsafe.... voltage is clearly higher than it is recommended and temps are getting there too - I would not recommend keeping or pushing those settings for 24/7 usage... that CPU is simply not that CPU which can do it... you know, the chip lottery...

The safest is to make step back with frequency and voltage...

Level 13
Well, your CPU is better than mine. I need 1.50 volts for stability at 4.6GHz. With that I get 30C core voltage at idle. You don't say what load gives high 70s temperature. When I do Aida64 FPU stress, the CPU pulls 186 watts and rises to 85C. That's an extreme. The point is not that those temperatures and voltages are good, but that they're normal for a dog of Haswell.

The recommendation is to run benchmarks if you're competing with Realbench or at HWbot at the highest voltage, stable clock and temperature you can stand. The limit is 100C where the CPU will throttle back. Then, for 24/7 use, cut back as suggested above. I need to back off to 4.4GHz to get core voltage down to 1.30. Temperatures drop accordingly to the 50s in stability tests.

There is more to be gained from delidding than from a second radiator. The biggest problem with cooling Haswells is inside the lid. The adhesive around the edge of the lid adds some separation between the chip and the metal lid. Intel filled that space with a so-so TIM. Delidding, removing the adhesive and replacing the TIM all work to improve heat flow. Such an effort needs a really good TIM. Arctic Silver and Gelid are good. Liquid metal is 9 times better, but adds to the risk because it's electrically conductive. I think, though, that liquid metal TIM is what lets me get high peak power with tolerable temperatures.