CPU voltage absolute limits

[shift8]

Hi guys, got a question for anyone who may know.

 

My understanding is that aside from heat, over voltage-ing a cpu kills it due to electron migration? So I guess my question is, assuming you can keep the temps down, can you take the voltage of a cpu up really high to increase the overclock for short periods of time? Like if I only wanted the cpu to last for a month, can I run it at 1.5 to 2 volts?

[imreloadin]

1 minute ago, shift8 said:

Hi guys, got a question for anyone who may know.

 

My understanding is that aside from heat, over voltage-ing a cpu kills it due to electron migration? So I guess my question is, assuming you can keep the temps down, can you take the voltage of a cpu up really high to increase the overclock for short periods of time? Like if I only wanted the cpu to last for a month, can I run it at 1.5 to 2 volts?

What CPU is it? Each architecture has different voltage limits.

[TheRandomness]

FX series chips can go to 2V under LN2 iirc. 

[shift8]

4670k

[themctipers]

1 minute ago, shift8 said:

4670k

i wouldn't go over 1.35v

 

[imreloadin]

Just now, shift8 said:

4670k

The highest I'd take it is 1.4volts on the core, if you don't care about the lifespan of the CPU then the sky is the limit.

[BrightCandle]

The last time I saw Intel make any recommendations around overclocking voltage was with Sandy Bridge where they set their maximum recommended as 1.35V. There were at least 2 people on overclockers.net's forums who were watercooling and had their CPUs instant pop at 1.4V but then there were also liquid nitrogen results at 1.6V and a variety of water coolers using 1.4V for years.

 

I used about 1.4V on an i7 920 at 4Ghz with custom watercooling and started getting degraded clockspeeds after 2.5 years, within 3 years it couldn't do stock clockspeed and at default voltage it was terrible. But I ran a 3930k@ 4.4 Ghz for 3 years and never saw signs of any degradation at all. A large part of whether it degrades and how it does so is based on the silicon lottery.

 

So you could go high voltage and get an instant pop or it could last 10 years and pretty much every outcome in between. Considering the voltages modern CPUs ran at is quite a bit lower than Sandy Bridge used I would hazard a guess that 1.35V is no longer even remotely safe from instant pops or fast degradation.