Place your bets people.........

[Brian y.]

While waiting for my 965 to return from rma I decided to pick up a i7 920 to play around with. Not turning out to be anything great by any standard I decided to run a little test.

I have been wondering how long one of these little bad boys would last under some "extreme" CPU PLL/VTT voltages.

A quick screenshot of one hour in.



Let the betting begin.



*Disclaimer*
Only one CPU was injured in the making of this thread.

[Falkentyne]

You have to overclock it if you expect to see any adverse effects within any reasonable length of time.

These things can last forever at stock....

[GRAFiZ]

You have to overclock it if you expect to see any adverse effects within any reasonable length of time.

These things can last forever at stock....

Nah dude, he's not testing CPU Vcore... he's testing the CPU PLL/VTT.

This is very interesting indeed, tons of reports saying high values will kill your CPU quicker than a high vcore, but other than Anandtech, ther really isn't any data.

While this might be a waste of a chip and expensive it's very interesting and I hope he returns some results!

Good show dude!

[mikeyakame]

This proves nothing.

We all know that too much QPI/Uncore Voltage will kill a chip. Intel have burnt through potato sacks of them to prove this.

Too much CPU PLL will potentially damage the internal PLL on the CPU and make clock jitter worse at high frequency. Thats common sense, don't need to kill a CPU to know that voltage sensitive circuitry such as a PLL circuit is going to either degrade or have negative gains with too much voltage.

Internal PLL design specs aren't nearly as open to interpretation as the other voltages on the CPU. Its a closed feedback circuit and those design parameters are there to make sure the feedback and compensation works as it should. Throwing 2.3v into the CPU Internal PLL is only going to make feedback values much less accurate and compensation values rely on accurate clean feedback to do their job. 2.0-2.1V should be fine for a chips lifetime IF absolutely necessary, but most of the time it'll be overkill and the problem isn't in the chip its in the board design itself.

[Falkentyne]

Nah dude, he's not testing CPU Vcore... he's testing the CPU PLL/VTT.

This is very interesting indeed, tons of reports saying high values will kill your CPU quicker than a high vcore, but other than Anandtech, ther really isn't any data.

While this might be a waste of a chip and expensive it's very interesting and I hope he returns some results!

Good show dude!

Well, I'm aware that he's not testing vcore, but my point still does stand: in that case he should overclock the chip to the highest possible setting on stock vcore, to determine the effects of this.

Damage is first going to be seen at higher clocks, that's pretty self explanatory. If the chip doesn't die outright. By overclocking as far as he can with stock vcore, that will cut down his testing time.