Thoughts on Phenom II and overclocking

[soccerballtux]

Does the fact that Intel uses Hi-K gates now affect safe voltage levels with them? If it's unsafe to push a Penryn past 1.4V (and assured destruction to go to 1.5v), does that mean we won't be able to go past 1.4 with AMD's 45nm chips as well? Why?

 

[SlowSpyder]

Unless I understood him wrong, I thought aigomorla said that CPU-Z was reading the voltages wrong. Could be wrong on that though.

 

[SickBeast]

The highest I've seen someone volt the P2 was 1.475v.

It could probably take 1.5v, but I don't think I would go any higher than that.

 

[BLaber]

For both Phenom 1 and Phenom II , 1.5v for daily use on exotic air cooling is ok.I have been running phenom 9950 @ 3.4 ghz / 1.45v for some good time now.

 

[Cookie Monster]

Im not too sure on the indepth details of the difference between intel's high-k 45nm process to AMDs 45nm (still SOI), but you can probably guess that both process will behave differently and will have different voltage limitations. So its premature to assume that voltage limitations found on the Penryns should be applied and carried over to the Deneb.

Hopefully AT will have an indepth article about this just like they did with the Penryn.

 

[Soulkeeper]

1.5v sounds insane to me

 

[Elfear]

Didn't AMD say that 1.55V would be pretty safe on air?

 

[Rhoxed]

with AMD i would say anywhere from 1.5 - 1.6 on air as long as temps are reasonable

 

[SickBeast]

Didn't Anand take it up to 1.575v on the stock cooler to hit 4ghz? I'm pretty sure someone did.

 

[Martimus]

It is comparing apples to oranges. AMD designs their 45nm chips to run at 1.35V, while intel designs theirs to run around 1.1V. So 1.5V would probably be safer for AMD than Intel just based on that. At the same time, I don't have one of these chips, so I would just be speculating on the safe operating voltages.

 

[soccerballtux]

Originally posted by: Rhoxed
with AMD i would say anywhere from 1.5 - 1.6 on air as long as temps are reasonable

This is on 45nm though!!!

SickBeast, Anand also took an e8400 to 1.5v. They hit 4.7Ghz IIRC; but it fried after about 20 hours.

 

[Fox5]

Originally posted by: Martimus
It is comparing apples to oranges. AMD designs their 45nm chips to run at 1.35V, while intel designs theirs to run around 1.1V. So 1.5V would probably be safer for AMD than Intel just based on that. At the same time, I don't have one of these chips, so I would just be speculating on the safe operating voltages.

I think AMD chips have generally not handled as big of swings in voltage as well due to the IMC and SOI. The IMC is very sensitive, and SOI I think just fails quicker under high voltage.

 

[nyker96]

I think I remember Anand took it to 1.5v, but you have to keep in mind, how safe the chip will be at high volt depends on design a lot. AMD has different design than Intel, this number will not be the same. Still I will be cautious and leave it at 1.4v initially before raise it some more unless your temp is pretty low at that voltage.

 

[aigomorla]

Originally posted by: SlowSpyder
Unless I understood him wrong, I thought aigomorla said that CPU-Z was reading the voltages wrong. Could be wrong on that though.

correct cpu-z does not show voltage properly.

this is directly from my friend who was writting a review for one.

"Real stock Vcore is 1.3125 V as selected by VID, 1.32 real.
Real Vcore on 2nd screenshot, at 3600 MHz is 1.4000V VID and 1.41 real - but this chip will do 3600 with less voltage after some tweaking."

The chip on cpu-z reported 1.520

so yes the cpu-z voltages are wrong..

And this was taken on cpu-z 1.49

 

[Idontcare]

Originally posted by: aigomorla
so yes the cpu-z voltages are wrong..

That has always been true though, just as DTS based CPU temps are never accurate except when CPU temp happens to also be TJmax temp. What seems relevant here is the apparent magnitude of the inaccuracy (approaching 0.2V).

The seemingly more important question, which you answered so thank you very much, is whether the BIOS selected Vcc is in fact the delivered Vcc. Knowing that this is the case means that we aren't entirely blind to the actual Vcc but rather we just need people who are reporting their PhII Vcore's to go the extra mile and report the Bios set Vcc instead of CPU-z Vcore.

Doesn't the AMD overdrive utility report Vcc? Is it less inaccurate than CPU-z?