6GHz Phenom II Overclock on LN2, 5Ghz on Dry Ice

@ 2008/11/20
As we all know, the Phenom II will be AMD's second 45 nm part (the first is officially the Shanghai server part) that is scheduled to hit Q1 09. This heavily redesigned processor based on the Phenom architecture has already showed some significant per clock performance gains over the older Barcelona product, and AMD considers this to be their best design ever in terms of performance, functionality, and eventual clockspeed. The first desktop products are expected to hit 3 GHz at the top end, but AMD has been hinting that the Phenom IIs can do a lot more in the hands of an enthusiast.

The first leaks were that the Phenom II could hit 4 GHz on air cooling alone, though we obviously wonder how extreme that air cooling is. Well today AMD had some actual demonstrations at their gathering, and the Phenom II was able to hit 5 GHz at 1.6v by using dry ice cooling. Dry ice is CO2, and it goes directly from solid to gas (sublimation) at a temperature of -109.3 F (-78.5C for the rest of the world) at 1 atmosphere. So with some pretty extreme cooling (non-LN) the Phenom II does show that it has some legs in the clockspeed department.


Comment from jmke @ 2009/02/06
Comment from Kougar @ 2008/11/22
Hum, sorry JMke, I meant Geoffrey.

Quote:
Originally Posted by Massman View Post
I don't trust Anand's reviews that much, he has been proven incorrect when comparing the Yorksfield with the Nehalem in the past, favoring the Nehalem too much.
He has also constantly written articles against Intel during the Pentium era and never seemed unwilling to pull punches. Regardless, there is no Nehalem system in their results. Just a 3Ghz Yorkfield 8-core server against a 2.7Ghz Phenom II 8-core server. The results are really closer than I would have expected, which is good for AMD... but they are not enough either in my humble opinion. Also insufficient results for a full comparison, but it is a start.

Massman, that's fine as I don't base my views off any sole site's results either, THG and Tweaktown made doubly sure of that. But do you know of any other sites that have done major Shanghai review comparisons...
Comment from jmke @ 2008/11/22
Quote:
Originally Posted by Kougar View Post
And to be fair JMke, I am sure AMD will have dual-core chips priced competitively.
why are you addressing this reply to me? I didn't post anything in this thread
Comment from Massman @ 2008/11/22
I don't trust Anand's reviews that much, he has been proven incorrect when comparing the Yorksfield with the Nehalem in the past, favoring the Nehalem too much.
Comment from Kougar @ 2008/11/22
Quote:
Originally Posted by Massman View Post
Apparently it's CPC as fast as the Nehalem, so if AMD manages to launch their products at similar price ratios as the first phenoms ... it's better than Nehalem
I'm not so sure. Anand already did 2socket 8core tests of Shanghai against Penryn, it was competitive but with plenty of caveats.

Against Nehalem it swings back to Intel's favor, because Intel will lose the power hungry FB-DIMMs and should gain significantly better performance scaling compared to Penryn. I expect Nehalem to dominate 4-socket systems....

And to be fair JMke, I am sure AMD will have dual-core chips priced competitively. They just won't have many of them unless they specifically make dual-core only chips.
Comment from geoffrey @ 2008/11/21
Just don't forget the Intel E8x00 series, mainstream priced cpu's which scale up to 4GHz rather easily, at these clockspeeds you have a very fast setup for daily usage/gaming. Quad cores will be either slower, cost more, use more power, require more volts and better cooling ... I honestly don't think AMD will sell their PII cheaper then Intel's duals. In other words, not that interesting for daily usage/gaming/apps which don'ts scale with multiple cores.

In other market segments, workstations for example, I don't think there will be too much overclocking going on, important is that AMD can get hold of the performance/watt crown at near 3GHz, with the correct price AMD can be very competitive. Just one thing, don't expect that these cpu's will come for cheap, expect to pay premium prices like we're used to with FX cpu's. If AMD can get more money of their product they won't hesitate, they're not better then Intel
Comment from Massman @ 2008/11/21
Apparently it's CPC as fast as the Nehalem, so if AMD manages to launch their products at similar price ratios as the first phenoms ... it's better than Nehalem
Comment from Kougar @ 2008/11/21
Quote:
Originally Posted by Oberon View Post
I sure am glad that all of you people saying this didn't happen were there to see it and everything.
I never said it didn't happen.

Quote:
Originally Posted by Massman View Post
AMD back in the game?
Pure performance, no. Factoring in price, sure.

It required 1.9v to reach ~6GHz. That's supposedly better than Nehalem, which is neat. But for the 99.9% of users that don't use LN2 or dry ice, I'm waiting to see what typical overclocks are.

If 1.6V for 4GHz holds true then it is no better than Nehalem. AMD would still be playing the cost/performance game.
Comment from npp @ 2008/11/21
"Cameras weren't allowed at the event, and so no clear pictures were taken..." Then I guess they could have taken it to 7Ghz s well...
Comment from Massman @ 2008/11/21
Well, apparently it IS true. Tony confirms the speeds and apparently Macci was controling the setup to, indeed, -196°C ... no coldbug :-O.

Apparently it's CPC as fast as the i7 as well

AMD back in the game?
Comment from Oberon @ 2008/11/21
I sure am glad that all of you people saying this didn't happen were there to see it and everything.
Comment from Kougar @ 2008/11/21
Quote:
Originally Posted by Massman View Post
Amd uses SOI, Intel uses High-K.
AMD was talking about using their IBM/AMD version of High-K since 65nm, they've simply pushed it back repeatedly. Whether you believe them or not they stated the next chips after Phenom II will use it.
Comment from Massman @ 2008/11/21
Amd uses SOI, Intel uses High-K.
Comment from geoffrey @ 2008/11/20
High-K, or what they don't have
Comment from Massman @ 2008/11/20
The biggest problem is not the actual frequency (which is quite unbelievable as well, seeing 4GHz is still nearly impossible), but the scaling with temperatures. AMD's production process makes it close to impossible to run your cpu close to -50°C, let alone -150°C.
Comment from npp @ 2008/11/20
Heavily redesigned? Ah, come on... This reminds me of the 3Ghz Phenom AMD showed some many months ago, unfortunately it got lost somewhere on the way. 4Ghz shouldn't be a problem for the second version, I guess.