[Waragainstsleep]

Simple enough question really.....

[Krypton]

90 nm 970fx

[Waragainstsleep]

OK, so the Xserve G5 has had up to 2 FXs @ 2GHz since January.
Given the Xserve has a much smaller enclosure than a PowerMac, it seems fair to assume that the same (rated 2GHz in Xserve) chip would be able to run slightly faster in a bigger enclosure. So how much faster? 10%? That would be dual 2.2GHz PowerMac if they had put the FX in it in January.

What if you then took this dual 2.2, and water-cooled it? How much faster would it go then? Another 10% would put it at 2.42GHz. If you could squeeze an extra 14% instead of just 10, you've hit 2.5GHz.

Steve understandably didn't dwell on the failure to meet the 3GHz target in his keynote last week, citing problems with the 90nm process, without being very specific. He did point out that the percentage speed increase from the previous 970-based PMG5s, was bigger than the increase achieved by Intel's P4 in the same period of time. But if you water-cooled a P4, I expect the difference would be all but gone.

If these figures, which I freely admit are no more than partially educated guess, are anywhere near the mark, then the actual speed increase in the FX since January is in the region of zero. I find it difficult to believe that IBM couldn't squeeze anything more out of their flagship CPU in six whole months.

Another recent development, which may seem unrelated, is the iMac statement. No iMacs until the new one in September. This has got people talking because it is unusual behaviour from Apple. They have never done this before. The fact that Steve used 'iMac' and 'Paris' as the search criteria when he demoed Spotlight has added to the speculation about the new iMac. This delay does say one thing, almost for certain: The new iMac is a G5. There's no reason a G4 based model would be delayed like this.
Whether its a 970 based machine, or an FX, is more difficult to say. But it does tell us one more thing, and that's why I'm bringing it up in a topic about PowerMacs. The production of G5s is going up. By September, Apple hope to have stuck a few in iMacs, and stockpiled alot more ready to go the same way.

If production has increased, then the yield of faster chips will increase along with it. Which means along with all those extra 1.8 and 2GHz chips to go in PowerMacs and Xserves, there are going to be a few 2.2s, maybe even some 2.4s. Factor in the watercooling, and these go up to nearly 3GHz.

I have said before that the 'new' dual 1.8G5 looks like a stopgap model. So when the new iMac comes, it may not be alone....

[Krypton]

You may be speculating a little too much

The 2.5Ghz chip represents something like a 25% performance increase, which isn't bad. If you read ArsTechnica you'd have seen that the problems with scalability regarding the 970Fx are due to it being designed and manufactured very quickly for a processor.

They are working hard on this, I am sure.

As to the iMac, a whole new redesign could explain the delay - the processor isn't really a valid argument, as they would have plenty of chips at the bottom end of the yield for iMacs (1.5-1.6Ghz).

[Maflynn]

Originally posted by Waragainstsleep:
If these figures, which I freely admit are no more than partially educated guess, are anywhere near the mark, then the actual speed increase in the FX since January is in the region of zero. I find it difficult to believe that IBM couldn't squeeze anything more out of their flagship CPU in six whole months.

I don't know what kind of math your using, because I'm totally confused. In May the fastest G5 was a 2.0 GHz. In June it the 2.5 was released/unveiled. That's a 25% speed increase. I don't understand a 2.0 -> 2.5 equates to "the actual speed increase in the FX since January is in the region of zero." Is this like the new math

I have said before that the 'new' dual 1.8G5 looks like a stopgap model. So when the new iMac comes, it may not be alone....

Maybe/maybe not but we are not discussing 1.8 G5's, so why bring it up.

Mike

[Leonard]

It sounds to me that he saying that the Dual 2.5 G5 PowerMacs are using overclocked 2.0 G5s which I don't think they are. The water-cooling system isn't required because Apple is over-clocking, it's required because of the density of the chip and the fact that the heat is now concetrated in a smaller area, as well as to keep the noise down. The G5 Xserve is a noisy machine and so would the G5 PowerMac if you didn't have the watercooling.

[RooneyX]

On Tom's Hardware they have tested water cooled PIVs at over 4Ghz running stable. For even more bizarre experiments, a liquid nitrogen chilled PIV running stable at over 5.2 Ghz!!!!! But it could only be done once because the very crude and large cooling system they built spills nitrogen all over the place. It was just an experiment.

[olePigeon]

Originally posted by RooneyX:
On Tom's Hardware they have tested water cooled PIVs at over 4Ghz running stable. For even more bizarre experiments, a liquid nitrogen chilled PIV running stable at over 5.2 Ghz!!!!! But it could only be done once because the very crude and large cooling system they built spills nitrogen all over the place. It was just an experiment.

Old news. There's a group (webpage someplace out there) that submerged a Mac Classic and overclocked it to something like 500MHz, that's up from 8MHz.

[Catfish_Man]

Originally posted by olePigeon:
Old news. There's a group (webpage someplace out there) that submerged a Mac Classic and overclocked it to something like 500MHz, that's up from 8MHz.

I don't remember exactly when liquid nitrogen cooled Crays were running at 400MHz, but it made my jaw drop. You can do amazing things with enough cooling.