Mac users and hardcore (well, at least firm) overclockers will now have something in common.
The next generation of top-end G5 Powermacs will come with a water-cooling system standard.
No, Apple doesn’t say “water” specifically, but it looks like the technology described here is what Apple is likely using.
Note the companies that have been working with this new firm, and note the price tag. Expect to see something like this in a PC near you sooner rather than later.
Act of Faith
I’ll leave the technical details for Joe to discuss another time, but the bottleneck for mainstream watercooling has never really been technical, but psychological.
More specifically, hydrophobia. Well, maybe that’s a bit unfair, because unlike the fear of water that comes as an extra feature with rabies, those with computer hydrophobia really fear leaks of water onto components that generally don’t play well with wet.
If computer hydrophobia affects most nuts-and-bolts guys, just imagine how Mac users are likely to feel about it.
This is supposed to be a completely maintenance-free, lasts forever kind of deal.
It will be very interesting to see how Apple markets this to Mac fans, and how they’ll respond. Will Apple bring this to their attention, or not? Will Mac fans have greater faith in Apple than they fear water?
This could be the test run for bringing this type of watercooling to OEM PCs.
Why Is This Now Necessary?
The new top-end G5 packs dual 2.5GHz 90nm IBM SOI chips.
Some of you may recall that the head of Apple promised 3GHz 90nm IBM SOI chips around this time.
Per Tom Boger, Apple’s Director of Power Mac Product Marketing, it’s not going to happen:
“It’s actually quite simple,” said Boger. “When we made that prediction, we just didn’t realize the challenges moving to 90 nanometer would present. It turned out to be a much bigger challenge than anyone expected.”
“All-in-all, no, we are not getting to 3GHz anytime soon (our emphasis) . . . . ”
Hmmm, the company that AMD paid $46 million to help them build 90nm SOI chips can only get out 2.5GHz chips, they can’t do 3GHz anytime soon, and even the 2.5GHz need water-cooling.
I grant you that the design and environmental constraints in a G5 (i.e., acceptable heat/noise levels from CPU cooling) are tougher than they’ll be for an AMD platform, but it makes you wonder about what kind of problems AMD may be having making 90nm SOI that are measurably faster than that.
The issue here is not SOI or IBM or AMD. The issue is 90nm process technology. From everything we’ve heard about it, the old rules and givens break at 90nm.
To quote some recent comments by IBM’s chief technologist:
“Basic technology simply is not driving performance at the rate it did previously. . . .
[T]he rate of performance enhancement is becoming impacted. We simply made transistors so small that as we continue to attempt to make them smaller, it requires a huge amount of innovation to get added performance.(our emphasis) If you were getting an improvement in technology of perhaps 30 percent a year, and suddenly that rate drops to half, the rate at which your systems improve would also drop by that amount.”
Power consumption is becoming a critical factor, limiting performance. . . . It is really an extraordinary event. I am hoping that people really understand the sort of discontinuity we are talking about here.”
So if people tell you, “What’s the big deal, AMD can get 2.5GHz out of a Hammer now?” they don’t get it.
Again, this is not AMD-bashing. Everybody who has gone to 90nm has hit this wall. This is an industry problem, not a company problem. The quick-and-easy-and-cheap days are over, at best for a few years, perhaps forever for the typical PC.
So don’t be surprised if AMD ends up having real problems getting fast (i.e. 2.6GHz or above) 90nm chips out. I don’t think they’ll have any problem getting slower ones out to replace current Hammers, but I think heatsink-and-fan people are going to be in big trouble cranking up these things to expectations.
Let me put it this way. I’m beginning to think that if I’m going to be doing any serious overclocking six months from now, for anybody’s processor; I may have to get myself some phase-lock refrigeration, simply because air won’t cut it, and water won’t be a lot better.
Don’t jump to conclusions, I’m just starting to think that way. Wouldn’t actually do it for another six months or so, seeing first what really happens, and hoping it doesn’t turn out as bad as I fear it will.
But things could get that ugly, which means the typical heatsink-and-fan overclocker will either have to lower expectations, upgrade the effort, or find something else to do.