jakemikey said:
But that's really beside the point. There seems to be (at least on the WWDC specs thread) a general prejudice against the 970 and its performance. It's referred to often as a "holdover" product which offers lackluster performance, and is there only as a spacefiller. I have to wonder how many of these people use a G5 on a regular basis. Performance, even on a stock G5, is anything but lackluster.
There's a reason for that, you know. The PowerPC 970 isn't all that much greater than the G4, especially when the massive changes in architecture are taken into account. Despite the five to sixfold increase in FSB, despite the full use of DDR, and despite clock advantages ranging from a fifth or more, the performance increase just isn't as great as the numbers would lead you to believe on many tasks, especially in the lower-clocked units. A 1.6ghz 130nm 970 barely beats out a 1.5ghz MPC7447A at Final Cut, for example, and it has the massive advantages listed above.
Why wait a year? Why the amateurish feel of a less-efficient AltiVec implementation on the G5?
Might it be because, as we guess over on the WWDC thread, you've got something even bigger in your pocket but it needs more time to be brought to market? Just a thought.
Other people say that if Apple releases anything but a 975 (a product which, even if mentioned once on an Asian IBM server, has had NO SPECS released, leaked, or otherwise made known by a reliable source), it will spell their doom. Come on people! Not one of us would really be able to tell the difference in daily routines between a 3GHz "975" and a 3 GHz 970FX.
The Power4 is being outperformed by something between a factor of three and four when it's put against its new big brother, the Power5. Are you telling me that you wouldn't notice a computer being four times faster? I probably would, though it might be a little less obvious if you didn't have the machines side by side. However, if you were to benchmark a 975 and it did four times better than clock extrapolation of the 970 would indicate, that would be pretty clearly backing up the whole point.
Also, I think you're overstating the general consensus of the 975 cheerleaders. We want it, we think it's likely to be better than the 970 in most ways, but there's not a one of us you'd find who think Apple will die without the chip. There is room to argue that there would be some serious consequences from not meeting the 3.0ghz promise, and from not moving on with technology as its adopted (Intel's already using SSOI, and AMD is about to use SSOI and Black Diamond low-k).
As to whether the FX could make it to 3 GHz, maybe not, but keep in mind that the original 970 was said to reach 1.4 or something before it showed up at 2 GHz in the G5. Right now on IBM's website, the 970FX says 2 GHz +.
It's conceivable that the 970FX will reach 3.0 ghz and beyond, but I doubt that's the chip that we'll see do it any time soon. Part of the problem is just the core that was used - the Power4 - is not intended for that kind of clock, even with the gate oxides thinned out. The parent chip in IBM Big Iron is running at 1.3ghz, after all.
There's a budget to be played with at this point, provided that the SSDOI process in the 970s has been resolved and is no longer a problem, and some room to play with the clock. The question is still whether or not it will allow a jump in the realm of 1ghz without soaring back over 51 watts, as the Prescott 90nm parts exploded when they were tried.
I highly doubt that if IBM had a chip (in high volume) as impressive as the 975 is made out to be, that they'd keep it so tightly locked down, just on Apple's account.
They would if Apple spent $1 billion to pay for the R&D for the chip, since that would come out to a little more than a tenth of IBM's revenue for last year. Apple's sitting on a huge pile of cash, and that would be one hell of a way to use it...
