I haven't posted in a while, nor have I read all of this thread, but the first few pages were more than enough to give me material to slam down a couple of lines with.
First of all, the 970FX is already being fabbed at 90nm - a previous poster seemed to think otherwise - and that doesn't exactly help the heat dissipation characteristics. You see, when a chip is shrunk in overall surface area, you're trying to edge out the excess thermal energy from a smaller space. True, the overall amount is lower, but that doesn't mean anything if there's a worse ratio of heat to surface. I've done the math before, but the 7447A (the G4 that's currently in PowerBooks) is a far more efficient design than the current 970 line. When you add to that the fact that Freescale's MPC7448 chip is going to drop roughly a third to a half of the existing heat (from 12-16 watts to a mere 10) while clocking up to 1.7-1.8ghz, raising the bus to 200mhz, and working with existing pin-outs... Is it really that hard to realize why the 970 is looking less and less likely?
I would be deeply surprised to see the 970 in any incarnation of the PowerBook before WWDC, and that's an extreme outside chance. For that matter, I would be disappointed to see the 970 in a PowerBook at alll. There are better options out there and people are seemingly forgetting that the big push this year is for dual-core designs that will be pushing multithreading and simultaneous multitasking in the mobile sphere. Let's not forget that the Dothan Pentium-M's replacement is due to show up as a dual-core under the Merom codename sometime this year. It's not for nothing that Intel trashed Tejas and the Prescott 4ghz+ designs and it looks as if all of the big players are looking towards paralellization as the new savior. Even IBM has plans for the 970MP sometime in the near future, though there's no chance in hell that will be a mobile processor.
Our best hope is still Freescalle and the MPC8641, whether in dual or single core, due to the SoC integration of many of the components and the massive efficiency when compared to the previous 'Books. Even with a single core, you'd be running a 1.8-2.0ghz core with a 1MB L2 cache, on-die 667mhz DDR2 controller (faster and cooler than PC3200 SO-DIMMs, incidentally), integrated SATA, SATA2, and PCI-E 24 lane controllers on the chip, four gigabit MAC controllers with hardware encryption, and a pair of 128-bit dual precision vector units that make IBM's implementation of AltiVec look like a pale, weak joke. The dual-core package, at 1.8ghz, runs a relatively svelte 22 watts (the same as a previous generation TiBook or a Pentium M 1.6ghz), and has advanced cache integrity features that let the processors share their prefetching at the L1 point.
So, why should we want the G5, again? Because the number is higher?
Feh.
