Found the following at :
More on Apple's upcoming 64-bit white horse from IBM
Posted 9/19/2002 - 12:55PM, by
eWeek is carrying a few more details on the upcoming 64-bit desktop Power4 derivative chips from IBM that caused so much enthusiasm in the Mac community a few weeks back. If you followed the analysis of the original information that was done here and here on Ars, then very little in this article will be news to you. Still, there are some new tidbits, which I'll list in a moment. But first, I want to note that the most important thing this article does is lend credibility to the claim that this chip is indeed intended for use in Apple machines, and that Apple and IBM are working together on it in some capacity. eWeek has been very reliable in their Apple scoops so far, and while they're almost never the first to leak new information (this particular Apple-IBM info showed up on a Mac rumor site earlier this week) the information that they do chose to publish has historically been spot-on. They've sort of become the last and most reliable link in the Mac rumor chain--a Mac story goes from being rumor to an eWeek story to "official."
Now that that's out of the way, let's look at what they've got. First, the name is GPuL, for GigaProcessor Ultralite (ewwwww...). Next up is the revelation that Apple is (thankfully) working on a new bus architecture, called ApplePI, to replace Maxbus for use with the GPuL. No details on what improvements this bus architecture will bring, though, except the obvious: point-to-point topology and higher frequencies. The GPuL chip is likely to debute on a 0.13u SOI process at speeds ranging from 1.4 GHz to 2 GHz, and it will feature four processor cores packed onto a single die for four-way on-chip SMP. A GPuL based system should scale to 8-way SMP using two processor packages. The system will support SIMD using IBM's VMX instruction set, which is another name for what Motorola calls Altivec. Initial benchmarks show excellent performance, and as I noted in my analysis IBM will be building Linux systems around this, possibly in cooperation with Red Hat. And finally, there's a projected release date of late Summer 2003, a date which I've seen bandied about for this chip before.
So there you have it. No really meaty microarchitectural details, but solid confirmation of what we'd already surmised along with some nice new details. Now, what does Moto have up its sleeve with the G5?