The Hammer core running at these lower speeds would have a hard time competing side-by-side, in a purely 32-bit software environment, against chips like the Pentium 4 and AMD's own "Barton" core Athlon XP processors. (Barton, which launches February 10th, is not fabbed using SOI, and should be readily available for sale on the launch date.) Rather than launch the Athlon 64 for the desktop early and see relatively lackluster performance versus current chips (as Intel arguably did with the Pentium 4), AMD is planning to wait until September, by which time the company hopes to have raised yields, and thus clock speeds, for its SOI process.
In this context, AMD's recent announcement of a technology agreement with IBM looks less like a forward-looking deal and more like a move aimed squarely at getting Hammer back on track as soon as possible. UMC wasn't jilted just because Hector likes Big Blue better, it seems. IBM pioneered SOI chip production, and AMD is already learning some tricks from IBM's foundry types. With luck, AMD may be able to avoid major reworkings of Hammer while bringing Athlon 64 up to speed.
That seems to be the plan, anyway. At this point, Hammer has been delayed more often than a French declaration of war. We'll see how it all works outin time.