Charles, although this wasn't the main point of your post, I thought I'd address it anyway:
<RISC has some inherent advantages over CISC, as well, even when the latter is implemented via the former. And Sun got on that wagon early enough that its considerable legacy of code has been able to maintain a high degree of continuity (in large measure both binary and source compatible).>
I fail to see what those "inherent advantages" of RISC over CISC are in the real world. Already the UltraSparc II is being trounced in the benchmarks by Xeon. Intel's Xeon platforms are also turning out to be a much better price/performance solution than UltraSparc. In my mind, those "inherent advantages" that you talk about are largely theoretical, which is a far cry from practice.
And speaking of which, UltraSparc is hardly the pinnacle of RISC. I'd consider Compaq/Digital's Alpha to be a much better RISC processor than UltraSparc, and in terms of performance, a much stronger competitor to Intel's Itanium than UltraSparc.
Finally, the "legacy code" argument has been used before to argue against the transition to RISC, but that didn't stop Sun. Now that "legacy code" argument is one of the main FUD arguments against Intel's IA-64 and Itanium. However, the FUDsters ignore the fact that Itanium will still remain backward-compatible with IA-32. And backward-compatibility with IA-32 has been Intel's gold mine for years and years.
Tenchusatsu |
