D_Fresh's Journal: More about "crippled' DDR in the new G4s

Well, the G4 DDR debate rages on with the posting of this story on Apple.Slashdot.com. This is a Good Thing, because /. posters are some of the best-informed folks out there, and this comment helped clear up the issue the most for me. He says that although the processors are not able to take advantage of a DDR front-side bus, the rest of the system components - hard drive, graphics, PCI bus, etc. - will see speed increases when accessing system memory, and this in turn will lead to real-world performance increases for things like hard drive memory caching, GPU main memory usage, and so on. Definitely good news, and I look forward to non-CPU tasks being benchmarked on barefeats and other places. (I've emailed Tom's Hardware in the past lambasting them for ignoring the Macintosh hardware in favor of Intel and AMD platforms - it struck me as simply ludicrous that they called themselves a hardware site while neglecting to admit the existence of alternative platforms that are interesting simply because they offer a different design philosophy from the "mainstream" Wintel/AMD system architectures. No reply, of course.)

Everyone is saying that Apple is gearing up for more than just two processors in a system (see this comment by foobar104, who always has interesting insights), which is intriguing to me because it represents such a different direction from the way Intel and AMD seem to be headed. Obviously this isn't coincidental - Apple is trying desperately to reach parity, psychological or otherwise, with the faster clock speeds of the Pentium/Athlon competitors through a combination of multiple procs and an OS that takes full advantage of them - but it *does* raise interesting questions about the capabilities of an Apple system with 4+ processors vs. those of a single proc Wintel box. I'd like to think, knowing what little I do about pipelines and multi-threaded apps, that a multi-proc Mac would get better performance when running several applications at once, even if the applications' processes were simply segregated by processor and not symmetrically divided among all of them. (Pardon me for not speaking the language very well - some of the terms are still new to me.) But then I remember that the system bus, memory bandwidth, and so on are usually shared by the processors, and this can become a bottleneck for performance. Still, it's nice to think of G4 chips becoming cheap enough to pile several onto a single mobo, which in conjunction with OS X's SMP functionality could really speed up power-user tasks like audio or video production. (Beowulf cluster remark summarily deleted - ed.) Maybe Apple is following the NASA doctrine of "smaller, faster, cheaper" when it comes to processors - a concept altogether foreign to they-of-the-single-super-clockspeed-chip at Intel.

Enough rambling - I've got to get up and do something. My wife is working all day at the hospital, poor thing, and I'm jonesing for stuff to keep me occupied. Woe is me. :)