when Ritsuko times out

… probably went like this, since I’m too poor to actually become an ARS subscriber, I’ll just guess at what he said and be happy.

So fast forward from 1999, ten years ago, and suddenly pretty much every big “RISC” ISA except POWER has gone extinct. Alpha, MIPS, and SPARC is only a matter of time. The remaining ones are POWER (doing very well in consoles) and ARM, MIPS has retreated into the embedded market and China. There are also rumours that POWER7, the strangely enough out of order successor to the in order POWER6, will be the last POWER CPU designed, but let’s not count that for now.

Bla bla bla, reiterating stuff that back in 1999, RISC and CISC had already begun to converge. Honestly, I don’t see what else is left to talk about RISC vs. CISC in this day and age, unless he tries to describe some new CPU architecture as “more RISC” or “more CISC”.

Hmm… he did mention he was gonna talk about SoCs, right? Well then, here goes nothing… originally any transistors saved by adopting a RISC approach would go towards integrating more components on a single chip thereby reducing costs, but hey whodathunkit RISC CPUs got more and more complex themselves (just look at POWER2). Now x86 is also invading that area with Tolapai, a Pentium M system on chip solution, and reportedly nVidia is also raring to jump in on the action. Well so much for every other competitor in the SoC market, it’s hard to pick a winner between ARM and x86.

Traditionally we’ve seen vendors pool together a ton of simple CPU cores usually based on some RISC ISA to make a supercomputer, like SiCortex, and IBM in its supercomputers (ever notice they always use tons of PowerPC cores in those, never POWER? And PPC750/G3-era/variant PowerPC cores at that). Traditionally, the x86 supercomputer vendors would be touting their multisocket Opteron/Xeon installations. But today, what do we have in the server room? What, a cluster of Atoms? Does this bode a trend? Will they ever move up into supercomputing territory? Wait… haven’t they? SGI just announced their Octane III (the case design is an insult to the Octane’s name, but at least it’s back) and you can have it with quite a few Atoms in it! And they’re even marketing it as a supercomputer!

And that’s about all there is to talk about RISC vs. CISC in 2009, I suppose. Didn’t Jon Stokes himself emphasize that the comparison was meaningless back in 1999?

Damn I really need to get in on that subscribers-only article.