Whoops! If, because of yesterday's spiel on QuickTransit, you're imagining the Windows version of Doom III running at blazing speeds on a mid-range Power Mac G5and a Golden Age of x86 Emulation free from Redmond's evil clutches, you may want to take it down a notch. Faithful viewer Colin Cornaby noticed something on Transitive's technology overview page that we didn't; it notes that QuickTransit maps between "similar operating system calls" that exist between "any two Unix/Linux-like operating systems." That strongly implies that one of QuickTransit's four key components-- and much of the software's vaunted speed-- goes out the window if QuickTransit is used to translate from one operating system to a dissimilar one, such as Windows to Mac OS X; no wonder they demo the Linux version of Quake 3 running on a PowerBook and not the Windows one. We imagine it'd be possible for Transitive to build a Windows-to-Unix/Linux operating system mapper to regain some of that performance, but that's obviously further down the line, assuming it's on the agenda at all.

That's not to say that it isn't still some promising software; the operating system mapper is just one of QuickTransit's nifty tricks, and we may still see Windows emulation at least partially based on its technology sometime in the future. But if it happens, it's not going to be for a while, and when running Windows apps the speed might not be that much better than Virtual PC's after all. But hey, at least we wouldn't be completely at Microsoft's mercy to get viable x86 emulation on our Macs, and if nothing else QuickTransit looks like it'd still do graphics subsystem mapping for native 3D performance, and even Virtual PC still won't do that.

That's right, we said it still won't-- at least, not in the long-awaited version 7 that's slated to ship next month. Oh, sure, we know that was one of the most eagerly-awaited features planned for Virtual PC 7, since it'd make at least some non-ancient Windows games adequately playable on fast Macs. But AppleInsider is reporting that Microsoft has chucked a "significant number" of planned features from version 7's spec sheet in order to hit its (already really late) October ship date. So wave buh-bye to native graphics card support and 3D acceleration; say sayonara to punching through the 512 MB RAM ceiling; bid adieu to RAM disk support; spit out a big, fat paalam na po to better multiprocessing usage. Shades of Longhorn, anybody?

The upshot, then, is that the one planned feature that Virtual PC 7 does retain-- its long overdue compatibility with the G5 processor-- might not be enough to deliver the blazing speed users have come to expect from this upgrade. Sure, if you have a G5 it'll be infinitely faster than not running at all, but most of the features that Microsoft cut stood a chance to deliver significant performance improvements. Don't like it? No problem, just go buy the competition's product instea--

Oh. Right.

Well, don't fret; AI says that "most, if not all of the features cut from the development of Virtual PC 7.0" will simply migrate to future upgrades. Sure, you've been waiting for over a year already, but c'mon, are you really going to need those features before "next summer" anyway? Besides, now you can spend all that time weaving your complex conspiracy theories about how Microsoft is "accidentally" keeping Windows emulation hobbled on the Mac platform just so the buying public doesn't see a Mac with Virtual PC to be a workable substitute for a real Windows system. Memo to the Transitive folks: if you are planning to bring QuickTransit to the Mac platform as a Windows emulation solution, you may want to beef up security around the office and hire a team of food-tasters. You know, just to be... cautious.

1,100 dual-processor Xserve G5s wired together into a crushing onslaught of computational brute force? Ah, that's nothing. Virginia Tech's System X cluster may have quickened our pulse back when it was the world's largest pile o' modern Macdom cobbled together into a single hive mind, but ever since we heard about the Army's new project last June, our techno-lust for raw power has been directed elsewhere. Remember? A military contractor by the name of COLSA Corporation planned to roll its own Virginia Tech-style Xserve cluster-- only with 1,562 Xserves this time around. How can we resist? Sure, the Virginia Tech folks have infinitely more style than any Army guys we've ever met, but c'mon... 462 extra Xserves? We're fickle supercluster groupies, man. We go where the teraflops are.

And clearly, Here There Be Teraflops. Faithful viewer klownhaus tipped us off to a fairly extensive (yet fluffy) three-page profile of the COLSA cluster, which makes a big thing about how the $5.8 million project-- named, as you may recall, the MACH5-- has a theoretical peak performance of 25 teraflops, compared to 40 for the $350 million Earth Simulator in Japan. (Of course, theoretical performance doesn't mean bupkis, but reasonable estimates place the MACH5's real-world performance to be roughly a third of that of the Earth Simulator... at a sixtieth of the cost.) Apple claims that's fast enough that the MACH5 could calculate in one measly second what it'd take a human being (a non-dying human being, but a human being nonetheless) with a calculator two million years to accomplish, and that's without breaks for eating, sleeping, or Laverne and Shirley reruns on the Lifetime channel.

Well, okay, yeah-- but can the MACH5 do that thing where you make it look like you're pulling the tip of your thumb off and then putting it back on again like nothing even happened? Ah-HA! See? Humanity wins again. Give yourselves all a pat on the back.

What's more, the MACH5 isn't even fully operational yet, so Immortal Calculator Guy would get a nice head start; the last we'd heard was that the system might be up and running by "late fall," hopefully in time to rank somewhere on November's list of the world's 500 fastest supercomputers. Based on our estimate of actual performance (15ish teraflops, give or take), the MACH5 would probably have slotted into the current list at number 3, so we're hoping that it'll stick a top-five position in the rankings this fall. See, photos of the five fastest are posted to the TOP500 home page, and since the chances of System X getting back up there are slim to none (it'd be ranked fifth in today's list, and newer and faster clusters are being built all the time), it'd be nice to have a snapshot of a Mac-based supercomputer up there, basking in the glory.

Anyway, if you're a Teraflop Slut like we are, check out Apple's pages on the MACH5; amid the technical details, the speed comparisons to some old guy with a Casio, and the well-deserved and utterly predictable self-congratulatory tone, you'll find a few requisite photos of racks upon racks packed full of Xserve-y goodness, and one small shot of a semi truck delivering 300 Xserves at a time. Wow. Is it... is it hot in here, or is it just us?

And the online digital music market continues to churn like the stomach of a man who just invented the Beer, Twinkies, and Soft Taco Smoothie. The Apple-vs.-Apple speculation continues to run rampant, with faithful viewer mrmgraphics noting that all this talk about a monster-sized settlement has some people figuring that the Beatles will become "major shareholders" in Apple and that Paul McCartney might even wind up on Apple's board of directors. Meanwhile, RealNetworks's Rob "I Swear I Am A Man of the People" Glaser is spouting off to nerds about freedom of choice over at Slashdot; feel free to wander in there if you don't mind getting lost for a couple of days. And as usual, other companies are still scrambling to line up for an iTMS courtesy kick to the head.

The latest soon-to-be-victim? Just say Yahoo! Faithful viewer Hatchet noted that, according to CNET, the Internet portal thingy is about to shell out $160 million in cash to buy MusicMatch, one of those services that keeps falling under our radar, partially because it's not Mac-compatible (duh), but mostly because we hear about it so rarely, we can only assume that it's not exactly kicking kiester out there. Also, while it does sell individual songs and albums, MusicMatch seems to put a real emphasis on its subscription service, which is another reason its name hasn't been as visible in stories about the Download Wars.

Anyway, Yahoo! says that it's "committed to being a major player in digital music," hence the MusicMatch merger. The company had already bought Launch a few years back, but that's apparently a streaming-only service, and Yahoo! is all gung-ho and foamy to get into the whole download thing, too, so that it can mix it up with the big dogs in what the BBC actually referred to as-- get this-- "the highly lucrative digital music market." (When faithful viewer iain sent us that we just about giggled ourselves into an aneurysm.) If Yahoo! wants MusicMatch to compete, though, it may need to rely on more than its sizeable audience of eyeballs-- er, eardrums-- to make it happen; MusicMatch's songs are 99 cents each on top of a $10 monthly subscription fee, making it one of the most expensive options floating around out there right now. But we're sure they'll get it all hammered out in the coming months.

Of course, speaking of "hammered," you really have to wonder what sort of mentality it takes to invest $160 million to buy your way into a market that seems to be break-even at best, and in which the competition is all packing switchblades and giving each other the hairy eyeball. This is a rough crowd, and the weak get weeded out pretty quickly. (Remember BuyMusic.com? No, neither do we.) So we have to wonder what Yahoo!'s been drinking to want in to this particular party. Buy hey, you know what? The more the merrier. We're sure Apple won't mind putting its boot upside one more cranium just for the sake of community. For it's own sake, here's hoping that Yahoo! doesn't sober up first.