A wheel has finally fallen off the Transmeta bandwagon, quite possibly taking some millions of dollars of next week's IPO price with it. IBM has spiked its plans to produce a Crusoe-based ThinkPad 240 this quarter, pleading mysteriously that the beast doesn't fit into its marketing plans.
Did Intel get to IBM? Or should we read "manufacturing" for "marketing"? Probably a little of both. Assuming that production hadn't started already, IBM had left it pretty late to push the button, and the "Q4" shipments would hardly have made it into the channel before Christmas interrupted. And despite having shown one of its famous "technology demonstrations" of a Crusoe 240 at PX Expo, IBM has been seriously outhyped by Sony et al, who have announced Transmeta products (several times apiece) and even started shipping some of them.
So maybe we could read that as "our marketing plans are toast, and we screwed up on manufacturing again."
But if the cold hand of Chipzilla didn't also play a part, it might as well have done. By putting the knife into Transmeta IBM will have boosted relations with Intel massively, and should be able to leverage this into rich, perhaps Dell-like, rewards. Historically IBM has been pretty well down-the-line Intel for ThinkPads, and the cancellation (they say "on hold," but in reality the hour has passed) of the Crusoe project nails that arrangement down more securely.
IBM's planners may also have been influenced by performance-based FUD. Crusoe doesn't perform well in traditional single-pass benchmarks, but as it 'learns' about the software it's running it does a lot better in real life scenarios. As we pointed out last week several largely meaningless benchmarks have already been published, and people who should know better have scampered off with "doubts about Crusoe performance" stories.
Internally IBM may have conducted more appropriate tests and concluded that Crusoe's performance is inadequate, but if the decision really was benchmark-related it's more likely that the marketing people, from the position of ignorance prevalent within their discipline, were worried by the spurious ones rather than impressed by more valid testing processes.
This does however point up a serious marketing vulnerability for Transmeta. Normal benchmarks don't work, but until the company can get benchmarking outfits to devise new systems that do, Crusoe will be a largely unknown quantity, which at best will be though of as probably a bit slower than an Intel equivalent. The Transmeta pitch is that in real life it'll win out, and that battery life advantages will impress users.
But first it has to get to users; the FUD and IBM's defection doesn't help this at all. ®