Industry sources say they believe the company will unveil an embedded-class very-long-instruction-word (VLIW) processor, called Crusoe, which will be showcased in a handheld Web pad or similar downsized browser equipped with an embedded version of the Linux operating system. The Web pad will have been designed with an unspecified partner, those sources speculated.

Transmeta has declined to comment, maintaining the same wall of secrecy it erected when it opened its doors in 1995.

"Transmeta is not giving any press interviews in advance," Dave Ditzel, founder and chief executive officer of Transmeta Corp., told EE Times this week. "So anything you see printed before [Wednesday] Jan. 19 is still guesswork."

At Comdex last November, speculation ran rampant that Transmeta was poised to field a downsized microprocessor capable of powering handheld information appliances a la Palm Computing's Pilot.

Linus Torvalds, the creator of Linux and Transmeta's best-known employee, further fueled those flames at Comdex when he stated in a keynote speech that Crusoe would be essentially "software-powered."

Though Crusoe "does have something to do with mobile computing," an info appliance is not necessarily the objective, said Ditzel. "There will definitely be a few surprises that no one has even come close to yet. Check our Web site, because on the 19th, the truth will be out there."

Pentium killer?

Indeed, the buzz at Comdex about an embedded info-appliance chip eclipsed longstanding talk that Transmeta was preparing a cost-busting rival to Intel's high-end 32-bit chips.

Such thinking evolved because of the expertise of the initial Transmeta team. In addition to Ditzel, Transmeta acquired ample hardware expertise when it hired many members of the Texas Instruments team that was disbanded after TI decided not to go ahead with a Pentium-class clone.

But industry sources speculate that Transmeta did build a processor but that chip could not outperform an equivalent Intel Pentium-class offering.

Perhaps that's why Torvalds' high profile had everyone from programmers to pundits thinking that Transmeta is a software-oriented company rather than a fabless chip vendor, and that Torvalds had become a driving force behind the company's activities. Some experts pointed out, however, that Torvalds did not join Transmeta until well after it was formed.

Indeed, some believe Transmeta is wisely using Torvalds' Linux connection as a form of "spin control" since there's no reason handheld Web browsers can't be built using other embedded processors or OSes.

In terms of facts, the only thing Transmeta has revealed is a posting on its Web site, which reads in part: "Crusoe will be cool hardware and software for mobile applications. Crusoe will be unconventional."

Patent info

Transmeta has been granted two United States patents in the past year. The more important of the two details a scheme for "code morphing," a technique that could be used to translate X86 instructions into native Crusoe instructions with a VLIW engine Transmeta has designed. Another possibility is that Crusoe will have a hardware abstraction layer (HAL) above a relatively simple VLIW processor. The chip itself will have a small number of functional (i.e., execution) units rendered in hardware. The (software) HAL will convert X86 code using its patented morphing technique, pack it into wide instruction words and dole it out to the chip.

Such an approach would fall within the definition of a "software-controlled" chip, another tenet Transmeta has set forth for Crusoe. More important, it would mean that Transmeta could ultimately field a family of chips ranging from the embedded debut model to heftier spins aimed at servers and such.

Transmeta chief Ditzel is in an ideal position to run a microprocessor company. As an architect, he cut his teeth designing a host of pacesetting microprocessors starting with Bell Labs' Crisp, but he is best known for his work as chief architect of Sun Microsystems' RISC-based Sparc family. As Transmeta CEO, he is considered to be a proponent of VLIW. In a December 1998 keynote speech at the IEEE's Micro-31 conference, Ditzel noted that VLIW has been around for more than a decade and suffered commercial setbacks during its first attempts to find market success in the mid-1980s. "Largely, VLIW has been a technique waiting for technology to catch up," Ditzel said at the time.

See related chart