Crusoe: Less power is more power

Till January 19, 2000, little was known about the highly secretive,

well-funded Santa Clara chip start-up that went by the name of Transmeta. The

only definitive information that were made available were the two patents it

filed last year and its staff, which includes the high-profile Linus Torvalds

and AT&T Bell Labs chip architect David Ditzel. Even the source for the

Transmeta Web site piqued industry curiosity with the message "Transmeta’s

policy has been to remain silent about its plans until it had something to

demonstrate to the world." It went on to read "Crusoe will be

unconventional, which is why we wanted to let you know in advance to come look

at the entire Web site in January…"

Crusoe debuted amidst great fanfare, being billed as the first smart,

low-power software-upgradeable microprocessor. Presented as a x86-compatible

family of solutions, it claimed to combine strong performance with remarkably

low power consumption. For a large part, the patents had already revealed this.

What many industry analysts did not expect was this chip attacked Intel and

every other chipmaker make on a new level — it replaced large numbers of

transistors with software.

While designing Crusoe, Transmeta engineers went back to the drawing board

and literally rethought the fundamentals of processor design. Instead of a

primarily hardware implementation, Crusoe consists of two components — a

hardware engine logically surrounded by a software layer. The engine is nothing

but a very long instruction word (VLIW) CPU. The surrounding software layer

fools x86 programs into believing that they are running on x86 hardware.

Crusoe processors have a proprietary native language, which is a 128-bit VLIW

(very large instruction word) instruction set. This works like RISC, only much

faster. What Crusoe also does is that it employs a technology at the software

level called Code Morphing. Put simply, Crusoe software not only morphs Intel

specific code into its own native language on the fly, but it also caches the

translated instructions, doing away with the need to retranslate them later.

What you have then is the advantage of being able to write software that would

enable a chip to effectively be any chip (a chip-upgrade is now a software, and

not hardware), and at the same time, replace the millions of energy-sucking

transistors other chips use.

Crusoe’s software has the smartness to judge exactly how much energy an

application needs. Correspondingly, it alters and adjusts both the clock

frequency of the chip and the voltage requirements. It’s much like the Turbo

button on older PCs (which regulated voltage to the chip to slow it down so that

older programs such as games would run at a usable speed), only that Crusoe's

software raises and lowers the voltage and the clock frequency automatically.

Something like chucking your spanking P-III out and substituting it with a P100

to save energy when you’re playing Minesweeper, then boosting it back up when

Outlook 2000 demands it. A degree in chip engineering isn’t required to

imagine the power savings a laptop could enjoy with this technique. And, for

those who do hold this degree, it results in cubic reductions in power

consumption (up to 27 per cent), as opposed to other chips, which adjust power

linearly.

Bottomline



At present, Crusoe is expected to be seen in a host of mobile appliances such as
handheld devices for connecting to the Internet (3120) and lightweight laptops

(5400). S3 has even planned the usage of Crusoe in its range of Web Pads.

Although the model TM3120 (with a specially designed version of Linux to take

advantage of the changes in chip design) and model TM5400 are impressive first

efforts, the significance, and more important, market acceptance, is going to be

more apparent over the next several years. The question will remain — like the

legendary fictional character after whom it is named — will the Crusoe sink or

learn to adapt to a hostile environment? Only time will tell.