Crusoe chip mystery to be revealed

Can the world's most secretive technology company change the world? The folks at Transmeta think they can.

But they're not talking. At least until Wednesday, when the company will lift its veil of secrecy at an event in Silicon Valley and reveal details behind its mysterious new processor named Crusoe.

The biggest mystery, however, is how Transmeta's new technology will change the landscape of computing -- although it is safe to say Transmeta intends to shake things up.

A not-so-secret message, embedded in the HTML code of the company's Web site, proclaims that "Crusoe will be cool hardware and software for mobile applications. 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, so that you can get the full story and have access to all of the real details as soon as they are available."

A statement from the company is somewhat bolder, asking reporters to "Please join us for the introduction of Transmeta and our new Crusoe processors, the world's first family of software-based smart microprocessors."

Crusoe, therefore, should include a number of whiz-bang technologies. But for all intents and purposes those technologies should exist under the hood and be completely transparent to end users, who should simply see an inexpensive and fast device. That appears to be the goal.

Before Wednesday, the easiest way to figure out what Transmeta is up to is to sift through its technology patents. The patents, including No. 6,011,908, granted Jan. 4, describe a technology that pairs a processor, known as the morph host, and software, called code-morphing software.

However, the processor is different from chips such as Intel's Pentium III in that it will likely have many fewer transistors. Instead, it will be optimised to work with software, including attributes such as large memory buffers, which speed performance of the code-morphing software.

Combined, the two would allow a chip to translate instruction sets not designed for it, such as Intel's x86, into its own native code. Crusoe is expected to be the first implementation of this technology.

Transmeta has been mum on the details of Crusoe, except to say that it is aimed at mobile applications. The company is expected to reveal other details, possibly including physical attributes of the chip, power consumption and cost, on Wednesday.

If Crusoe works as Transmeta's technology patents suggest it might, the devices using it would run a wide range of applications, including Java applications, Windows applications and even those written in PostScript.

This flexibility would grant buyers of Crusoe mobile devices the ability to run Windows or Java applications without any fuss. Device makers and software developers would benefit as well, because they would not need to worry about writing new applications or porting older applications to a new device.

A device maker could use Crusoe in any number of products, because in theory it would work the same in a PC running Windows as it would in a set-top box running a Java-based OS. However, it does not appear likely that Crusoe will be aimed toward desktop or notebook PCs.

Instead, it more likely will be aimed at cellular phones, Internet appliances and possibly handheld devices such as Microsoft's Pocket PC.

Besides a few hints from Transmeta's most famous employee, Linux creator Linus Torvalds, during his Comdex/Fall keynote, there's not much else to go on. But with Torvalds onboard analysts expect the company to be developing Crusoe to work with Linux or possibly delivering the chip with its own version of Linux.

But it won't only work with Linux. Transmeta's morph-host hardware technology is geared toward enabling its accompanying code-morphing software to run applications written for other processors by translating their instruction set, on the fly, into its own set. It does so by essentially extending a set of techniques used by application programmers to port applications from one processor type to another, according to Transmeta patents.

Normally, this translation process would encumber a processor, causing it to run slower than a hardware-only chip, such as a Pentium or a PowerPC. Transmeta, however, has developed a combination of techniques to speed up this operation, in part by using a large translation buffer to store translated instructions so they need not be translated a second time. The chip also employs a technique to alias certain pieces of frequently used data that would normally be stored in memory, in order to allow the data to be retrieved more quickly, its patents state.

This, in other words, will allow the chip to run software written for just about any processor, such as a Pentium or CISC or a PowerPC chip used in Apple Computer's Macintosh computers. PowerPC is a RISC processor.

When it comes to startups bringing out new PC processors, the rate of success is slim to none. The only exception is AMD. While not exactly a newcomer, the company has seen great success recently with its Athlon processor.

If Crusoe is targeted more at Internet appliances or cellular phones, as has been suggested, the impact on its PC processor business should be minimal. But Crusoe could give StrongARM a run for its money.

Intel and AMD, in fact, won't see much threat from Transmeta, analysts said. "The concern that AMD and Intel have right now is growing the market (for PCs)," said Martin Reynolds, a research fellow at Dataquest in San Jose, California. "If Transmeta goes out and makes $1bn (£0.62bn), then maybe Intel will notice."

But that's not to say Transmeta won't find a niche. "They're going off into new and uncharted territory. There's something there," Reynolds said.

That something, he said, is room for a new processor architecture that will work with Internet appliances, handhelds and cellular phones. But in order to be successful, that architecture must have a number of values, including low cost, low power consumption, high performance in multimedia and flexibility.

The marketplace for chips for Internet appliances is expanding from zero as new devices come on the market.

Of late National Semiconductor, whose x86-based Geode chip powers Internet appliances such as Boundless Technologies's iBrow, has seen some design wins from Internet appliances there. Geode will also be in forthcoming versions of NetPliance Inc.'s I-opener and Qubit Technologies's Qubit, both companies have said.

Transmeta could pose a threat to National Semi and also to companies like Rise Technologies, which recently refocused its efforts to develop an x86-based PC processor, called mP6, toward Internet appliances.

"But at the moment, there's no market (for Internet appliances), so there's nothing to lose," Reynolds said. "They've got everything to gain."

What do you think? Tell the Mailroom. And read what others have said.

Take me to the Linux Lounge