Was it more a question of over-exuberance on the part of the marketers and the press, or does the potential of the new software-based processor really deserve the hype?
Everything known about the Transmeta microprocessor prior to its launch Wednesday was based on information gleaned from patent applications, seasoned with a fair bit of good old fashioned speculation. However, the company has been assiduous in preparing the ground -- more than 2000 separate non-disclosure agreements with third parties were signed prior to the unveiling. Appetites were further whetted by the knowledge that the creator of Linux, Linus Torvalds, was involved in developing the technology.
According to Transmeta, its prized technology consists of a specially designed CPU working in conjunction with "code morphing" software. Code morphing translates instructions written for one processor to those used by Crusoe, and thus can let it run applications software and operating systems already in existence. The emphasis, unsurprisingly, is on Intel-based programmes such as Windows and its applications software. This isn't the first time for such techniques -- under the general name of emulation, they've been used for at least thirty years -- but Transmeta claims to have improved the state of the art. The patents reveal many key innovations in software design, and the chip itself has been designed to make the translation process particularly efficient.
One of the benefits is in power consumption: one of the original design goals for Crusoe was to use as little power as possible. Dynamic speed and voltage control allow the chip to maintain application performance while running longer at a cooler temperature. These factors are vital in mobile computing, as they translate directly into better battery life. While an Intel processor typically uses between 4 and 10 watts -- and thus also needs a power-hungry fan to get rid of the heat -- the Crusoe chips use an average of just 1 watt and doesn't require active cooling.
One of the downsides of emulation is that the translation process takes processing time, slowing down performance when compared to a chip that's designed to run the software directly. No independent benchmark results have been made available yet, leaving this aspect of Crusoe's claims yet to be verified.
For this and other reasons, some experts have expressed concern that the hype outweighs the real potential. Others believe the Crusoe could herald a revolution in modern computing.
Richard Gordon, senior analyst with Gartner Group research, is one who remains unconvinced. He says, "I think you can safely say that their marketing and PR campaign has been very successful. There's been lots of hype but they've yet to prove the technological innovation."
Gordon also has reservations about the specific technology involved. He continues, "If you read between the lines, it's not quite as quick and as clever as some think it is. When you have the ability to run different OSs it has to be an emulation technique. Then there has to be a performance penalty. From what I understand the headline speed is 700 MHz but emulation reduces this to 500."
According to Gordon, Transmeta could also have difficulties competing with the likes of Intel in a real business environment. "Lots of people have made attempts at challenging the Intel standard and you've seen how successful they've been," he says.
Editor in chief of The Microprocessor Report, Keith Diefendorff, is much far more impressed and even believes that the hype surrounding the Transmeta chip may have been justified. He says, "Actually I think what they've done is fairly impressive. If it lives up to claims, then they've made a serious reduction in power consumption."
Transmeta's biggest accomplishment, according to Diefendorf, is in totally rethinking CPU architecture. He thinks that by doing "an entire engineering job" and designing a completely new CPU Transmeta may not only have achieved excellent power consumption and temperature control but could also have succeeded in releasing the "real potential of software emulation" for the first time.
Joe Bryne at Gartner Group Research in the US points out that by simply streamlining the hardware, Transmeta has improved processor technology. He compares the principle behind the Crusoe to those behind the Palm Pilot. "Look at the Palm," he says, "It is based on very simple hardware and set of features people really need. Competitors [that] have used more complex hardware haven't been as successful. It is a far more useful, more elegant design."
Diefendorff also believes that the technical innovations Transmeta has made are so important that it shouldn't have any difficulties competing with the major microprocessor companies. He adds, " It seems that they've made enough of a difference in power that they can lay down a enough of a reason for anyone to buy their chips."
Who is right? Tell the Mailroom