|[Home] [Credit Search] [Category Browser] [Staff Roll Call]||The LINUX.COM Article Archive|
|Originally Published: Thursday, 2 November 2000||Author: Brian Richardson|
|Published to: enhance_articles_hardware/Hardware Articles||Page: 1/1 - [Std View]|
Transmeta: Revisiting Predictions
So, IBM dropped the Transmeta ball. What does it mean for the future of this secretive company? Brian Richardson takes a look back, and then a look forward, at Transmeta and the future of Crusoe.
Blast From The Past
On September 14, I described the "fabless" design philosophy implemented by Transmeta, and how it might affect their future as a processor manufacturer. I also tried to shed some light on the cancellation of their licensing deals with IBM and Toshiba, and how that action might undermine the Crusoe product.
The three possible futures for Transmeta were classified simply as "Good", "Bad" and "Ugly".
Good: Pre-product buzz and smart design by computer manufacturers will prove the success of Transmeta's code morphing CPU.
Bad: Dissolving agreements with IBM and Toshiba will cause Transmeta to lose quality control of its "fabless" manufacturing process, leading to poor quality silicon.
Ugly: Transmeta's technology will not deliver the advertised results. Hardware vendors will implement the technology poorly, generating bad product reviews. Dissolving the IBM and Toshiba deals will cause revenue problems and generate doubt in the minds of investors.
It may be a bit early to see if Transmeta has won the race, but we can at least check to see if they are still on track.
Only one solid Crusoe-based product, the Sony PictureBook, has arrived on the market. The Sony product is a traditional notebook computer ... small and light, looking like a full-size portable that ran the wash cycle one too many times. Magazine reviews report a good battery life (over 2.5 hours on a small battery), but the product hype disappointed reviewers who thought that it should run even longer.
The real problems come in performance reviews, which rate the 700 MHz processor at the speed of an Intel Pentium III 500 MHz. Traditional benchmarks don't take Transmeta's "code morphing" algorithms into account. These constantly adapt the Crusoe's microcode. By tweaking the Crusoe to work best with the program currently being run (making commonly run instructions faster than ones that haven't been used in a while), performance improves over a short time. The typical "single pass" benchmark won't profit from this new technology, making it hard to nail down the Crusoe's true capabilities. Of course, the fact a magazine reviewer may only spend a day or so with the product might hurt too, since Transmeta's technology is aimed at a long-term use scenario.
While Transmeta's technology can improve the battery life of a "standard" notebook, it will not change it overnight. Notebook computers suffer from power-hungry displays, hard disk drives, sound cards, and PCMCIA devices. Just changing the processor doesn't cut it. The best utilization of Transmeta's Crusoe lies in products that haven't hit the market ... mostly "Webpads" that utilize flash memory and smaller displays. Once Transmeta moves into this arena, the true benefits of Crusoe may be revealed.
Since the early days of Transmeta, IBM has been looking at making a Crusoe product. Well, until yesterday, when they publicly announced the ThinkPad 240 was put "on hold" (right ... like my job at AT&T GIS in 1995, when they put the factory in Liberty, S.C. "on hold"). This could be for any number of reasons, but there are two obvious ones:
(1) IBM read the Sony notebook reviews and decided to hold their product.
(2) IBM isn't Transmeta's good pal anymore, since their licensing deal was canceled.
Whatever the real reason, this maneuver is seen by most as a technological "bitch slap" by IBM. Announcing this type of thing weeks before an IPO can't be good, even if the product's "issues" have nothing to do with Transmeta's technology.
Inquiring Penguins Want To Know
There is one thing that has annoyed me, which may or may not have anything to do with Transmeta's future success. No Linux-based Transmeta products have yet appeared. We know S3 has a Webpad in development, and that AOL and Gateway are releasing their Transmeta/Linux "counter-top appliance" one of these days. We don't know when these devices will surface, what types of performance levels are expected, or what role the Crusoe's VLIW instruction set will take in the Linux world. Will it run an x86 Linux kernel? Will some special "Crusoe Linux" be ported to take advantage of the processor's power? Will Linux.com ever get a Transmeta platform for review (I hope David Ditzel gets the hint)?
Linux is one of those "emerging products," much like the Crusoe itself. Porting Crusoe to more Linux platforms might get the army of Linux-crazed penguins to flock to the stores, snapping up nifty Transmeta gadgets (since the code is free, we spend disposable income on gadgets).
The Jury Is Still Out
It's still too early to declare Transmeta as "Good." "Bad," or "Ugly." I can't say I'm totally right or wrong, but some elements of my forecast came true ... just a little bit from each side. In the end, I hope Transmeta can overcome the ramblings of computer journalists (myself included) to create a truly great product. Of course, I also hope they've learned lessons from those who came before, like Cyrix, to prevent the same old mistakes from kicking a good idea squarely between the legs.