Originally Published: Thursday, 14 September 2000 Author: Brian Richardson
Published to: enhance_articles_hardware/Hardware Articles Page: 1/1 - [Printable]

Understanding The Fabless Transmeta

Transmeta has had a lot of buzz ... and they've had some recent bad press. Find out what could happen to this hot new company.

   Page 1 of 1  

A recent EETimes story announced that Transmeta had changed its production agreements with Toshiba and IBM. In the beginning, Transmeta had sold production rights to the Crusoe processor to these companies. While Linux fans everywhere are rooting for Linus Torvalds' latest venture, this move begs the question ... is this a good thing for Transmeta?

What Happened

Transmeta's Crusoe processor design is just that ... a design. Every modern processor goes through years of simulations before the earliest silicon is spun. Intel, IBM and AMD maintain very expensive chip fabrication facilities, funded by decades of revenue and investment. A small startup like Transmeta cannot afford to start a fabrication facility, even with Paul Allen's deep pockets.

Like most companies producing custom ASICs, Transmeta uses the "fabless" approach to chip design. The simulated design is given solid form by an outside production house. With NexGen (the first Pentium clone, later sold to AMD to make the K6), the design house produced chips under contract. These chips were sold under the NexGen name.

The original Transmeta deal with IBM & Toshiba is similar to how Cyrix did business. Fabrication rights were sold to IBM, allowing IBM to make their own brand of CPU. This gave Cyrix two sources of revenue: direct marketing of chips made for Cyrix by IBM, and royalties from the sale of IBM-branded processors.

The Good

The problem with the deals between IBM & Toshiba is the cost to Transmeta. If the IBM and Toshiba branded products were a success, Transmeta gets very little outside of name recognition and the hope of future revenue. If IBM and Toshiba make products that fail in the marketplace, Transmeta gets blamed (even if the CPU design isn't to blame).

Being free of the IBM/Toshiba agreements gives Transmeta a better chance to control their own destiny. The "Transmeta buzz" has already landed customers like FIC and Sony. If the relationship with IBM was not harmed by this latest move, Transmeta could make serious money when Crusoe hits the marketplace.

The Bad

Computer history buffs will note that the two examples of fabless production given above (Cyrix & NexGen) were sold to other parties after disappointing financial performance. Most of this can be attributed to the design of the CPU, which is not always the fault of the fabricator. But a third-party chip fab could be problematic for Transmeta.

Fabless production had a few disadvantages. First, the CPU designer can simulate a good CPU that is very hard to mass-produce (Example: Intel's original PentiumPro). Second, the CPU designer doesn't control the fabrication environment. This can lead to those nasty "non-technical problems" that kill a product.

The deals with IBM and Toshiba gave Transmeta a very tight relationship with two major silicon manufacturers. Removing the ability to manufacture the Crusoe processor under the IBM name might affect how IBM handles production of the Transmeta CPU.

The Ugly

The buzz surrounding Transmeta's processor design has been a steadily increasing noise the past few years. But buzz does not necessarily generate revenue ... especially for a company that still has no product in the marketplace. The revenue generated by these deals helped fund Transmeta's initial Crusoe development, and was an indication that the Transmeta product would make a return on investor's money.

It's not a good sign that Transmeta bought rights back from Toshiba just a few days after Toshiba questions Transmeta power claims in a vnunet.net article. While Transmeta has several product design wins in Taiwan and Japan, none of these have gone to market yet. Bad press does have the potential to hurt early sales and the upcoming Transmeta IPO.

So ... What Does It All Mean?

In the end, Transmeta's product will speak for itself. Transmeta could be the best thing to happen to the mobile computing market since the LCD, or it could become the NexGen of the 21st Century. Remember that Crusoe is just a processor: putting a good processor in a bad design still makes a bad product. The ability of the manufacturer to create a stellar product based on Crusoe may be the key to Transmeta's survival.

Brian Richardson will spend today on a 450 MHz computer writing 16-bit code running in real mode, checking to make sure tomorrow's computers are compatible with yesterday's standards. Don't think about it too much ... it's very depressing.

   Page 1 of 1