- Transmeta's Crusoe promises a lengthened notebook battery life, but now that the much-hyped chip is being cut short by IBM, some are questioning Transmeta's staying power.
IBM's decision to cancel a proposed Transmeta-based Thinkpad comes on the heels of laggard processor performance results reported by PCWorld.com two weeks ago. Later reports by other media outlets have offered similar findings.
IBM confirmed Wednesday that the company has dropped its plans to ship a ThinkPad with Transmeta's Crusoe processor in the fourth quarter of 2000. Transmeta officials declined to comment on the news due to the company's impending initial public offering.
IBM's decision to skip Transmeta at this time may bode ill for the upstart chipmaker, or it may just be a bump in the road toward success. It depends on which expert you ask.
Is IBM's Dismissal Not a Worry?
Transmeta may be displeased with IBM's decision, but it doesn't drastically reduce the company's chance at success, says Mike Feibus, principal analyst at Mercury Research.
"It's less of a blow than some people are making it out to be," Feibus says. The hype surrounding Transmeta and its code-morphing processors pushes everything out of proportion, he says. Transmeta's early notebook design wins from companies such as Sony and Hitachi weren't as important as they were made out to be either, he adds.
Sony shipped its Crusoe-powered PictureBook in October, and both Hitachi and NEC have announced plans to use Crusoe in notebooks that will ship first in Japan.
Most vendors are still experimenting with Crusoe, and none has committed to using it in major notebook lines yet, Feibus says.
"It's still untested and unproven," he says, and these companies are simply taking baby steps with a new technology. "The fact that IBM's baby step isn't coming out this month is not a big hit [to Transmeta]."
Intel's Revving its Engines
On the other hand, IBM's decision may signify a very big problem for Transmeta, says Kevin Krewell, senior editor of Microprocessor Report.
IBM's decision seriously hurts Transmeta because the notebook giant's Crusoe endorsement lent credibility to the new processor, he says. IBM's decision to forego a Transmeta product at this time shows the company is more comfortable with other options, he says.
Those options are likely supplied by chip giant Intel.
IBM apparently delayed long enough to see Intel unveil processors designed for longer battery life and strong benchmarking performance on a standard platform that won't require reworking the system, Krewell says. Intel previewed an ultralow-power processor in October.
Intel moved remarkably fast to answer Transmeta's threat, Krewell says. Before Crusoe showed up on the radar, Intel was focusing on faster mobile processors, just as in desktops.
Crusoe Stokes Road Warriors' Hopes
Clearly, Crusoe has already made a splash by putting a spotlight on low-power processors that keep notebooks humming on cross-country flights.
The hype around Crusoe's introduction and first appearances in notebooks shows just how many people are on the lookout for long-lived mobile systems, analysts say.
In fact, Intel was probably already working on low-power processors, but fast-tracked those chips when Transmeta surfaced, says Krewell of Microprocessor Report. He expects Intel's new very low-power processors to offer excellent Pentium III performance while consuming power at levels comparable to Crusoe.
While the Crusoe does seem to offer some battery life savings, performance has thus far proven to be its weakness. Early PCWorld.com benchmarks of the first Crusoe-based notebook, the Sony PictureBook, showed an improvement in battery life, but at a hefty cost in performance.
Quibbling Over Benchmarks
Transmeta supporters argue that benchmarks don't tell Crusoe's whole story because the chip actually runs applications faster after repeated use. This capability has yet to be fully tested or confirmed by independent sources, however.
The first time the Crusoe receives instructions from an application, it must engage a two-step process, Krewell says. First it must translate the x86 instructions into the instructions Crusoe understands, and then it must execute those instructions. That's why initial performance scores may significantly lag other processors, he says.
But the chip saves those translated instructions in an on-chip cache or in a system's main memory for use later, he says. So, in theory, the next time those same application instructions hit the chip, Crusoe checks the cache instead of translating the instructions again, which speeds up the process. (PCWorld.com's testing labs are examining new ways to benchmark Crusoe-based notebooks to test this premise.)
Watch for Round Two
Mercury's Feibus says he isn't surprised Crusoe scored poorly on benchmark tests. When Transmeta unveiled the Crusoe in January, a company spokesperson stated that benchmarks would not be a good measure of Crusoe's performance. "I've heard that before," he says, referring to other processor companies who have similarly discounted benchmarks in advance.
Despite the early poor performance scores and the blow from IBM, Feibus says Transmeta is far from beaten. It will simply take time for Transmeta to break into the notebook market, he says.
It's a notoriously hard market to crack, he says. For example, it took ages for ATI to break into notebook graphics, even though it offered a better product than the entrenched competition, he says. Transmeta may have to follow the same path. "Making your way into the notebook market is a long and winding road," he says.