New Transmeta patent reveals x86-killer design
The product will be cheaper and faster than existing chips, and it'll run anything - honest
Another patent granted to Transmeta seems to have established what the mysterious outfit is up to pretty clearly. Briefly, the company's processor is intended to be faster than anything built using current technology, and to be able to run any of the operating software for any existing processors - faster than the original. This of course sounds like complete hokum, but Transmeta has the patent, US Patent Office number 5,958,061, and the application, which can be read here, explains in some detail how the company proposes to achieve this. Transmeta claims: "The present invention overcomes the problems of the prior art and provides a microprocessor which is faster than microprocessors of the prior art, is capable of running all of the software for all of the operating systems which may be run by a large number of families of prior art microprocessors, yet is less expensive than prior art microprocessors." The mechanisms described in this patent tally closely with those outlined in a previous Transmeta patent (Transmeta reveals radical new chip design). Transmeta won't be using "a microprocessor with more complicated hardware to accelerate its operation," but instead will be combining a hardware processing portion it refers to as a "morph host" and an emulating software portion, "code morphing software." These two, the company claims, will work together as a microprocessor "with more capabilities than any known competitive microprocessor." The relative simplicity of the morph host is significant, because it should mean Transmeta's fab costs will be a lot lower than the rivals it proposes to outpace massively. The morph host "includes hardware enhancements to assist in having state of a target computer immediately at hand when an exception error occurs, while code morphing software is software which translates the instructions of a target program to morph host instructions for the morph host and responds to exceptions and errors by replacing working state with correct target state when necessary so that correct retranslations occur." Losing it again? Yes, so are we. In English, it seems that Transmeta is going for simple hardware that can achieve very high clock speeds early on in the ramp, and using this to more than compensate for any speed degradations caused by using software rather than hardware optimisation. Multiple instruction sets can be built into software (although the company's description talks largely of x86, so it's clear who's being gone after), and Transmeta will also cache translated instructions in a "translation buffer," so the amount of translation needed is minimised. It may not however be the case that Transmeta's use of software will cause degradation, as the software itself includes go-fasters: "Code morphing software may also include various processes for enhancing the speed of processing. Rather than providing hardware to enhance the speed of processing as do all of the very fast prior art microprocessors, the present invention allows a large number of acceleration enhancement techniques to be carried out in selectable stages by the code morphing software. Providing the speed enhancement techniques in the code morphing software allows the morph host to be implemented using much less complicated hardware which is faster and substantially less expensive than the hardware of prior art microprocessors." Transmeta also includes what you might call a teaser: "As a comparison, one embodiment of the present invention designed to run all available X86 applications is implemented by a morph host including approximately one-quarter of the number of gates of the Pentium Pro microprocessor yet runs X86 applications substantially faster than does the Pentium Pro microprocessor or any other known microprocessor capable of processing these applications." Pentium Pro? So what? But remember, the application was filed in July 1996, so Transmeta undoubtedly has some far groovier stuff by now. To be fair though, Transmeta does point out that, although the application refers heavily to x86, the techniques are equally applicable to other platforms. But don't you think that, with the latest information going into the public domain, that Transmeta's secret is more or less blown? Might as well announce it then - over to you, Linus. ®
Sponsored: Fast data protection ROI?