Return to the archive index

Re: Crusoe versus StrongArm, Super H, etc.

From: bph@primenet.com (Blair P. Houghton)
Date: Sat Feb 5 11:16:24 2000
Newsgroups: comp.sys.wearables

When I read the Transmeta website, I got a strong whiff of the
RISC vs. CISC debate.

The chip appears to be VLIW, meaning "variable length instruction
word," but what that means internally I don't know (yet; I signed
up as a developer, but they pleaded inundation and haven't sent
a byte of data..).

It could be a varaible 8-32 bit word.  But so's the Intel
architecture, when you look inside it.  The new Intel/HP
designs have VLIW up to the multiple-word range.  But maybe
Crusoe can do that, too.

What made me laugh, though, was their claim that they were improving
the microprocessor field by moving operations from hardware to software.

First, RISC used simple instructions and implemented complex
ops by having the compiler unroll them.  And second, all CISC
chips--and some RISC chips--use internal microcode, meaning
they do some ops using firmware control of simpler hardware
capabilities.

About the only thing novel I can see is the claim that Crusoe
can remember certain snippets of software as though they were
microcode.  So you might be able to run an emulator-setup
program on it, then just run your off-the-shelf software as
though you owned the right processor for it.

Which brings up another non-innovation innovation.  DEC did the
chameleon platform thing with the Rainbow computers.  And lost
its ass.

Transmeta is taking a leap into empty space.  It will be
interesting to see if it can flap its arms hard enough to keep
off the rocks below.

				--Blair
				  "Or it's all hype and another Microchip."

Return to archive index | Next message in thread


From Comp.sys.wearables Newsgroup Archive (CSW)
Maintained by R. Paul McCarty

Archive created with babymail