• Join Us |
  • |
  • Sign in with:

A New CPU?

by Steve Gibson, Sep 23, 1999 1:55pm PDT

Thanks Brent for the lead, this has got to be one of the most interesting articles to pop up in quite a while. Silicon Secrets- Playing 20 Questions with Transmeta over at pathfinder.com is a speculative article which talks about the possibilites of the purpose of Transmeta. The company which Linux creator Linus Torvalds is now employed, and has funding from all over the place including ex-Microsoft employees. The speculation right now is that they are working on a new CPU architecture, dig this quote:

Most evidence suggests Transmeta is inventing a new kind of microprocessor so fast that it'll make a Pentium III feel like an abacus soaked in Jell-O. The chip, it's rumored, is built around a wholly new architecture, supposedly fast enough to run software emulating a standard PC's Intel x86 chip faster than any x86 chip could run itself. John Dvorak, a respected columnist for PC Magazine, stated confidently in February that the chip would be a "generic processor engine" that can run code for any kind of microprocessor -- Pentium, PowerPC, SPARC, you name it -- and translate it on the fly to its own native instruction set.


Read all the News


Comments

34 Threads | 34 Comments
  • #34
    Refresh Thread
    By: kerrigan x
    Reply
    #33 - no, no, no. An Athlon can obviously run Pentium instructions. How do you think it runs the PC version of Windows? Also, the register size in a CPU does not depend on instruction length AT ALL. An x86 (Pentium/Athlon/etc.) processor always has 32 bit registers, but the instructions can be anywhere from one to a dozen bytes long. I have no idea what \"programmable registers\" means--aren\'t all registers \"programmable\"??

    The way this chip can handle multiple instruction sets is because it is full of programmable logic gates. You can just load in some circuitry to handle x86 instructions, or different circuitry to handle PowerPC instructions, etc. The glaring disadvantage to this is that programmable gates are huge compared to \"normal\" gates, so you can only have a few thousand on one chip (not millions, like other processors) and they\'re slower, too.

    -Thomas.Kerrigan@Colorado.EDU
    Sep 25, 1999 2:41pm PDT