home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.intel
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!wupost!cs.utexas.edu!asuvax!chnews!hfglobe!mithril!flamer
- From: flamer@mithril.intel.com (Jim Trethewey @mithril)
- Subject: Re: Idea for the P5!
- Message-ID: <flamer.722187076@mithril>
- Sender: news@hfglobe.intel.com (News User)
- Nntp-Posting-Host: mithril.intel.com
- Organization: Intel Corporation
- References: <1992Nov18.203321.18612@msuinfo.cl.msu.edu> <1992Nov18.222927.20422@nas.nasa.gov>
- Distribution: usa
- Date: Thu, 19 Nov 1992 15:31:16 GMT
- Lines: 16
-
- In article <1992Nov18.203321.18612@msuinfo.cl.msu.edu>, sundaram@eecae.msu.edu writes:
- > Instead of devoting part of the die to emulate 80x86 code, why not include some
- > instructions that will allow the P5 to act as a master processor to a true
- > (or slightly modified) 80x86 chip.
-
- As we have seen with other co-processor scenarios (most notably 8087, 287,
- and 387), you sacrifice a lot of speed by having to "go off chip" as
- opposed to having the functionality onboard. Although many people think
- the 486SX upgrade "trick" is stupid, I think it was a smart idea for
- exactly this performance reason. Another factor to consider is that a
- large percentage of the power drawn by a microprocessor is devoted to
- driving the I/O pins. The less you wiggle those lines, the cooler it will
- run. Hence another advantage of onboard cache and coprocessor functions.
- --
- --Jim Trethewey, $WORK = flamer@mithril.intel.com, $HOME = flamer@alfirin
- 5200 NE Elam Young Parkway HF3-73, Hillsboro OR 97124-6497 (503)696-5444
-