home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.intel
- Path: sparky!uunet!munnari.oz.au!labtam!graeme
- From: graeme@labtam.labtam.oz.au (Graeme Gill)
- Subject: Re: 960 XA/MX Questions
- Organization: Labtam Australia Pty. Ltd., Melbourne, Australia
- Date: Mon, 11 Jan 1993 02:28:41 GMT
- Message-ID: <1993Jan11.022841.17445@labtam.labtam.oz.au>
- References: <1993Jan6.023733.5129@netcom.com> <1993Jan08.173711.14431@donau.et.tudelft.nl>
- Lines: 19
-
- In article <1993Jan08.173711.14431@donau.et.tudelft.nl>, wolff@erasmus.et.tudelft.nl (Rogier Wolff) writes:
- > I am not sure what the 960MX/XA have that the 960CA doesn't, but our Xterminals
- > seem to run just fine with the CA. As far as I know, there exists 960's
- > with floating point coprocessors that are Military only like the MX/XA.
- > Isn't the X protocol completely floating point free as to make a floating
- > point coprocessor useless in such an environment?
-
- I'm sure they do run fine, but only because they're running
- floating point emulation code. If you ever get the chance to benchmark a 960CA and
- a 960KB terminal you will notice that the arc performance
- of a CA is noticeably worse than a 960KB (but be careful - arc caching
- covers this up for repetitive tests). When it comes to a terminal
- running PEX, then the 960CA is awesomely slow compared to a processor
- with an f.p. unit. And if you would like a choice of downloadable clients
- or graceful memory degradation on your terminal, then your out of luck
- unless you've got a cpu with a memory management unit.
-
- Graeme Gill
-
-