lharris@utgpu.UUCP (12/02/87)
Does anyone have circuits to make a 68020 pin compatible (and timing compatible) with a 68000. I've heard of a 7 chip solution somewhere but it wasn't a direct plug-in replacement. Also - what sort of speed increase could I expect if all else were kept the same (8 meg clock, 150ns ram, etc...) Thanks /leonard
gnu@hoptoad.uucp (John Gilmore) (12/03/87)
Motorola put out a tech note on this some time ago; below is a copy of
a message on it.
When I was at Sun, we made a couple of "turbo" piggyback boards that
did this, so we could play with 68020 samples without building a whole
new CPU board. It is really easy to do, I think it took 1 or 2 pals
and some sockets.
With the cache off, it runs SLOWER than the original 68000, since it
always prefetches 4 bytes of instructions, even when it won't use the last
two. With the cache on, it's somewhat faster than the 68000 -- the
simplistic benchmarks I can recall ran something like 10-50% faster.
One good reason to do this would be if you want floating point -- the
turbo board can easily have a 6888[12] on it, running full speed.
You might be able to run the 68020 at double the clock speed of the
68000 socket, making it a bit faster, but this will only win while
executing out of the I-cache, otherwise you wait for the old RAM.
Our turbo board didn't do this since we didn't have 20MHz 68020s then.
More interesting would be a 68030 piggyback, since it has not only an
MMU, but a data cache that could conceivably decouple you from the
slow memory bus enough to make a difference. I don't have the 68030
datasheets to see how easy this is, though -- I hear they didn't
make it pin compatible with anything.
John
From: randyb@halley.UUCP (randy banton)
Newsgroups: comp.sys.m68k
Subject: 68020/68010 app note
Message-ID: <283@halley.UUCP>
Date: 25 Aug 87 21:06:06 GMT
Organization: Tandem Computers, Austin, TX
I found one dated 4/29/85.
The title is:
MC68020 & MC68881 Platform Board for Evaluation in a 16-bit System
***Revision 2***
Advanced Microcomputer Applications Engineering
Microprocessot Products Group
Motorola Inc.
Austin, TX
The trouble is there is no official app note number on it.
It is about 20 pages long with text, schematics, pal listings, and
benchmark results.
Randy Banton
--
{pyramid,ptsfa,amdahl,sun,ihnp4}!hoptoad!gnu gnu@toad.com
"Watch me change my world..." -- Liquid Theatre