🚀 go-pugleaf

in article <11988@steinmetz.ge.com>, perley@caesar.steinmetz (Donald P Perley) says:

> I don't know about the CMI board, but some 68000 speedup boards have
> floating point processor sockets.  If you use it, and are running 
> programs with a lot of FP computation that support it, you should
> get a dramatic speed increase, yes/no? 

You'll get a speedup over no 68881 at all (assuming 1.3 math libraries).  What
you don't get is a 32 bit interface to the 68881, or a true coprocessor
interface to the 68881, so this is still considerably slower than a similarly
clocked 68020.  Given the speedups you can get with any kind of math chip,
it's probably a good idea for someone into image rendering or other heavy
duty math operations who doesn't have a 68020 type budget.

> If you do a lot of floating point stuff (and a lot of graphics programs do),
> it sounds like a lot more bang for the buck compared to a 68020 board
> with coprocessor at ~$1000 more (plus 32 bit memory to get a real advantage
> over the 68000).

Note that for math operations, a 68020 with 68881 is several times faster than
a 68000 with 68881, even on a 16 bit bus.  The difference is of course the CPU
interface that the 68020 uses, plus the wide data bus.  The 68000 or 68020
have to fetch data for the 68881 in either case.  But while the 68000 will fetch
at 16 bits, then transfer at 16 bits, the 68020 will do it's transfers at 32
bits.  Of course, if you can only afford a fast 68000 board at $200-$300, it's
going to speed up your math more than that cash sitting in the bank waiting
for a 68020 board will...

> -don perley
-- 
Dave Haynie  "The 32 Bit Guy"     Commodore-Amiga  "The Crew That Never Rests"
   {ihnp4|uunet|rutgers}!cbmvax!daveh      PLINK: D-DAVE H     BIX: hazy
		"I can't relax, 'cause I'm a Boinger!"