🚀 go-pugleaf

In article <2278@hudson.acc.virginia.edu> gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) writes:
>In article <23970@cup.portal.com> Xorg@cup.portal.com
>(Peter Ted Szymonik [ quoting dave small ]) writes:
>
>[ Dave Small's blurb about this maybe-030 upgrade for the ST... ]
>
>> We will also "clock" the 68030 processor at a much higher speed
>> than the ST's 68000. The ST runs at 8 mhz (millions instructions / 
>> second); we'll go for at least 16 Mhz and possibly more.
>
>This is wrong. ...
[Rationalization deleted]
>Gosh, I don't really remember anymore. Anyway, it isn't 8 million
>instructions per second.

The 68000 processor takes about 8-16 cycles per instruction, on the
average.  The minimum instruction time is 4 cycles, up to a worst
case of 170 cycles for one instruction (a signed divide on an
absolute memory address).

The ST therefore operates at about .75 million instructions per
second, or MIPS.

With a 68030, however, there are three speed improvements without even
changing the clock--the instructions take fewer cycles, memory
is accessed in 32-bit chunks instead of 16, and the 68030 has a
cache, meaning that it reads data and instructions before they are
needed, while it's doing other things.

Result: You can get a 4-8 times speed improvement without even
changing the clock rate, and with twice the clock, 8-16 times...

These numbers are approximate, and if you want to argue them with
me, don't waste net bandwidth, please!  Respond via email and we can
hash it out there...

-------------
Tim Mensch.
Internet: tmensch@ucsd.edu
Bitnet:   UCSD!TMENSCH

P.S. I agree that the Atari vs. the world bickering has gone on too
long, and should be rerouted to either email, alt.flames, or
/dev/null.  I'm not picky.