dorian@eddie.MIT.EDU (Dorian Garson) (12/03/87)
{I'm posting this at the request of a friend, but I'm interested, too!}
Has anyone attempted to replace the slug-bait 68000 in their ST with a
68010 or 68020?
If you were successful, what procedure did you follow?
If you blew it, what went wrong?
Thanks 10^6 -Dorian->
I-net: DORIAN@AIM.RUTGERS.EDU
Bitnet: dorian%gold.decnet@zodiac
UUCP: rutgers!unirot!dorian
CMS: ...!njmcsip!argo!SYSOPdorian@eddie.MIT.EDU (Dorian Garson) (12/05/87)
I received the following message in response to my request for info
about installing a 68010 or -020 in my ST:
+----------------------------------------------------------------------+
Hi,
I did the software part of the replacement of the 68000 with an
68020 in an ST, and maybe this is a good moment to report our
results. Note that although I have obtained a degree in EE long
ago, I'm not a hardware guru. This was a two person project
(Henk did the hardware stuff, I did the software).
Forget about using the 68010. It brings a lot of trouble without
any significant speed change (expect a few percents due to the
loop mode).
We therefore went straigt to the 68020. We decided to use the
PAK-68 card, as described in c't, august 1987 (a German
magazine). This card replaces the 68000 with an 68020 and
optionally an 68881. The 68020 works with a 16 bit data bus.
Also additional logic is provided for synchronous bus cycles.
As already stated by c't, and confirmed by Atari Benelux, this
board does not work with TOS. Probable cause: timing loops
and weird code (example: build your own stack frame and do an
RTE). The whole thing works with RTOS (a real time os from
Germany, not related to TOS).
Since RTOS is not that cheap, we decided to do some own
experiments first. The following was done:
>From Johan Stevenson, I obtained a beta release of MINIX (thanks
Johan). I made a new bootstrap routine. It worked.
(Ok, it was not that easy, its hard to debug bootstrap routines)
The results were a litte discouraging though.
We only tested with the dhrystone benchmark, as posted on the
net before. Minix/68000 did around 714 dhrystones/second.
This is almost equivalent with the DRI compiler from TOS, but
since both minix and the benchmark program on minix are compiled
with this compiler, this is not surprising.
The results on the 68020 were:
cache enabled: about 769 dhrystones/second
cache disabled: about 515 dhrystones/second.
There is no explanation why the latter result is so much worse
than the 68000 results. The difference with cache enabled and
disabled is about what was expected.
And this is where the story ends...
Henk sold his system, and I did not feel capable enough to find
the cause of this delay. I don't feel that the gain obtained now
is worth the investment in this card.
Any other experiences/results??
By the way: the response from Atari Benelux has always been one
of laughing: what you're trying to do is impossible.
Well, it can be done. Not with TOS, but only with real operating
systems (which use interrupts for I/O instead of timing loops).
Anyway: does anyone know exactly what to do when writing your
own loader for TOS. I know the story in the hitch hikers guide,
but would appreciate some additional info. Anyone?
Oh yeah; about that PAK card. In the november issue of c't they
describe a small card to switch between 68020 and 68000. For
those who want to go to the 68020 but don't want to give up TOS
completely (to play PSION?)
Please reply to the address below, and not to the address in the
header, since this is a shared account. Thanks in advance.
Frans Meulenbroeks
Centre for Software Technology
...!mcvax!philmds!prle!cst!meulenbr
or ...!uunet!prlb2!cst!meulenbr
+-------------------------------------------------------------------+
-Dorian->bryce@hoser.berkeley.edu (Bryce Nesbitt) (12/09/87)
In article <7567@eddie.MIT.EDU> dorian@eddie.MIT.EDU (Dorian Garson) writes: > >Has anyone attempted to replace the slug-bait 68000 in their ST with a >68010 or 68020? Drop a 68010 in and the Operating System (TOS) dies a horrid death. No electrical problems, just the software crashes out. The same problem would exist for the 68020, problems multiplied many times. You can boot an alternate OS, but then you don't get to run any of your ST software. I use an '010 in "that other machine". It speeds up the multiply/divide instructions *a lot*. Other than that the increase in speed is minimal. Since I had to replace a physically damaged 68000 in the machine anyway, the '010 made sense. |\ /| . Ack! (NAK, SOH, EOT) {o O} . bryce@hoser.berkeley.EDU -or- ucbvax!hoser!bryce (") U WARNING: hoser's spool directory eats a *lot* of mail. :-(
sl131002@silver.bacs.indiana.edu (12/11/87)
/* Written 5:55 am Dec 9, 1987 by bryce@hoser.berkeley.edu in silver:comp.sys.atari.st */ /* ---------- "Re: Turbo charging the ST" ---------- */ In article <7567@eddie.MIT.EDU> dorian@eddie.MIT.EDU (Dorian Garson) writes: > >Has anyone attempted to replace the slug-bait 68000 in their ST with a >68010 or 68020? Drop a 68010 in and the Operating System (TOS) dies a horrid death. No electrical problems, just the software crashes out. The same problem would exist for the 68020, problems multiplied many times. You can boot an alternate OS, but then you don't get to run any of your ST software. I use an '010 in "that other machine". It speeds up the multiply/divide instructions *a lot*. Other than that the increase in speed is minimal. Since I had to replace a physically damaged 68000 in the machine anyway, the '010 made sense. |\ /| . Ack! (NAK, SOH, EOT) {o O} . bryce@hoser.berkeley.EDU -or- ucbvax!hoser!bryce (") U WARNING: hoser's spool directory eats a *lot* of mail. :-( /* End of text from silver:comp.sys.atari.st */