denbeste@bbn.com (Steven Den Beste) (10/03/89)
Last summer, with dreams of megaflops dancing in front of my eyes, I picked up
something called a "midget racer" - a 68020 drop-in board - to put in my Amiga.
Later I got a 16MHZ 68881 and a 16MHZ crystal. Sigh. whenever the 16 MHZ
crystal is installed, I got to visit the guru-farm every few minutes - but only
if my Starboard II (2 meg, SCSI) was installed.
Calls to CSA led to a suggestion that I do the PAL grounding mod (funny, I
thought I HAD done it, but turns out I did something else) - which didn't help.
I decided that the problem was that the grounding to the Starboard wasn't good
enough. I had heard from someone that THEIR Starboard was connected with
screws, so I spent a frustrating afternoon trying to install screws in mine
(only to find out there are two case designs, and mine can't use screws - this
was later) - and I fried my Starboard.
It was away for 2 months, during which Microbotics couldn't find anything
wrong. (They only charged me $25 for what seems to have been a cockpit error -
I thought I had toasted some CMOS - I now think I shook the SCSI daughter board
loose. They're good folks.)
So, it's back now. It works, but I'm having some sort of noise problem with the
SCSI interface, where my machine locks up with a guru #3 or #4 during some disk
operations.
This happened once during a file write, and now my disk has to validate every
time I boot - and it brings up a requester reporting "Key #XXXXX already set"
with "retry" and "cancel" buttons. "Retry" just reports the problem again, and
"cancel" aborts the validation. (What I WANT is an "IGNORE IT AND PROCEED"
button, damnit!)
I don't know how to clear this condition, and the opsystem won't let me write
to the drive as long as it persists. I don't have the actual number with me
right now - if it is important, I'll send it to whoever asks.
The only way I can think of right now to clear this up is to roll all 45 MB
onto floppies with Quarterback, reformat the drive, and roll it all back. The
prospect of doing this 5 hour job just doesn't thrill me somehow, but that's
what I'll do if I have to.
Is there a simpler way?
[Oh, by the way, after all of that, the 68020 running at 7.16MHZ doesn't give
really impressive performance gains. Oddly, it does slow down the little tune
during boot by probably a factor of 4 - which makes no sense to me. On the
other hand, it speeds up the intro to Scrabble by probably a factor of 3 (cache
effects on busy-wait loops?).]
Steven C. Den Beste || denbeste@bbn.com (ARPA/CSNET)
BBN Communications Corp. || {apple, usc, husc6, csd4.milw.wisc.edu,
150 Cambridge Park Dr. || gatech, oliveb, mit-eddie,
Cambridge, MA 02140 || ulowell}!bbn.com!denbeste (USENET)daveh@cbmvax.UUCP (Dave Haynie) (10/04/89)
in article <46373@bbn.COM>, denbeste@bbn.com (Steven Den Beste) says: > [Oh, by the way, after all of that, the 68020 running at 7.16MHZ doesn't give > really impressive performance gains. Oddly, it does slow down the little tune > during boot by probably a factor of 4 - which makes no sense to me. On the > other hand, it speeds up the intro to Scrabble by probably a factor of 3 (cache > effects on busy-wait loops?).] That's not surprising -- if you turn the cache off, you'll go slower almost all the time than the 68000. The deal is that the 68020 really wants a 32 bit bus. It always prefetches 32 bits for it's instruction pipleline, even if only 16 will be needed. So you end up throwing out words with the '020 that you'd never have fetched with the 68000. The enabled cache generally overcomes this slowdown, plus maybe 10% or so, so on the average, the '020 in its normal setup will go a tad faster than the 68000. The only real reason to buy a Midget Racer, or any 68020/68030 board without 32 bit wide memory is for the floating point unit. > Steven C. Den Beste || denbeste@bbn.com (ARPA/CSNET) -- Dave Haynie Commodore-Amiga (Systems Engineering) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy Too much of everything is just enough