Path: utzoo!utgpu!attcan!uunet!ncrlnk!ncrcae!hubcap!gatech!rutgers!bellcore!texbell!killer!aimania
From: aimania@killer.DALLAS.TX.US (Walter Rothe)
Newsgroups: comp.sys.amiga
Subject: CMI accelerator problems
Keywords: CMI 68020 68881 crosstalk
Message-ID: <6052@killer.DALLAS.TX.US>
Date: 10 Nov 88 06:42:30 GMT
Organization: The Unix(R) Connection, Dallas, Texas
Lines: 52


I bought a CMI accelerator board at a good price from AMAZING, one of the two
amiga stores in Dallas and have run into some interesting information about
the board. I have a 2000 from the first trade in your 1000 deal and a
Starboard II with adapter card for conversion from 1000 to 2000. I also own
a Stardrive, also a Microbotics product. I have been real happy with the
service from Microbotics but it sure helps having them in town.

Anyway, I bought the CMI board with a 12MHZ 68881 and proceeded to try it out.
I took it home and found that it would not fit into the 68k socket well. It
interfered with the chip immediately to the right of the 68k. Being a
computer designer from way back, I had no qualms about buying a gold
extender socket and trying it out. This seemed to work. However, the Amiga
seemed to guru alot. I took the 68881 out and it worked fine.
I also tried the board without the Starboard and with the 68881 and it
worked fine. I called CMI and they said that they had run into some
crosstalk problems and were going to go to a multilayer board soon. They
said that they had a fix that seemed to correct the problem so I said I would
give it a try. They suggested cutting the etch to pin 6 of the 68k socket
and wiring a solid wire above the board to U3-12. Note that this does not
functionally change anything, it just may add some noise immunity or change
the reflection characteristics. They also suggested improving the grounding.
As a side note, the Lucas board has a 30 ohm resister in series with the
address strobe but I did not try this.
I also asked them what to set the jumper configuration to since the pamphlet
I got was missing the part about jumper settings. They suggested taking the
jumper off so that there were no jumpers on the board. The combination of
doing these two things helped alot. My system now made it through bootup
consistently and seemed to work fine most of the time. However, when I would
run my telecommunications package in the background trying to connect to
another machine over the phone and at the same time tried to load a large
program into Matt's DME editor, I could consistently Guru.
I called CMI back and they started accusing Microbotics and asked that
I call them and find out why it didn't work. I did this and Microbotics said
that they got a CMI board in and it did not work with any expansion devices.
At this point, I decided to give the board back to the dealer and let them
send it back. One of the dealers also sent all their CMI boards back also.

I really like the idea of having a cheap accelerator board but there seems
to be a few problems left to work out. Hope someone from CMI will address this.

By the way, the floating point speedup achieved from any 68881 connected to
a 68k will not be as great as that you can get by connecting a 68881
directly to a 68010 or 68020. There will still be a significant speedup on
double precision floating point calculations but only a small( 2 or 3 times )
speedup over the single precision Motorola FFP format.



-- 
Walter Rothe at the UNIX(Tm) Connection, Dallas, Tx
UUCP: {rutgers}!smu.killer.aimania