[net.micro.amiga] Amiga shown with 68020

bobp@amiga.UUCP (Robert S. Pariseau) (11/21/85)

TITLE:  Amiga shown with 68020

At Comdex yesterday, Computer System Associates (CSA) showed their
68020 board for the Amiga.  This board includes a 68020, a 68881 math
processor, and 256K bytes of 32-bit bus RAM.  The board replaces the
68000 in the Amiga, plugging into it's socket.  The board fits under
the Amiga FCC shield.  No release date was announced.

The board was demonstrated using a pre-release version of the new
Amiga V1.1 system software.  The V1.1 software is compatible with
the 68010 and the 68020 processors.  Application programmers wishing
to take advantage of such products should be sure to use the Exec
GetCC() function to get the processor condition codes.  (GetCC() handles
the differences between the processors such as the fact that MoveSR
is an invalid user mode operation in the 68010 and 68020.)  In addition,
you should be sure to NOT use the upper 8 bits of a pointer for storing
unrelated information, since the 68020 uses all 32 bits for addressing.

For those doing systems work, Exec maintains flags in the AttnFlags
field of ExecBase which describe the type of processor in your
machine.  Exec's coldstart procedure will update certain library entry
vectors as necessary to maintain compatability.  If you write code
that uses the Supervisor() function, keep in mind that your stack frame
is processor dependent.  Use the new SuperState() function for
processor independent supervisor mode entry.

V1.1 ROMWack will correctly display information for 68010/20 address
and bus errors.

When the 68020 is detected during coldstart, its instruction cache is
enabled.  This has implications for programmers writing self-modifying
code.

The new Exec function TypeOfMem() should be used by programs wishing
to know whether a given address is in Chip memory or Fast memory.  This
function will insure compatability with future Amiga architectures.
It should be used instead of assuming that certain fixed address
boundaries define the break between Chip and Fast memory.

-----------

Computer System Associates, Inc.
7564 Trade Street
San Diego, CA 92121
(619)566-3911

kim@mips.UUCP (Kim DeVaughn) (11/24/85)

[ ... go ahead, eat my bits ... ]

> TITLE:  Amiga shown with 68020
> 
> At Comdex yesterday, Computer System Associates (CSA) showed their
> 68020 board for the Amiga.  This board includes a 68020, a 68881 math
> processor, and 256K bytes of 32-bit bus RAM.  The board replaces the
> 68000 in the Amiga, plugging into it's socket.  The board fits under
> the Amiga FCC shield.  No release date was announced.


I called CSA to get more information right after I read this posting.
I spoke to Pat (Patricia), who was very helpful, and quite personable;
she answered most all of my moderately technical questions.  Here's
the info I got:

	o  With the piggy-back 68020 board, an Amiga *System* is
	   ~$3075, and Pat quoted 4 wks ARO.  I'm not sure how "full"
	   the CSA board is at that price, as I was only interested
	   in their 020 board.
	   
	o  The CSA 68020 board, with a 68881 is $1480; 1 wk ARO.
	
	o  The CSA 68020 board, with 68881 and 256K of RAM is $1985,
	   and is ~3 wks ARO.  Pat said the machine at Comdex was the
	   1st prototype with RAM, but since they had had no trouble
	   with it, it was going into production NOW.

	o  The CSA board is also available without the 68881, but I
	   didn't get a quote.  I also got the *impression* that you
	   can buy the bare board from them.

	o  The question Pat couldn't answer was what Rev of the 020/881
	   chips are being used (the parts still have a couple of bugs
	   in them depending on Rev level, according to Motorola).  Of
	   course the parts are socketed, so you can get a newer Rev of
	   the chip(s) if you run into one of the bugs (after Moto fixes
	   it, of course).

	o  The 256K of RAM does not conflict with the Amiga internal
	   expansion RAM.  Pat *thought* it was contiguous with a 512K
	   Amiga.  According to the BYTE memory map, that'd put it in
	   the 1.5M reserved area beginning at x'080000'.  Hmmmm?
	   
	o  They don't yet have a really technical spec on the product,
   	   but Pat'll send you what they have ... sounds like marketing
	   fluff, and maybe a price-list.


Sounds like it could make the Amiga into a SuperAmiga!  Now if we could
just clock the thing at 16.7MHz (or even 12.5MHz).  I suspect that that
is a bit too tough to do though, since the 7.159+MHz clock is so intimately
tied to the video display, etc;  besides, who knows what the AmigaBus and
the custom chips can handle?  Of course the lack of an MMU becomes more of
a liability now, but it should still end up being quite a machine.


> The board was demonstrated using a pre-release version of the new
> Amiga V1.1 system software.  The V1.1 software is compatible with
> the 68010 and the 68020 processors.

Nice to know that the Amiga development organization is looking toward
the future ... sorta makes you wonder what goodies might be coming from
Commodore, doesn't it?  'Course I don't suppose that Commodore can make
any comments "on products that may or may not be in development"  (to
borrow an expresion from Blue :-) ).


>                                      Application programmers wishing
> to take advantage of such products should be sure to use the Exec
> GetCC() function to get the processor condition codes.
> 
> For those doing systems work, Exec maintains flags in the AttnFlags
> field of ExecBase which describe the type of processor in your
> machine.
>                          Use the new SuperState() function for
> processor independent supervisor mode entry.

Hopefully, this information has been "impressed" upon 3rd party ISV's
with sufficient emphasis and with enough lead-time so that most (if not
all) commercially available s/w will run correctly.  Some people still
do use s/w timing loops, etc. tho ...


> V1.1 ROMWack will correctly display information for 68010/20 address
> and bus errors.

Are the 68020/68881 opcodes, regs, etc. supported by the Assembler; by
the Lattice C Compiler?


I'd be interested in hearing from anyone who's seen the CSA board;  what
are your impressions of the performance improvement one can expect?  Anyone
have any benchmarks, or other technical info?  Opinions from you 68020
gurus?


Let's hear it for OPEN architectures, and 3rd-party vendor support;
Thanks, Commodore!!!

/kim


P.S.  What will happen to my warrenty if I install one of these puppies?


Boring disclaimer:  I'm not associated with Computer System Associates (CSA)
or Commodore in any way (except I own an Amiga).  All opinions are strictly
my own, though they can occasionally be leased.
-- 

UUCP:  {decvax,ucbvax,ihnp4}!decwrl!mips!kim
DDD:   415-960-1200
USPS:  MIPS Computer Systems Inc,  1330 Charleston Rd,  Mt View, CA 94043

tim@ISM780B.UUCP (11/26/85)

> == bobp@amiga in net.micro.amiga

> is an invalid user mode operation in the 68010 and 68020.)  In addition,
> you should be sure to NOT use the upper 8 bits of a pointer for storing
> unrelated information, since the 68020 uses all 32 bits for addressing.

Well, the 68020 puts out all 32 bits.  But does the system use them?  If
it doesn't, then it should be ok to play games with these bits.

						Tim Smith
						ima!ism780!tim
						ihnp4!cithep!tim

Disclaimer:  This posting is in no way an endorsement by me or my employers
of using bits in pointers for any evil, immoral, or illegal purposes.

max@plus5.UUCP (R. Max Mutrux) (11/27/85)

> Sounds like it could make the Amiga into a SuperAmiga!  Now if we could
> just clock the thing at 16.7MHz (or even 12.5MHz).  I suspect that that
> is a bit too tough to do though, since the 7.159+MHz clock is so intimately
> tied to the video display, etc;  besides, who knows what the AmigaBus and
> the custom chips can handle?  Of course the lack of an MMU becomes more of
> a liability now, but it should still end up being quite a machine.

As long as Amiga didn't do anything wanky like counting on the AS-DS delays
or the DS-/AS delays you should be able to supply the 68020 with a "private"
clock source. And since the 680xx is an async design when it goes outside
of the private 256k ram it should slow down. When I get my tech specs
on the amiga I'll check this out. I want to put the 20MHz 68020 on my amiga
and then run Drystone 24 hours a day.

Also for general info. CSA is using BIG static rams on there 68020/w 256k
board so don't think you can just change those 64k chips to 256k chips
and have a meg.

-- 
		(char)Max_Mutrux	..!{ihnp4,cbosgd,seismo}!plus5!max

dale@amiga.UUCP (Dale Luck) (11/28/85)

In article <39700008@ISM780B.UUCP> tim@ISM780B.UUCP writes:
>> == bobp@amiga in net.micro.amiga
>> unrelated information, since the 68020 uses all 32 bits for addressing.
>
>Well, the 68020 puts out all 32 bits.  But does the system use them?  If
>it doesn't, then it should be ok to play games with these bits.
>

CSA's amiga/68020 board has 256k of memory. It was configured to be
at 7f000000-7f03ffff. This caused a hickup in one of our libraries since
it was internal using some of the 8msb as some flags. That got changed
REAL FAST.  We advise all those that wish to write software that is
hoped to run on compatible but more powerful products to assume that all
pointers are 32 bits.
BTW: The 256k or more of memory that CSA puts on this 68020 board is
not included in the total 8mbytes of memory expansion available via the
expansion bus.  It lives in an address space above the 68000 16mbyte
space.

kim@mips.UUCP (Kim DeVaughn) (11/29/85)

> > Sounds like it could make the Amiga into a SuperAmiga!  Now if we could
> > just clock the thing at 16.7MHz (or even 12.5MHz).  I suspect that that
> 
> As long as Amiga didn't do anything wanky like counting on the AS-DS delays
> or the DS-/AS delays you should be able to supply the 68020 with a "private"
> clock source. And since the 680xx is an async design when it goes outside
> of the private 256k ram it should slow down. When I get my tech specs
> on the amiga I'll check this out.

One can *hope* that this level of info is in the H/W Ref Man ... I wonder if
CBM-Amiga has a set of maintenance manuals for 3rd-party repair people?  That
should include the schematics, timing diagrams, etc.  'Suppose you have to be
a Registered Something-Or-Other to get 'em tho ... sigh!


>                                   I want to put the 20MHz 68020 on my amiga
> and then run Drystone 24 hours a day.

I've not worked with 680x0's until now ... is also possible to run a 68881
with a fast '020 (I don't *think* fast 881's have been announced yet)?  If
not directly, then by playing some reasonable h/w tricks (split clocking, or
some such)?

/kim
-- 

UUCP:  {decvax,ucbvax,ihnp4}!decwrl!mips!kim
DDD:   415-960-1200
USPS:  MIPS Computer Systems Inc,  1330 Charleston Rd,  Mt View, CA 94043

gnu@l5.uucp (John Gilmore) (12/03/85)

In article <249@mips.UUCP>, kim@mips.UUCP (Kim DeVaughn) writes:
> > > Sounds like it could make the Amiga into a SuperAmiga!  Now if we could
> > > just clock the thing at 16.7MHz (or even 12.5MHz).
> > As long as Amiga didn't do anything wanky like counting on the AS-DS delays
> > or the DS-/AS delays you should be able to supply the 68020 with a "private"
> > clock source. And since the 680xx is an async design when it goes outside
> > of the private 256k ram it should slow down.

Probably the Amiga really does depend on all those unspecified delays.
When you're trying to cram 3 RAM cycles into every CPU memory cycle, you
use all the kludges you can...

In any case, it should be possible to run a 68020 at exactly twice the
standard clock rate of the Amiga (or any other 68000 system).  This
would make the interface synchronous to the rest of the Amiga board and
you could tune the timing of the various signals to match what the
Amiga expected.  Given that a 68020 cycle would take 3 new-clocks and a
68000 cycle would take 8 new-clocks, there's plenty of time to fiddle
with generating all the right fake strobes.  If you took the time to test
it on all of them, you might even be able to build one piggyback board
that would run in a Mac, Atari, or Amiga, by emulating the 68000 timing
exactly enough.

> I've not worked with 680x0's until now ... is also possible to run a 68881
> with a fast '020 (I don't *think* fast 881's have been announced yet)?

You can run the 68020 and 68881 on totally independent clocks.  It uses
standard (wait for DTACK) bus cycles to talk back and forth.

herbie@polaris.UUCP (Herb Chong) (12/03/85)

In article <39700008@ISM780B.UUCP> tim@ISM780B.UUCP writes:
>
>> == bobp@amiga in net.micro.amiga
>
>> is an invalid user mode operation in the 68010 and 68020.)  In addition,
>> you should be sure to NOT use the upper 8 bits of a pointer for storing
>> unrelated information, since the 68020 uses all 32 bits for addressing.
>
>Well, the 68020 puts out all 32 bits.  But does the system use them?  If
>it doesn't, then it should be ok to play games with these bits.

suppose i were rich and managed to both buy the 68020 board and 32M of
memory to go with it?  'nuff said.  don't ever use those "spare" address
bits or you'll eventually go through what IBM is now with upgrades to 31-bit
addressing unless you know you will never upgrade and/or will junk your
machine before the amiga comes standard with a 68020.
 

Herb Chong...

I'm still user-friendly -- I don't byte, I nybble....

VNET,BITNET,NETNORTH,EARN: HERBIE AT YKTVMH
UUCP:  {allegra|cbosgd|cmcl2|decvax|ihnp4|seismo}!philabs!polaris!herbie
CSNET: herbie.yktvmh@ibm-sj.csnet
ARPA:  herbie.yktvmh.ibm-sj.csnet@csnet-relay.arpa
========================================================================
DISCLAIMER:  what you just read was produced by pouring lukewarm
tea for 42 seconds onto 9 people chained to 6 Ouiji boards.