Conference hydra::amiga_v1

    
    I haven't seen much info about an A3000.  Someone from Commodore
    did make a few comments about the 68020 board which they are developing
    which goes in the A2000.  Maybe someone has this information lying
    around.  One important thing I remember is that it will have
    an MMU which would significantly aid in an implementation of UNIX
    or for that matter enhancing AmigaDOS.  There is a slot in the A2000
    specifically for this board and it should not be considered an
    afterthought.
    
    regards,
    
    steve mcafee

    As a matter of fact, there is an interesting article in the February
    '88 issue of AmigaWorld. It talks about the CSA 68020 coprocessor
    and memory that can be added to any Amiga (500, 1000, or 2000).
    Though the CSA product is expensive, there appears to be another
    outfit that offers the same product at a lower price.
    
    The article is NOT just marketing hype. It points out several points
    that should be considered.  Such as, the fact that the 68020 will
    not always speed things up, due to the existing bus, and the special
    graphics chips....
    
    However, the interesting thing to note is that the product is available
    today. Not later, not with the A3000, nor with and Atari system.
    
    =jbf=

    re: .1
    The Commodore 68020 board will plug into the "CPU" slot of the Amiga
    model 2000, and run a 68020/floating point/memory management chip
    combination over the Amiga 16-bit bus.  Several Commodore-Amiga
    employees have hinted on the Usenet Amiga newsgroup that these boards
    are working in their offices and will be announced Real Soon Now.

    Processor speed will be either 7.16 or 14.28 MHz, don't know which,
    but I don't believe there will be any 32-bit-wide memory.  So, as
    they say, your mileage may vary.  The board can probably be adapted
    to the A1000 but NOT the A500.
    
    Re: .2
    Looking over the AmigaWorld review of the CSA board set, several
    points sounded kind of funny.  This was the first 68020 article
    I have ever seen which never mentioned the word "cache."  Given
    the mumbo-jumbo about 24-bit vs. 32-bit addresses, and the author's
    ignorance of the DeciGel patch, which allows user programs to get
    the condition code register, the quality of the review doesn't look
    very good.  I don't trust the few performance numbers in the article.
    
    Without going into great detail, the CSA boards for the A2000 look
    like a repackaging of their stuff for the A1000.  CSA breaks Amiga's
    rules for hardware expansion, though apparently with no performance
    penalty.  Other people have told me that their older CSA boards
    speed up the Amiga even without 32-bit memory.  But with wide memory
    and a floating-point chip, the machine really screams.
    
    BTW, the operating system and both popular C compilers already handle
    the 68020 and 68881, but most commercial software titles don't take
    full advantage of these chips. 

    Finally Technologies is shipping their Hurricane board, which includes
    a 14.32 MHz 68020/68881.  The board includes a connector which will
    accept 2 megs of 32-bit RAM.  Retail price is $950, mail order should
    be around $800-850.
    
    Absoft Fortran, Manx C (Lattice too??) include support for the 68881
    math chip.  See note 600 for some interesting performance benchmarks
    on math-intensive computations.  A 14 MHz 32 bit Amiga is faster
    than a VAX 8650 by a slim margin (at least for this silly benchmark).
    
    I read a message on PLINK from a person who witnessed a demo of
    release 1.3 of the Amiga system software.  Somehow, the new release
    will always use the 68881 chip if present, even if the program was
    not compiled for it.  Maybe they're finally going to use the Amiga
    math libraries to their full potential.
    
    Anyway, Dave Haynie (the A2000 guy) has publicly stated on PLINK that 
    the Amiga 3000 is quite a ways off, and is more of a concept than a
    real project.  It would probably have much higher screen resolution,
    more bit planes, a 68030, special video DRAMS, etc etc.  Should
    be a dream machine, but don't hold your breath waiting.  
    
    The MAC ][ really isn't all that expensive; a color system with
    2 megs and a 40 meg HD is around $5500 if you shop around.  It's
    actually pretty tempting.  However, I've been spoiled by cheap software
    and the Amiga's use of multiple virtual screens, not just windows.
    And multitasking on the Amiga is a breeze.  The few Multifinder
    demos I've witnessed were rather underwhelming.
    
    Ed.
        
    

    re: .0
    
    as the previous notes say, you don't need to wait for an A3000 to
    get a system with a 68020/68881.
    
    The Feb 88 issue of AmigaWorld has an ad for a CSA 68030 board called
    the "Over 030" board.  Says they will begin shipping next month after
    the initial debut at the CEBIT Hannover Computer Fair in Germany.  Also
    says they are already shipping the 68882 math coprocessor as an option 
    for their 68020 boards, and also says they have a single board that
    holds up to 32Meg of 32-bit dynamic ram.
    
    In the A2000, Commodore has finally implemented slots instead of
    just documenting how they should work.  That will go a long way
    for encouraging add-on hardware.  The A3000 may end up being whatever
    the market determines is a popular "stuffed" A2000 configuration
    (just move the board stuff to the motherboard to free up the slots).
    
    -dave

    Re .3
    
    	As far as I have seen none of the current C compilers generate
    WORKING 68881 code. MANX said they do but their are so many
    bugs that you can't get a program to work.
            
    	None of the C compilers even try to generate 68020 specific
    opcodes.
    
    			Steve Peters
    

    re: .2, .3
    
    I asked a couple of questions on Usenet about Amiga performance
    and the 68020, and the upshot of a few replies was that:
    
    a) _Amigaworld_ is still not to be trusted about anything technical.
    
    b) If the CSA board really does perform that badly then CSA did
    a lousy job of implementation.

    c) Any well-designed add-in 68020 board should give you a 30-50%
    performance improvement by itself, and 200-400% relative performance
    with 32-bit wide memory.

    So, you gotta be cautious, but _some_ '020 boards will give you
    a big speedup.