Conference 7.286::atarist

        Ken,
        
        GEM has nothing to do with the calculations.  It is only used for
        screen/user interface management.  If you want to speed THAT
        portion up, then the TurboST software product is all you need.
        
        As for the calculation section - does your Rainbow have a FPP
        (8087) installed?  If so then that may be the reason that the
        Rainbow performs calculations faster than your 1040.  Without a
        FPP the 1040 should be about 2x faster than the Rainbow on
        calculations.  The only things that I know of to improve the
        basic calculation speed of the 1040 is to add a FPP processor,
        which is not easily done and your spreadsheet program would have
        to be recompiled with FPP support (not very likely) or make your
        calculations simpler (which defeats the purpose of using a
        computer in the first place).
        
        If you really think that VIP is the problem, then you might look
        at the new spreadsheet by LDW - Power Spreadsheet.
        
        Bob

    
    The same company that made my 16 MHz Processor Accelerator for the
    Amiga is working on an Atari and Mac version.
    
    It's basically a small board that plugs into the 68000 socket and
    has it's own clock and 16 MHz MC68000.  The Amiga version also is
    socketed for a MC68881 math chip.  Software is provided that allows
    you to toggle from high to low speed and to enable and disable the
    math chip.
    
    I don't know how math is handled under GEM/TOS, but the Amiga uses
    a special disk based library that is called by applications that
    want to perform FFP operations.  When the math chip is present and
    enabled, the library will automatically use the math chip instead
    of the software routines.
                             
    This board DOES provide real speedups, ranging from 10% to 40%.
    My own benchmarks confirm this.  Applications that call the math
    library (and hence the math chip) see astounding speed increases,
    up to 1000%.  In general, the machine just 'feels' faster.
    
    The Amiga version costs $169, I expect the Atari version will be
    similarly priced.  It's a good, cheap alternative to a 68020 board.
    
    CMI stands for Creative MicroSystems Inc.  If anyone is interested,
    I'll post there number here.  There's a very knowledgeable fellow
    named Bill Caldwell whom I've spoken to a few times.
    
    
    
    

>    It's basically a small board that plugs into the 68000 socket and
>    has it's own clock and 16 MHz MC68000.  The Amiga version also is
>    socketed for a MC68881 math chip.  Software is provided that allows
>    you to toggle from high to low speed and to enable and disable the
>    math chip.
    
  Just out of curiousity, does the board have cache memory? Or does it
just run the processor fast, and insert a lot of wait states to get to
memory?
    
>    The Amiga version costs $169, I expect the Atari version will be
>    similarly priced.  It's a good, cheap alternative to a 68020 board.
    
  Humm, for this price, probably no cache...

>    CMI stands for Creative MicroSystems Inc.  If anyone is interested,
>    I'll post there number here.  There's a very knowledgeable fellow
>    named Bill Caldwell whom I've spoken to a few times.
    
  Please post this. I'd love to get some cheap, quick MIPs!
    
    

    Thanks folks,
    
    	for your help. 
    Just to add to the knowledge, my Rainbow has no FPP, and has only
    750K memory. I know, it seems cock-eyed to me too, I would have
    thought that the 1040 would be much faster.
    
    I'll try and get my hands on the TurboST gadget, and see if that
    helps.
    
    Merry Crimble!
    
    Ken.T.

    
    Re: .3
    
    According to Bill at CMI, instructions that execute in four cycles
    or less see no speedup.  Instructions longer than four cycles will
    average 20%, with shifts, multiplys, and divides hitting as high as
    50%.
    
    The biggest perceived speedup comes from reading the ROM routines
    at high speed.  However, you'd better hope that your ROMS can in
    fact BE read at high speed.  I've had no problems with my Amiga
    1.2 ROMs, but I've heard of some marginal 1.3 ROMs.  Bill was thinking
    of adding a few more waits to read the ROMs at around 12 MHz.
    
    Creative Microsystems Inc.
    Portland Oregon
    (503) 684-9300
    
    Ed.
    
    

    Postscript to above:
    
    I tried TurboST, guess what, as predicted, no change to the
    computational speed.
    
    I think I'll look at some alternatives, and publish if I find anything
    faster/better.
    
    Happy new year!
    
    Ken.T.