Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!UREGINA1.BITNET!GORRIEDE
From: GORRIEDE@UREGINA1.BITNET (Dennis Robert Gorrie)
Newsgroups: comp.sys.amiga
Subject: Re: (A) Amy 500 -> 2000? (B) 68040 vs. gfx coprocessor
Message-ID: <9002130521.AA09589@jade.berkeley.edu>
Date: 13 Feb 90 05:04:30 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Lines: 57

>People at NewTek agree that while the mode can be used to produce some really
>great pictures there are many limits to what can be done.   However, the
>NewTek advertising people prefer to really hype these new modes for what they
>are worth.  Anyway I don't think you need to take my word, since there are

>Also you mention dithering.  Actually color dithering is very difficult.  I

My  point  was that you overly criticized the lack of colors; you seemed to
have forgotten that dithering CAN achieve impressive results.

>The blitter can easily be replace with code.  The only operations that are
>software also.  It is just a question of speed.  A 68040 is so much faster
>than a blitter that there is no reason to waste time on a blitter.

Almost  ANYTHING  can  be  replaced with code.  But some code for the 68040
might  replace  a  slow  blitter  operation, and actualy run faster.  BUT I
clearly  said,  there  is no reason why a faster blitter could not take its
place.  A blitter with speed and access times comparable to the 68040 would
not  be  a  'waste of time'.  The major bottle neck would be the RAM speed.
However,  when  you  realize  that  our  future  GUI's and applications are
entering  the  relm  of  1kx1k  resolutions,  ANY co-processor that is of a
comparable speed to the cpu, can reduce the work load for the cpu.

>Now for the TI 34010 chip, I can only assume you don't know want it is.  It
>is a general purpose processor that has hardware and instructions to make
>graphics easier.  It isn't limited to just graphics usage although it is

Thats news to me.  I've been writing ML code for a NEC Graphics Engine for
the  last  few  months.   AND using instructions comparable to blits, I was
sorry  to  see it was NOT that much faster on equivalent resolutions.  MOST
of  the  speed  improvement  that  was there was due to the fact that it is
running a 50mhz TI 34010 with 1 meg of 70ns RAM.

>Then you could still do many more things with it besides.  The Amiga blitter
>is a small limited toy in comparison to the TI 34010.

Exactly.  The blitter is only a very small subset of what is available on
the TI.  WHY?  Because , as you stated, you can do the rest in software.
The fact is, a TI in a graphics card for the Amiga would be MUCH more
expensive than mass-producing a faster blitter.  And, it would have some
limitations, not being to access chip memory directly.  It could not replace
the current hardware implementation of functions of a blitter.

A TI right on the co-pro bus would seem more useful, but it would  be
limitted to the bus and ram speed, and for some operations, you would see
no improvement over the blitter.  For others, there would be a dramatic
improvement.  But, can Commodore actually expect to sell a machine like
this  that would cost the user $1k or $2k more in hardware?  Not to mention
the  added  overhead  of  software  support  interface with the different
memory graphics types and other co-processors.  Even when it becomes
economicaly feasible, it will take some time before hw/sw developement has
reached the point where this could be part of a basic Amiga x000.

+-----------------------------------------------------------------------+
|Dennis Gorrie                 'Chain-Saw Tag...                        |
|GORRIEDE AT UREGINA1.BITNET                    Try It, You'll Like It!'|
+-----------------------------------------------------------------------+