Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!jarthur!nntp-server.caltech.edu!tybalt.caltech.edu!toddpw
From: toddpw@tybalt.caltech.edu (Todd P. Whitesel)
Newsgroups: comp.sys.apple2
Subject: Re: apple ii's with 68k's
Message-ID: <1990May5.075048.15379@laguna.ccsf.caltech.edu>
Date: 5 May 90 07:50:48 GMT
References: <9005042139.AA03276@apple.com>
Sender: news@laguna.ccsf.caltech.edu
Organization: California Institute of Technology
Lines: 45

[ these are intended as friendly comments, it's late so please don't take any
offense if I sound like I am flaming ]

It's a lot harder to turn your mega-add-on system into a single computer than
you think.

One thing that startled me about board design is that unless you use really
rare or expensive parts, motherboard costs are dominated by board size and
through-holes, and not by parts cost. Your proposed system would not be very
cost-effective compared to a 65816-based system with coprocessors custom
designed by Apple. (By the way, the Amiga's coprocessors are all 16 bit --
they have to be because they all fight for the the same block of RAM.)

Your add-ons probably total up to the cost of a Mac II, so I would compare
their combined resolution/speed/etc to a Mac II and not the IIGS.

Don't knock the 65816 until you've tried it. It doesn't make programming
trivial but is efficient enough to work well once you understand it. I have
written a bresenham's line drawer for the GS (optimized for 320 mode) that
is entirely software and whose critical loop takes 70 clock cycles to execute.
I am about to build a frame grabber board that uses a 65xxx to control VRAMs
by using NMI's on horizontal blanking.

The 65816 is deceptively simple. Many high-level operations on the 68000 and
8086 require short instruction sequences on the 65816, but often execute
in about the same number of clocks. The frequency of appearance of these
operations is such that the 65816's one byte opcodes make for a nice tradeoff.

The 65816 has a lot of stack primitives which make compiled languages easier.
Zero Page has been turned into Direct Page, and like the Stack may by ANYWHERE
in the lowest 64K of memory. The address space has been bumped up to 16 Mb and
there are enough addressing modes to adequately support it. There are a couple
things missing but since the opcode space is full (with two 'escapes' for 32
bit and math coprocessing) I finally had to admit that Mensch did a good job
of extending the 6502. (But he can't make faster chips to save his life. Thank
God for ASIC! 20 mhz to the people!! AE get your grubby paws off!)

The basic point is, to build a better Apple //, you don't need a 68000 and in
many ways it is actually a detriment because by the time you have supported it
properly -- You've got a Mac! The GS's 2.5 Mhz clock is the real problem --
GS/OS and a Transwarp (7 mhz) are proof that you don't need a 68000 to have a
responsive desktop machine.

Todd Whitesel
toddpw @ tybalt.caltech.edu