$CSD211@LSUVM.BITNET (Mark Orr) (11/24/90)
In response to Todd Whitesel's response to my response to Todd's response
to Apple Defender's idea to build a renegade apple //gs:
Okay..all right, mabye emulation with another processor isn't such a great
idea after all...but it did have some good points, notably it would
eliminate the need to write GS-clone ROMS (i.e. you could write a '816
emulator, put the hardware in the right places (memory locations), and then
copy the ROM's of a GS and install them as ROM images, as per the UNIX //e
emulator, only this time it would have some hardware support).
But still, the ideas of the Apple //f and a RISC-Apple II aren't mutually
exculsive. One could build a modified //f and include on the motherboard
a RISC (either the enhanced ARM (VLSI V86C020), or the AMD 29000) as a
coprocessor. The advantage to this is that you gain a second address space;
useful as many of the your ideas involved mapping hardware to address spaces.
The system could be RISC based but using the '816 as an I/O processor, this
way no emulation would be done to run GS programs. Furthermore, existing
programs could gain from the presence of the RISC...in '816 mode, the RISC
could be used as a math coprocessor such as the FPE, or as a graphic processor
(i.e. by mapping the video/sound hardware to the larger address space of the
RISC, while keeping the I/O system mapped to the '816) The two processors
could communicate through an area of dual-ported RAMs (such as is done on
the Amiga's PC Bridgeboard. Imagine programming the system using a two-
processor assembler, allowing access both address spaces, both instruction
sets, and all those extra registers. Or possibly having a two-processor
system monitor that allows you to flip back and forth between processors
and use each others data.
This may be hard to do with the AMD 29k, since it uses the Harvard Architecture
(i.e. two input buses, one output bus). Both RISC's have their good and bad
points...The AMD is lightning fast but requires a sizeable, and well designed
cache (because it's so heavily pipelined), plus the non-von Neumann design,
and it it more expensive. The ARM has a small instruction set, a very
high processor-to-memory bandwidth (~18 MHz when the ARM is run at 5 MHz),
but half of its 32 registers are unavailable until there is an interrupt.
The enhanced ARM has its own cache, but still only a 26-bit address bus (gee,
only 64-megabytes).
Well, it's an idea anyway...I think there's a market for a low cost UNIX
machine.
I'm concerned that merely fixing the defencies of the Apple II may not be
enough. If someone were going to assemble a team and build such a machine,
it would have to have, at least from a marketing standpoint, an edge over
other machines in its price range. Implementing the //f would make II users
happy with better animation, more speed, and a better interface to the
Ensoniq; but how would it attract new users. The power of a RISC processor,
would give us many benefits:
1) The power to run UNIX and X-Windows in a low cost platform
(though it wouldn't be as complete a system as a NeXT, it would
probably suck away all their customers)
2) The ability to accelerate existing programs, by configuring as a
math coprocessor (accelerating SANE, or IEEE math routines)
3) It would serve as impetus to add a "industrial-strength" bus
such as NuBus (or FutureBus or Multibus II or VMEbus, or whatever)
4) It would allow graphics to be accelerated (if graphic hardware were
mapped to it) far beyond what it could with the '816 alone. RISCs
such as the ARM can manipulate memory regions faster than the '816
making your fill-mode much faster. Also, neither of the RISCs have
segmented memory.
5) It would remove the stigma that the II is old technology. Further,
it would stimulate interest in the II. People will ask "Why are they
trying to save the II? Why are they putting this heavy technology
behind what we thought was an obsolete frame?" A system so eminently
hackable as the II is what's lacking today; there are no other
computers on the market today which are as open to tinkering as is
the Apple II. (Actually the worst
thing that Mac Inc. could do to us is not to sue for copying ROM's,
but coming out with an upgraded GS to compete with our machine.)
In addition, I propose that the machine have a graphic resolution of at least
640x480 (resolutions at the Super-VGA level, i.e. 800x600, and 1024x768 are
quickly becoming commonplace, and at 640x200, the GS is quickly beginning to
look like a joke). Monitors capable of handling VGA-level resolutions are
cheap enough.
I beseech you to look at the October 1987 issue of BYTE. The article on
the Acorn Archimedes A310 is not to be missed...it contains many of the same
ideas as the Apple //f. (but this machine actually exists)
BTW, have any ideas about how we'd get around losing the SmartPort
and the Apple Desktop Bus?
------------------------------
! Mark Orr !
! $CSD211@LSUVM.SNCC.LSU.EDU !
! @LSUVM.BITNET !
------------------------------philip@utstat.uucp (Philip McDunnough) (11/24/90)
In article <9011240342.AA29693@apple.com> $CSD211@LSUVM.BITNET (Mark Orr) writes: [stuff deleted]> >Well, it's an idea anyway...I think there's a market for a low cost UNIX >machine. There already is a very low cost Unix computer available. It is based on the 386. Many companies make them.> [more stuff deleted] >I beseech you to look at the October 1987 issue of BYTE. The article on >the Acorn Archimedes A310 is not to be missed...it contains many of the same >ideas as the Apple //f. (but this machine actually exists) The Achimedes line of computers is indeed most attractive. I've used their 410 and it is one of the nicest computers you will find. The problem here is finding software for it, as we have to go through Olivetti or directly to Britain. In fact the Archimedes is a very natural successor to the GS and I would not hesitate in recommending it were it not for the software issue. Philip McDunnough University of Toronto->philip@utstat.toronto.edu [my opinions]