S-100 tech Summary (was Re: This is new...)

From: Jack Peacock <peacock_at_simconv.com>
Date: Wed Apr 15 02:34:32 1998

>I am completely unfamiliar with S-100 systems, so could you
please
>explain... were S-100 technologically superior to PCs (i.e. IBM
>PC 5150), or just aesthetically? As far as I know, they used an
>older processor...

S-100s started out with an Intel 8080, then Zilog Z80. That was
pretty much the end of the 8-bit version, although there were
several other variants (8085, NSC 800, Hitachi 64180, Signetics
2650, etc.). Hmm, I never recall seeing a 6502 or 6800 CPU
board. Once the S-100 was standardized around the IEEE-696
specifications the good 16 bit systems started appearing. One
of the earliest was the 8086 by Seattle Computer Labs (IIRC).
Famous because it came with the ancestor of MS-DOS, which
Microsoft later bought for their IBM PC deal. You might say all
the PC software now in existence owes its existence to the
S-100....

Anyway, there were quite a few 16-bit systems. Many 80286
variants, including a nice one by Macrotech, a dual Z80H/80286
CPU. There were several 68000s, TI 9900, National 32000, Zilog
Z8000, and a variant on the Western Digital LSI-11 bit slice
chipset called Alpha Micro. The AM ran its own multi-user
operating system, very reminicent of RT-11 (there was a strong
DEC influence there).

Was the S-100 technically superior? Well, if you go by the
IEEE-696 specifications then compared to either the XT or AT it
was quite a bit better. Maximum memory was 16MB (24 bit
address, 8 or 16 bit data path), 8 interrupt levels (open
collector! which meant multiple boards could use the same IRQ),
16 DMA levels (still better than the 7 DMA levels on the current
PC). Most 286 based systems ran up to 8Mhz reliably, some made
it to 10 or 12 Mhz (compared to the original AT at 6Mhz). Even
better, it was common practice to use static RAM memory on the
better business systems. If that term doesn't sound familiar,
cache memory on modern PCs is static RAM. The significance is
that there were no wait states or lost cycles to refresh. Think
of how fast your Pentium would run today if all 64MB of RAM was
cache, not DRAM.

S-100s were also very expandable. Motherboards usually had
between 18 and 22 slots for full sized machines. You could put
a lot of RAM, serial, and disk controllers in that many card
slots. I built custom 286 based multi-user systems that
supported 10 or more users running production business work.
The response time compared quite favorably to contemporary low
end DEC PDP-11s, and for a fraction of the cost.

S-100s were also early adopters for much of the current crop of
PC peripherals. Networking, using ARCnet at 2.5Mbps over coax.
Digital Research supported network access to disk drives using
CP/NET on top of MP/M II. No, it wasn't TCP/IP, but it still
compares quite well to a basic Netware system. Disk drives,
both the 5.25" floppy and the 5.25" hard drive showed up on
S-100s before PCs. S-100s using MP/M II could support disk
drives up to 512MB, long before MS fixed the 32MB barrier in the
XT and AT.

The S-100 did have some drawbacks, mostly from the weird control
signals the CPU had to generate (anyone remember how difficult
it was to simulate a PSYNC on a non-8080 processor?). It
suffered terribly from early failures to standardize the bus.
Many of the 8-bit systems had unsolvable compatibilty problems.
(on the other hand, it did make for some extra pocket change for
struggling college students who knew how to tune an S-100 to
make everything work :) )

Another problem was the unregulated power supply. Unregulated
+8 and +/-16VDC was run over the bus itself, right next to
signal lines. Every board required it's own regulation, which
could take 20% or more of the board space, as well as being a
nightmare to keep cool. If you see early pictures of loaded
IMSAIs, the cover was always off. This was a necessity, the
heat was too much with the cover on. I had to use a 16" fan to
keep mine running with 64KB of 2102 based static RAM (not 21L02s
BTW, they cost more than the fan did).

The single worst problem was the absolute lack of any hardware
standardization beyond the 696 specs. There were no I/O
addresses for anything. One manufacturer might use a WD 1791
floppy controller at port 7xH, another would use the NEC 765
floppy controller at address 9xH. There were no standardized
BIOS ROMs either. Systems came with some basic boot ROM for one
particular disk controller/serial interface, and that was it,
everything else was supported by drivers in the machine specific
version of the OS. A boot disk for an Altos wouldn't run on a
North Star, not even close. For those who think it an evil that
the world has standardized on the Wintel architecture for PCs,
trust me, the other choice is far worse.

>Was it just an issue of being used to them?

Sure, but then if you wanted a good hardware oriented micro with
lots of support from 3rd parties, in the mid 70s the S-100 was
the only choice. You could get a board to do just about
anything, tho you had to program it yourself. Nearly all boards
came with schematics, if you didn't like the design or it wasn't
quite compatible, you could cut traces and rewire to your own
choosing. My own IMSAI is far from a standard out of the box
model.

>As for laissez-faire, I never have believed in it. It makes
society
>too concerned about money. This is proven when complete crap
hardware
>is released now, and people don't care because it's good
_financially_


You don't like the profit motive? *gasp* That's, well, that's
just plain un-american (understandable and excusable if you
happen to be european tho).
    Jack "show me the money" Peacock
Received on Wed Apr 15 1998 - 02:34:32 BST