On Jan 30, 2022, at 2:43 PM, Chris Zach via cctalk
<cctalk at classiccmp.org> wrote:
Hm. Looking i see the dj11 had a silo but not dma.
The only thing I can see that did Async DMA was the COMM-IO-P board. That was a product
in 1980 timeframe, based on the KMC11 processor. With six DZ11's you have 48 serial
ports in a 9 board configuration.
Now for trying to unravel what the hell a DMX11 was...
The 9 board concept for 48 ports matches what I see in the Sha Tin documentation from mid
1977. Apparently the DMX11 was developed by DEC in 1975-1976 and was a product certified
for block mode devices, but the first version delivered in early 1976 was literally a
In the notes it says that DEC originally expected the DMX11 to provide block mode only
(to the cashier and ticket printing machines which operated in block mode) but signed a
contract addendum saying that the DMX microcode would be upgraded to support block and
async mode on the same controller so they could support VT52's for interactive
applications as well. Apparently this would make the DMX a better product to sell into
other environments and would be a "simple upgrade".
Turns out it was not. And this became a CSS Australia disaster that sank the whole race
Is it possible that the DMX11 was a CSS product? If so, each region or country might
develop their own products and other parts of DEC would never know it existed. About the
time you are talking about (1975-1976) I was working on a steel works railway automation
system in Sweden utilizing custom ruggedized terminals (RT04, maybe 10 keys, each 1.5? x
1?, RS232) and a pair of reed-relay switches so the terminals could be switched between a
pair of active/passive PDP-11/40?s. Until we actually moved into the custom built
machine-room, nobody realized that reed-relays would be located <50ft from a pair of
blast furnaces and, whenever they were switched on, the terminals ended up randomly
connected to one of the systems.
The following year I had moved to Maynard and I could find any information about the
project and/or the custom hardware.
Constant hardware and software patches were delivered
as DEC tried to fit the increased microcode on a single board as the HK Jockey Club held
insanely firmly to the contract requirement that the 48 port serial interface fit in a 9
board layout. Given that they couldn't change the DZ11's, they tried to mush down
the logic in the DMX11 control board and that plus the change requirement to make it
handle block mode devices and interactive devices doomed the whole project.
From what I can see, the the kmc11 was an M8204 single board which is different from the
8200 used in the dmc11. I had a DMC11 somewhere.
From the books, the kmc11 had an "lsi11 on board", 1k of 16 bit ram, 1k of 8
bit data memory a 300ns cycle time, 16 bit microprossor with a 16 bit micro-instruction
bus and 8 bit data path. This is according to the 1980 Terminal and Communications
handbook, so it's a few years after the 1976 timeframe of Sha Tin.
Now the original LSI11 processor was 4 main chips, an EIS/FIS chip (or the CIS lite chip
or the weird 1k*20 bit micro-ram board which I have somewhere). The DCT11 was a single
chip lsi11 that had an 8 or 16 bit outside bus and a 16 bit internal structure and ran
pdp11 instructions. So the KMC11 probably had the DCT11 chip.
The LSI11 chipset was around in 1975, so it makes sense that DEC could use it. The
SBC11/21 came out in 1981 but the chip was probably avail internally by 1980 so I'm
guessing that the KMC11 and the COMM-IO-DP was using the DCT11.
While, in retrospect, it may have made sense to use a DCT11 and the, later, Falcon chip
that was never the case with the KMC11. The documentation on Bitsavers indicates that the
instruction set is most definitely not PDP-11 like.
But Sha Tin was 1975-1976. Perhaps the DMX11 was an
early concept version of a KMC11 that had the original LSI11 chip set of 4 chips running
the show with a set of six DZ11's and a smaller amount of memory. But there wasn't
enough space for additional memory to handle the extra features and DEC got stuck. Maybe
part of that mushing work led to the LSI11 being fewer chips (the PDT11 has only 4 chips
but EIS/FIS thanks to one of the chips being a dual carrier of MicroRoms) and ultimately
to the DCT11.
One item that could really help would be RSX11D version notes. According to the
documentation the DMX11 was supported in RSX11/D but support was dropped in 1976. They
switched to RSX11/M where it was still supported.
RSX-11D V6.2 was released in June 1975. This was the last release of RSX-11D as a separate
product, subsequently it was merged with IAS with an option to generate a ?real-time?
system from a common distribution.