Jon Elson via cctalk wrote:
[..]
> >
> > Later I could try to build something with the 49C402, AM29116 or WS59032
> > to save PCB space.. All that stuff is already in my stash of parts,
> > besides many 2901,2903,2909,11,10 etc...
> > Maybe I'm trying to get that PCB with the four AM29203 and AM2910 from
> > GE to life first.
> >
> The 2903 was a VERY nice bit slice chip, it could do single-cycle multiply
> and divide, as it had an extra shift register for the multiplier bits.
...and you can expand the Registers externally with AM29705 (AM29707)
> The
> 29203 is pin compatible, but faster. Once I got the whole system running, I
> planned to upgrade to the 29203. I had planned to make device controllers
> with 4 74S181s and a dual-port ram chip for registers, and the control store
> would be 27C32 EPROMs of the fastest grade. I was hoping to get 4 MIPS from
> these.
The 49C402 has an expanded Register set (64 words instead of 16), an double
set of ALU Commands. Don't know if it is the same as on the 2903/203. It
contains
the carry readahed circuity too and is 60% faster IDT says...
The 27C32 would be the bottleneck. I've bought some used AM27C291 in the
past (and compatible Parts from Cypress and TI).
I've pesteres the german company Conitec to implement device support for
those for ther GALEP Devices. Send 10 chips to them and got a Mail later
that they would need more..they where noe all defective now...
The Read Support was added for the GALEP III and IV ..but I'm would'nt
pay additionally and finally build a prommer fort those differential
Eproms myself. I'ts working and no Chip hat to die for it...
But even with those Eproms (25..55ns) it would be impractical to build
an control store for a single experimental device out of it. I've build
an control store out of old cache RAMs (8kx8). (A friend of mine collected
many of them for me..out of old 486 Mainboards) An Atmega16 with an 4Mbit
Atmel Dataflash could store the Microcode and load it in the WCS.
There are still some Pipeline Registers to solder in and finally to write
the Software for the AVR. ..it's laying around for years now. :-|
This all was before I knew about IDT71502 WCS Chips that include serial
shift registers..
We had an Company here in Germany that sold used chips from recycled pcbs
one..the new management endet this. But before it ended I've bought a
bunch of them, besides things like J11, Bus drivers AM29828 and much
other things for a very small amount of bucks.
(From the point I've told them that I wanted the chips to build something out of
it and I'm not interrested to only look at nice gold ceramic Chips w/o
any scratches..it seems that they send me the best looking Ic's they
had.. :-) )
>
> Of course, now this could all be built on a few FPGAs, and get vastly higher
> performance.
That's the point where the fun ends..at least for me. Yes, I've done
some designs on Altera CPLDs (schematic entry only), bought a book about
Verilog and one about VHDL .. but spare time is low...
I don't even want to build a "fast" computer ...nice if it's fast, but
it isn't necessary.
I'm sitting on a PC with an Ryzen 5-3600 and 64Gbyte RAM (no, no Flash
disk, but a RAID arry of fast SAS "spinning rust" disks..)
>
> Yes, a small peripheral controller would be a much more practical project.
>
> Jon
Exactly.
First is to get something running..
In the old AM29xx Docs (Donnamaye E. White) are even "controllers"
described that don't use any ALU at all, but an AM2910 only.
As long you don't have to shuffle some data, this will be sufficient..
(Coffee machine?)
Regards,
Holm
--
Technik Service u. Handel Tiffe, www.tsht.de, Holm Tiffe,
Goethestrasse 15, 09569 Oederan, USt-Id: DE253710583
i...@tsht.de Tel +49 37292 709778 Mobil: 0172 8790 741
