I have an AT&T PC 6300 (Rebadged Olivetti M24) with some 16-bit slots in it, proprietary of course. The 16-bit extensions on the slots are literally ALLLL the way on the opposite side of the motherboard from the rest of the ISA slots. 🤣

Awesome machine! Thanks a lot for the picture! 😁

Jo22 wrote:

Hi, just seen an interesting mainboard at eBay..: […]
Show full quote

Hi, just seen an interesting mainboard at eBay..:

- Junior XT -
http://www.ebay.com/itm/292052372176

Apparently, this is an XT-class machine with 16bit (ie, 16bit wide) "ISA" PC slots.
I knew that these things exist (Olivetti, etc), but I've never seen such a mainboard in IBM PC form factor.

Also fascinating is the CGA card - it's unbelievable that they also existed with a 16bit connetion! 😳
I always thought CGA was only ever made for 8bit slots..

I'm curious, has/had got anyone of you such a machine ?
If so, did these slots follow some kind of standard (I mean, were they used by other manufacturers, too ?)

Weird machine. It's a Soviet-era clone so they had no incentive to follow the "western" way of building them. The CPU is an AMD 8086 so fully 16-bits wide; no surprise whoever designed it wanted to take advantage of that. Wouldn't have expected a genuine AMD CPU & support chips (at least one, to its right) on a Soviet clone board, unless they're replacements (or rebadges.)

Thanks for your reply, xjas! 😀
I also heard of XT machines having the real thing (8086) and torindkflt posted a wonderful photo about one of them..
If memory serves, my father even used an 8086 machine himself (a PC1512, I believe) when he wrote software way back in the 80s..
I'm surprised, though, that eastern people spent so much effort into improving that ageing architecture.
Someone would think they would have rather built systems around 286 or 80186 processors.
So I wonder, why was it (the 808x series) so popular over there ?

As I recall the days of the cold war, it was more a question of people inside USSR and other communist countries, having a hard time getting access to western produced technology. You had to work high up in the circles in order to be able to work with this stuff. I think it was a mix of trade embargo, price and political ideologies that were getting in the way of "just" picking up something like a 186 or 286 CPU. Things were a bit different back then. Things were no near the way they are today. Getting different stuff into Russia or China, would sometimes result in an execution or a 20 year "holliday" in Siberia. Back then, it was like sneaking stuff into North Korea today.

That's super cool. Somebody should develop a LIM 4.0 EMS card and VGA adapter to take advantage of those special 16-bit slots.

I was actually really hoping the guy who designed that new 8088/86 passive backplane kit (Sergey?)would have tried to hack 16-bit AT slots onto it. Although many adapters would not work due to the missing signalling, if special boards could be developed there would be a fairly significant speed boost. I really regret that my JUKO 8086 motherboard for example can only access the memory on my EMS card over the 8-bit bus...the onboard memory is considerably faster.

I would absolutely love to yak with some former iron curtain developers / engineers over beers sometime. The lengths they must have gone to make those machines functional is astounding. Things like cloning smuggled western chips from scratch, or writing down-to-the-metal operating code for architecture that might have had a production run in the hundreds... Insanity. It's no wonder why so many hackers (good or bad) come from Russia & eastern Europe now.

Sorry to ressurect an old thread, but I have some additional info.
Junior XT was a Romanian, NOT Soviet clone of PC XT; indeed, back then in the late 1980s Romania was a communist country.

The low-end PS/2 models (1987) used 8086 CPUs but had no expansion ports at all.

fillosaurus wrote:

Sorry to ressurect an old thread, but I have some additional info.

Nah, that's fine to me. New information is always welcome. Don't worry. 😀

fillosaurus wrote:

Junior XT was a Romanian, NOT Soviet clone of PC XT; indeed, back then in the late 1980s Romania was a communist country.

Ah, I see. Thanks for the correction. I went with the information I saw on eBay and didn't know that.

xjas wrote:

I would absolutely love to yak with some former iron curtain developers / engineers over beers sometime. The lengths they must h […]
Show full quote

I would absolutely love to yak with some former iron curtain developers / engineers over beers sometime.
The lengths they must have gone to make those machines functional is astounding. Things like cloning smuggled western chips from
scratch, or writing down-to-the-metal operating code for architecture that might have had a production run in the hundreds...
Insanity. It's no wonder why so many hackers (good or bad) come from Russia & eastern Europe now.

I'd love to see/hear their point of view, too. 😀 But I wonder why they would have had to lower themselves by copying western tech,
when the much better asian/chinese/japanese tech was "just around the corner" and could have had been provided by their neighbours discreetly.
A japanese V20/V30 was much, much better than a stinky 8088 from intel (by comparison; 808x hadn't even got a dedicated unit for
effective address computation. It had to use the ALU for that. 286+ fixed that design flaw).

Errius wrote:

The low-end PS/2 models (1987) used 8086 CPUs but had no expansion ports at all.

The Tandy 1000 SL also uses an 8086, and has ISA slots, but they are 8-bit only.

Jo22 wrote:

A japanese V20/V30 was much, much better than a stinky 8088 from intel (by comparison; 808x hadn't even got a dedicated unit for
effective address computation. It had to use the ALU for that. 286+ fixed that design flaw).

That is not an entirely fair comparison. The 8086 dates from 1978. The V20 wasn't released until 1982, so could take advantage of more modern technology, and use more transistors to improve performance (I wouldn't call it a design flaw, just a result of the limited technology available when the 8086 was designed, a compromise).
The 286 was also introduced in 1982, and was far superior to the V20 and V30 CPUs.

Back in '94 I had a russian XT clone, called ISKRA 1030.11. Since it came with complete schematics (and it used a 8086 clone), I can definitely tell it used a 16-bit bus, but the connectors were completely different. All the boards (CPU, RAM, CGA, MFM+Floppy ctrl, serial) were fitted in a rack onto a passive backplane containing those strange connectors. I remember trying to adapt a standard ISA VGA, it was a couple of days' work... It didn't POST, but amazingly I managed not to fry anything.
Most of the chips were russian clones, but a few of them were clearly made by Intel and had the original markings grinded away, black lacquer applied and onto that new markings written by hand. Such were the times...

Gallery (not mine, since it's long gone):

Iskra 1030.11

Scali wrote:

That is not an entirely fair comparison. The 8086 dates from 1978. The V20 wasn't released until 1982, so could take advantage […]
Show full quote

Jo22 wrote:

A japanese V20/V30 was much, much better than a stinky 8088 from intel (by comparison; 808x hadn't even got a dedicated unit for
effective address computation. It had to use the ALU for that. 286+ fixed that design flaw).

That is not an entirely fair comparison. The 8086 dates from 1978. The V20 wasn't released until 1982,
so could take advantage of more modern technology, and use more transistors to improve performance
(I wouldn't call it a design flaw, just a result of the limited technology available when the 8086 was designed, a compromise).
The 286 was also introduced in 1982, and was far superior to the V20 and V30 CPUs.

I apologize. 🙁 I really should have had worded it differently. I was rather thinking from a mid-1980s perspective.
I was assuming that the eastern countries weren't bound so much to western standards and could have gone their own way in certain respects
(such as expansion slots, code pages, disk formats, BIOS, memory layout). It wasn't until early 90s when they got a proper MS-DOS (V4.x?), anyway.

Of course they did use DOS software before, but plain CP437 textmode software such as Norton Commander, Turbo Pascal or dBase
(-NC apparently was very popular over there; it was claimed that no russian/eastern DOS user ever saw DOS without NC-)
did run on various DOS compatibles all over the world. Be it something like a Sirius 1, PC-9801,
Robotron 1750, an Apple II with 8086 card or a BBC Master 512 with 80186 board and DOS Plus.

On the other hand, the competitor of the 8086 predcessor (->8080), the Zilog 80, also was from 1970s (~76) but could do more
than the 8080 (~74) or even 8088 in that respect. Unlike the 8080, it was able to perform "single-bit addressing,
shifts/rotates on memory and registers other than the accumulator, rotate instructions for BCD number strings in memory, program looping,
program counter relative jumps, block copy, block input/output (I/O), and byte search instructions." (source)

Edit: Never mind, I didn't mean to criticize the 8086/88 in first place. I rather wondered why they tried so badly to copy western tech,
when they got better CPU designs in their own countries or their nearest neighborhood (for example, a NEC µPD7220 was much
more sophisticated than the Motorola 6845). That's why I would like to hear some interviews with ex-developers of these countries.
It would be interesting to hear who had the idea to built that Junior XT with 16-Bit slots, for example.
And which background story was related to that (which application needed the extra memory performance, etc.)

Jo22 wrote:

On the other hand, the competitor of the 8086 predcessor (->8080), the Zilog 80, also was from 1970s (~76) but could do more tha […]
Show full quote

On the other hand, the competitor of the 8086 predcessor (->8080), the Zilog 80, also was from 1970s (~76) but could do more
than the 8080 (~74) or even 8088 in that respect. Unlike the 8080, it was able to perform "single-bit addressing,
shifts/rotates on memory and registers other than the accumulator, rotate instructions for BCD number strings in memory, program looping,
program counter relative jumps, block copy, block input/output (I/O), and byte search instructions." (source)

The Z80 was/is considered as one of the best 8-bit CPUs of the era.
The 8086 however is a 16-bit CPU, so Intel decided to spend their transistors in a different way.
I think what's interesting is that the limit is clearly the technology of those days. Especially if you look at the 8088, the CPU has quite a powerful 16-bit instructionset, and a decent amount of registers etc. But its raw performance is still in the ballpark of the Z80 and 6502. They are all limited by the memory technology of the day.

I suppose the CPU from that era that really stood out was the Motorola 68000. However, it was a lot larger and more expensive than the 8086.

As for the clones... my guess is that it depends on what they had available, how long it took to clone and debug a CPU, and the limits of their own manufacturing.
I'm quite sure there were also communist Z80 clones (there are quite a few Russian sceners even today, who target ZX Spectrum-like machines. I believe this is because ZX clones were popular back in the day).

Speaking of limits of manufacturing, that is one reason why Commodore's Amiga died: Commodore had bought MOS in the 1970s, so they had their own chip manufacturing facilities. This had been an advantage for their VIC-20, C64 and original Amiga machines, because they could develop their own custom chips.
But after the original Amiga, their manufacturing technology had fallen behind, and they ran into manufacturing limits for future chipsets: clockspeeds, power consumption, max nr of transistors etc.

Several months back on vcfed, somebody posted a photo of a board with an 80186 CPU with 16-bit ISA slots. Almost all the pins were said to be connected to where they would go on an AT connector. I don't know if the board was ever powered up and tested, but it would certainly be interesting to follow up on that.

The 80186 board was also interesting in that it apparently did not use the integrated crap which typically makes 80186 systems less than 100% PC compatible, and isntead used discrete controllers like on the 8088/86.

That's interesting! Thanks for mentioning that! I heard something similar too, but can't remember where I got it from.
Apparently, PC makers of the mid-1980 were satisfied with the improved 80186 core, but not the "extras".
So they ended up adding some discrete pieces like DMA controllers (?) or timer chips.

That beeing said, I'm still learning about the PC/XT platform, so anyone feel free to correct me. 😀
I'd be interesting to know which market value the iBM PC had when the 80186 was made or released.
Since DOS-compatible PCs (not necessarily IBM compatible PCs) were said to be the future, this could explain
the improved, but slightly incompatible extensions to the 8086 eco system.
From what I read at redhill page, the 386 was in development before the 286 was released,
which in turn also draws the question whether the 80186 or 80286 was completed first.

Edit:@Scali Thanks for the explanation, too. 😎 The reason why I mentioned the Z80 was that it is a "cousin" of x86 family
(NEC V20/30/.. had some sort of emulation mode of the basic 8080 instruction set, as we know).
I also agree with the 68K, btw. To me, 68000/68010 are on par with the 286 as far as features go (MMU, privileges, adress space, ..)

Jo22 wrote:

Apparently, PC makers of the mid-1980 were satisfied with the improved 80186 core, but not the "extras".
So they ended up adding some discrete pieces like DMA controllers (?) or timer chips.

I suppose it depends on when the machine was designed.
Some very early x86 machines were mainly designed to run an OEM version of DOS.
The Philips : YES is an interesting example of a computer designed like that: it uses the 80186, and has its own graphics standard: https://en.wikipedia.org/wiki/Philips_:YES
It simply uses all the onboard logic of the 186, and doesn't aim to be hardware or BIOS-compatible with a PC.
But clone makers quickly found out that killer PC apps such as Visicalc required more than just DOS and an x86 CPU (and ironically enough even IBM fell into that trap with the PCjr). That made the 186 a very uninteresting option.
Apparently some manufacturers still figured the 186 may be an interesting option, even if the onboard logic is bypassed, and conventional chips are added on the board. I wonder how compatible these machines are.
But the NEC V20/V30 CPUs were a better option, because they gave you the same benefits as the 186 (extended instructionset, improved performance) without any of the drawbacks... they were pin-compatible with 8088 or 8086 motherboards.

I think the same. Sometimes I wonder what would have happend if an 8086 clone with 286 instuction set (incl. P-Mode)
and 64 to 256KB of built-in extended memory was released. Some sort upgrade chip, like these later MakeIt486 or 486DLC chips. 😀