Why did DOS/86 overtake CP/M-z80 ?

[cj7hawk]

Found them... Interesting... From 1988.

A column on "upgrading" for die-hard Kaypro users.

Situation 13: "I'm familiar with MS-DOS and its associated
software, hardware, etc. But I don't need bells and whistles and
14 megabyte spreadsheets; in fact, my extensive CP/M software
library is quite adequate and I never seem to find a project I
can't do with it. Besides, I have untold hours of experience
with CP/M and numerous data disks in that format. I have a
Kaypro 4-84 with one DSDD drive and a quad drive. Is there some
way to soup up the machine to perform near AT standards?"

Recommendation 13: Good question. The new Z280 chip is
roughly equivalent to the Intel 80286 in performance and capa-
bility. (If anybody wants to quibble I'll merely fall back on
the operant "roughly" - so save your time!) The capabilities of
the Zedux board are still unknown to me, but the Ultraboard
would satisfy your needs. The basic board with one meg of RAM
plus the battery-backup for the RAM will cost about $650 (check
with High Tech). Then each additional meg of RAM will cost about
$100. With, say, five megs of RAM (put application programs and
often used files in RAM; and data mainly on disks) you'd have
one spiffy machine.
You can speed up your machine in other ways not involving
Ultraboard, such as a hard disk and/or RAMdisk, but these routes
do nothing about improving CPU SPeed and performance. further-
more, by going the Ultraboard route you can load the board with
enough RAM chips to give yourself a permanently powered RAMdisk
that the Z280 can directly access and, in the process, probably
eliminate the need for a hard disk.
The catch here is cost. The basic board with backup and
four additional megs of RAM will cost about $1100. What kind of
005 system you can buy for the same money, and the caPabilities
of that system, would depend on your shopping ability. So I
presume your need/desire for high performance, and your willing-
ness to stick with CP/fvl software, is worth $1100 to you!

So I guess these conversations are just echoes of the past that reflect the current. Interesting that they thought this might continue... This was the final edition of this newsletter too.

 

[alank2]

You can only fight the tide for so long, when it gets stronger and stronger.

 

[Chuck(G)]

Of course. And who would want to lug a "portable" Keypro around today?

 

[hairyvisionary]

Could I trouble you to provide a link to the scans? I'd love to read them.

I see cj7hawk found 'em, but they're in:

http://bitsavers.org/pdf/morrow/bamdua/
http://bitsavers.org/pdf/morrow/morrow_owners_review/

 

[Eudimorphodon]

Recommendation 13: Good question. The new Z280 chip is
roughly equivalent to the Intel 80286 in performance and capa-
bility. (If anybody wants to quibble I'll merely fall back on
the operant "roughly" - so save your time!) The capabilities of
the Zedux board are still unknown to me, but the Ultraboard
would satisfy your needs.

I'm curious if the person responding with this had actually touched the board, or was just rattling off promises. Did some digging around for other mentions of this board, and I haven't been able to find an actual review, just vaporware mentions.

It's always interesting to read through the archives of magazines and newsletters documenting the twilight years of any computer platform. It's kind of remarkable just how long the afterlife of even some of the earliest computer platforms were, and as you move forward through the archives you'll usually see the same mix of defiance, resignation, delusion, and even guilt as people adjust to the reality that there just isn't any future for the machine they'd invested so much time and money into over the years. The guilt in particular is... sad, but understandable? It's a perfectly human quirk to attribute feelings of loyalty to inanimate objects like vehicles, tools... and computers. Saying goodbye is hard even if the thing you're trading it in for is objectively better in every single way... and with switching computers there's always the huge hassle of porting your data, learning new software. etc. CP/M might have been a pretty least common denominator computer, but spend enough time with something, learning the ins and outs of it... there are going to be some loyalists that try to stick with anything until the bitter end.

Not coincidentally, near the end of almost any 8-bit platform's life, you'd have a few little diehard operations advertising in these shrinking magazines some absolutely crazy accelerator and memory cards that offered to cram a few megabytes of memory and quintuple the CPU speed into machines that barely knew what to do with 64K. Heck, they might even provide patches for a handful of popular, but old, programs that could *kind* of let you use it, although typically you'd be lucky if you got much more than a RAMdisk and vague promises for an enhanced OS "coming soon". Sometimes they'd even manage to sell a few hundred units. But, obviously, the existence of these cards was meaningless in the grand scheme of things. Without software to really use the thing and otherwise being hamstrung by the remaining fundmental limits of the platform they were novelties unless you really had a specific use case for it, and upgrading an old machine with one plus a hard disk and other parts you'd need to *seriously* make any kind of comparison to a rotgut PC clone would be a financial disaster. Nonetheless their existence was often fuel for the true believers to renew their vows to never quit their computer until it's pried from their cold dead hands; maybe evenTHE PROOF that their platform is destined to rise like a phoenix from the ashes!

... Huh. I guess maybe we can't even limit this to 8-bit computers, because, well, Amiga exists. doesn't it? Some of those guys still act like they're in their island bunkers refusing to acknowledge the thirty year old surrender order even today. I assume at this point it's all performative, but I guess it's hard to know for sure...

 

[NeXT]

Of course. And who would want to lug a "portable" Kaypro around today?

I'll carry around my Osborne and its battery pack, if not to spite people at Starbucks while using Wordstar.

 

[Chuck(G)]

You might as well carry around a bunch of wet clay and a stylus.

 

[tradde]

Maybe he does that too?

 

[cj7hawk]

I'm curious if the person responding with this had actually touched the board, or was just rattling off promises. Did some digging around for other mentions of this board, and I haven't been able to find an actual review, just vaporware mentions.

It wasn't just the Kaypro diehard enthusiasts... Xerox users wanted it as well.

http://www.retroarchive.org/cpm/cdrom/CPM/XEROX/ULTRA-XE.INF

The Z280 ULTRABOARD is High Tech Research's ultimate add-on
board for all '84 Series CP/M Kaypros (including the 10-83.)

ULTRABOARD will turn your present machine into a revolutionary
new CP/M computer, with twice the processing speed and with 16
times the RAM capacity of an IBM-AT.

With the ULTRABOARD installed, you can immediately
- turn your Kaypro loose with a processing speed of 12MHz, up
from its' current 4MHz!
- address up to 16 megabytes of RAM!
- choose your own foreground and background hues on an
external RGB color monitor!

The High Tech Research ULTRABOARD is now in beta-testing for a
Summer release. CP/M is back, with a vengeance.


Take that you foul skanky 386 - ( yeah, they really did make that kind of comparison in the article. )

But as it notes, It's in "Beta Testing" - So most likely never went into production. Or they got it into production then discovered the distribution channels were now limitted to one user event per year for a system with no local support. By this time, PC shops were appearing all over the world like a newly formed cancer and spreading clones like crazy. And the Ultraboards rotted away in a shop somewhere waiting for a market that no longer new how to find them.

So it seems like the best answer to my question isn't that these things didn't compete. They clearly did - or at least the immune system of competition tried to respond.

But x86 came in like a tsunami and just covered the landscape leaving nothing but ruin and waste for everything else. There was no competition, because there was no real awareness of options - the typical consumer could name a few low-end home computers and then there were IBM PCs. Like magic, only with 4 color graphics.

As you note, software simply wasn't designed to use this extra memory, and DRI was AWOL flying planes - There's a lot of reason DRI could never have succeeded when Microsoft failed. Microsoft knew how to keep people coming back to the platform... How to standardise everything so our systems grew with us rather than being discarded the way an arthropod discards it's exoskeleton when it grows too large for it's body. DRI produced a product but Microsoft was a leader in every sense of the term.

Software vendors flocked to the PC. DOS allowed for 640K and back then at least, it really was enough for anyone. Between IBM standardising graphics modes all the way up to VGA, the emergence of VESA modes and standard memory access models for extended memory supported by Microsoft, along with early releases of Windows and pretty soon Windows 3.0 came out and then we got Wolverine and Trumpet and Netscape.

The operating system really did make up the difference, and since everyone was pirating the same operating system at a time when the second computer revolution was occuring, everything was standard and even idiots in large numbers can affect world-shattering change.

But aside from feeling like their support was cut off, for most people, for most of their work, 64K was still enough. Wordstar still worked, disks were fast enough and no one needed to upgrade. When the time came, there was only one choice and it wasn't the Ultraboard These old systems got stored in back sheds and now I'm buying them.

If fairness, Apple did die, then rose like a zombie from the grave somehow turning the world on it's head. In my fantasies, there's a cosmic event any everything in the nanometer range of computer chips become unreliable and we're all forced back to 8 bit machines once again Then I wake up and write this on a machine with 12 cores running at 4.8 GHz, before playing some games on my Meta Quest 3.

If there's one thing I learnt from the demise of the Sinclair Spectrum - Root for the underdog, but move to the side of progress. Did the 286 seduce me? Heck, I assembled it right in front of my Spectrum after shoving it off of the desk to make room. I too was disappointed with how they added memory to later 128K spectrums, and programmers never used the extra memory. At best, it was a useless RAM disk lacking capacity. What I really wanted was software that made use of this space. What I got was a 48K spectrum with better sound and a RAM disk, that software doesn't even use.

When my PC started to run out of memory, there was HIMEM.SYS and later EMM386.SYS and pretty soon it was just Windows. Which is why I still have a PC on my desktop today.

The 86 didn't compete with the z80. The z80 just didn't grow well out of the exoskeleton that CP/M had created for it and it's rival simply grew too quickly. DOS didn't replace CP/M because there was no relevance for it to do so. DOS was just a means to an end. What it did do though is prevented anything from doing to the PC what time did to the z80 systems of it's era. Between IBM and Microsoft and Intel, they really did start something incredible. Moving away from backwards compatability with the 8080 was probably a big risk for IBM. Had it gone with CP/M, I think that we'd be using a different architecture today.

 

[cj7hawk]

I'll carry around my Osborne and its battery pack, if not to spite people at Starbucks while using Wordstar.

You know you're going to start a trend if you start doing it at Starbucks.... Pretty soon none of us will be able to afford an old Osborne.

 

[Eudimorphodon]

And the Ultraboards rotted away in a shop somewhere waiting for a market that no longer new how to find them.

I’m pretty confident that the number of “Ultraboards” ever made, if any, coincides with how many Z280s Zilog would give you for free if you requested samples. There’s no way this was ever anything but a pipe dream.

 

[krebizfan]

Tillman Reh's article on a Z-280 board appeared in 1991 and the related kit seems about the only Z-280 anything that actually existed in numbers. Looks like another case where Zilog's marketing was trying to sell a CPU before the engineers got it working.

 

[Pepinno]

The real reason is in the apps. "Desktop" apps (like Visicalc, Wordstar, etc.) for CP/M were not binary-portable from one CP/M system to another, mainly because there was no common video standard for the CP/M platform (also, no common serial port standard, etc.). Therefore, software vendors had to publish their apps in a variety of formats/targets, which meant they had to own every CP/M system and its SDK for which they were targeting their apps. The IBM-PC platform solved this problem which software vendors had, and software vendors jumped ship to the IBM-PC platform en masse.

I know because I have an Amstrad CPC6128 with a z80 CPU, with which Amstrad shipped CP/M 2.2 and CP/M 3.0 (a..k.a. CP/M Plus) diskettes. I cannot grab from the Internet any CP/M-80 version of Visical or Wordstart or Kermit, because they won't run properly on my Amstrad. Instead, I need to find and grab versions of those programs specifically targeting the Amstrad CPC family of computers.

So it is not that MS-DOS overtook CP/M, but that the IBM-PC single-plataform, with its army of clones, overtook the fragmented CP/M-80 platforms. The official OS for the IBM-PC platform was DOS, therefore DOS displaced CP/M as the majority of the people just run the OS supplied by their computer's OEM.

 

[Chuck(G)]

Wordstar and Kermit are customizable by the user--there was no secret about that; for example, the "patch area" in WS is well documented in Micropro's user manual. Kermit uses "overlays" and the documentation describes those. I can't speak to Visicalc, as I never used it--always Multiplan. Supercalc or whatever.

 

[hairyvisionary]

The real reason is in the apps. "Desktop" apps (like Visicalc, Wordstar, etc.) for CP/M were not binary-portable from one CP/M system to another, mainly because there was no common video standard for the CP/M platform (also, no common serial port standard, etc.). Therefore, software vendors had to publish their apps in a variety of formats/targets, which meant they had to own every CP/M system and its SDK for which they were targeting their apps. The IBM-PC platform solved this problem which software vendors had, and software vendors jumped ship to the IBM-PC platform en masse.

In general, CP/M provided a "console" abstraction via its BIOS but you might have any kind of serial terminal or a memory-mapped video console connected to it. The IBM PC wasn't that much different but limited the choice of consoles to MDA, CGA, then Hercules-extended MDA and EGA.

In USAnia, the CP/M system vendors were, um, solving this problem in their own way. That was Adam Osborne's real innovation, not only did Osborne have a portable CP/M system, it shipped with a bundle of CP/M and applications that included a word processor and a spreadsheet and a database, all already ported to the Osborne 1. NLS did similar deals with the Kaypro (though the one I saw had shipped with Perfect Writer instead of Wordstar), and George Morrow put this idea back on the desktop with the Micro Decision. So your choice of hardware might be influenced by the software bundle.

 

[Eudimorphodon]

Moving away from backwards compatability with the 8080 was probably a big risk for IBM. Had it gone with CP/M, I think that we'd be using a different architecture today.

FWIW, here’s another place where the “Z80 = CP/M” framework you’ve adopted here breaks down. ”Proper” portable CP/M software isn’t written for the Z80, it targets the 8080, and while a few programs violated that rule late in the life of the platform for the most part it stuck. (I’m not going to count here programs that were really written specifically for Amstrad CPCs or MSX or whatever because those are obviously *not* portable CP/M programs.)

The Z80 is 8080 binary backwards compatible. Great. But it’s also not really better than an 8080 unless you take advantage of its enhancements and then, BAM, you’re not actually writing according-to-Hoyle CP/M programs anymore, are you? The 8088 is *not* binary compatible… but Intel went to significant lengths to make it Assembly *source* compatible. If you machine translate 8080 source to x86 the results aren’t as good as if you sat down and rewrote it from scratch, sure, but for a quick and dirty start of a port *it worked mostly fine*. So… this seems obvious, doesn’t it? If enhancing for Z80 breaks backwards compatibility for new programs for your computer with the legacy platform it’s supposedly compatible with, why not just rip the bandaid off and go to an even more capable CPU that makes it easy enough to bring stuff forward that a lot of software just needs to be reassembled? Because of the nature of CP/M moving to a new computer always needed patching for the video terminal and stuff anyway…

I dunno, it’s almost like Intel knew what they were doing or something.

 

[DeltaDon]

I ran a small business with two CP/M 2.2 S-100 computers. They did the job, but only because I had all the software I needed at the time. DBase II, SuperCalc and WordStar and a few more were the main programs I used. I could see the issues that was killing not only CP/M, but also the S-100 bus computers. First off, the only "standard" disk formats that software was available on were the IBM std. 8" disks and the Osborne 5.25" disks. If you had some other disk format, you had a problem, with a few exceptions. Then there was the lack of a "standard" CP/M that would plug and play on anything but the system it was designed to use. This was a problem due not only cost, but the sudden failure of your computer maker. Then there was the problem of video. Not just the lack of graphics, but the need to customize each program to work with your CRT terminal of choice. No such thing as a standard for reverse video, underlining or blinking. Need color? Good luck!

On the PC side of things there wasn't as many issues. Once the clones arrived and the same software/hardware could be used in them, that was the end of CP/M for even holdouts like me. Standardization wins. Features, didn't matter. My 6MHz Z80 with a large RAM disk ran circles around early PC's when processing a database.

Interest as a side note, my first laptop was also a CP/M machine. Now this was very early on, mind you. I had a NEC Starlet that ran embedded CP/M, had a static CPU that allowed instant boot to where you left off plus it ran on "C" cells that could be purchased at any local 7-11. The B&W LCD was bright and easy to read in the sun. If came with Word*, Calc* and other software in ROM. Limited as all heck, but it worked. My first 486 DOS laptop was a PITA to use while traveling between customers. The screen was hard to read in the sun and it took minutes to boot. To save the battery there was a lot of shuting it down and waiting to reboot. Lost time, lost money.

 

[lowen]

I'm curious if the person responding with this [information about Z280 boards] had actually touched the board, or was just rattling off promises. Did some digging around for other mentions of this board, and I haven't been able to find an actual review, just vaporware mentions.

I asked someone closely involved in the Zedux board about Z280 stuff several years ago. I got a nice, but very long, rant-filled (and private at this person's request) reply about the bugs that prevented it from coming to fruition. Even after all these years this person was still rather PO'ed about the situation. I learned several things I didn't know.

Peak Z280 design is Tilmann Reh's CPU280, and it works well. But too little, too late. Bill Shen has since done up several Z280 RC bus boards as well as a CP/M for them, and they work fine as well. But that's in the retrobrew hobbyist space, definitely not mainstream.

A TRS-80 model 4 replacement main board was announced by Anitek called TRX-280, but likewise never came to fruition. The original Model 4, better known as the 'non-gate-array' version, had provisions for a Z800 chip already on the board at the Z80 socket. A working Z800 in 1982/83 would have been good; a working Z280 in 1990 not so much (yes, Z280 was available before then, but early steppings had some rather frustrating bugs). Zilog had feeping creaturitis with the Z280 design; they bit off more than they could chew with all the fancy peripherals, cache, and advanced instructions. The Z280 instruction set is tantalizing; a straight Z80 replacement with less peripherals and no cache might have flown ok. But, not to be.

There is a thought experiment what-if article about this at http://www.desertpenguin.org/blog/what-if-the-zilog-z180-z280.html that I think is a good read, even if it does gloss over the 68000 and 6502 and all the systems based on them. The 68000 was and is a great chip, and really should have been the basis for the IBM PC, but not to be. 8088 was just too similar to the 8085 and associated peripheral chips....but the reasons IBM went 8088 are better documented by experts with more knowledge of those circumstances than I in other places.

6502, for all its impressive efficiencies, was a deader end than Z80 and kin for what IBM had in mind, and 65816 was as frustratingly late as Z800.

(You may need to force non-https on that link)

 

[lowen]

Tillman Reh's article on a Z-280 board appeared in 1991 and the related kit seems about the only Z-280 anything that actually existed in numbers. Looks like another case where Zilog's marketing was trying to sell a CPU before the engineers got it working.

I have a CPU280. It runs Tilmann's banked CP/M 3 very well. I also have access to the gerbers (with Tilmann's blessing) and bare boards if anyone wants to make a CPU280. I might have a set or two of GALs programmed, too. It's not an inexpensive board to produce, using several hard to find chips. Z8028012VSC is easy to find compared to two of the other chips; used Specialix SI/XIO boards are on eBay, and each and every one of those includes a Z280 MPU. A major European point of sale manufacturer, Nixdorf, used Z280 in several models of cash registers. (Reference: https://techmonitor.ai/technology/n...f_z80_at_low_unix_at_high_end_for_retail_line ).

There's way more to this story than meets the eye. Search on Google groups for posts by samiam (at) moorecad.com

 

[Eudimorphodon]

There is a thought experiment what-if article about this at http://www.desertpenguin.org/blog/what-if-the-zilog-z180-z280.html that I think is a good read, even if it does gloss over the 68000 and 6502 and all the systems based on them.

There are some pretty major boo-boos in that article. First off, he seems a bit muddled about what the Z180 actually is, referring to it as ”basically a 16 bit CPU with 20 bit memory addressing, just like the 8086”. No, it’s not. I’d call it a badly constructed sentence where he’s trying to say it has 16 bit addressing inside a 20 bit physical space, but later he explicitly refers to an alternate history IBM debuting a “16 bit OS” for it sometime after apparently releasing the machine originally with “8 bit” CP/M. No. Just… no.

I mean, sure, that said you could probably still run with the argument that the Z180 is at least as capable as the 8088 and you could have in theory made a pretty good ”IBM PC” out of it; it’s faster per clock, sometimes at least, than the Z80, and in addition to the MMU (which, downsides, is kind of limited and confusing to use; it is *not* anywhere near as good as the 8086’s segment registers) it has a built in DMA controller that can do memory-to-memory transfers *way* faster than either the Z80’s or 8088’s string move instructions. If it’d been available in 1980 it could have made for a pretty compelling CPU option not just for IBM, but for, say, Radio Shack and other existing Z80 users…

(Arguably the best CPU accelerator ever made for the TRS-80 Model 4, the XLR8er, used the Z180/Hitachi 64180, and some people wrote some pretty interesting proof of concepts with it, like using the DMA controller to do extremely fast graphics updates on machines fitted with the optional graphics board.)

… but the Z180 *was not available in 1980*. Hitachi didn’t come up with it until 1985, with Zilog licensing it back and spitting out their version shortly thereafter. Obviously that’s way, way too late for any of this alternate universe fantasy. If Zilog had actually gotten the *Z800* out the door in 1983, when they’d first started floating it (as noted, Radio Shack believed them enough to make provisions for it in the Model 4, I would guess they were far from the only disappointed party) then, I dunno… honestly I think it still would have been too late overall to stop the PC wave, but it might have made for an interesting last crop of home computers. A Z800/Z280 based Amstrad CPC, for instance, would basically be as capable as an Atari ST; in fact, in this universe maybe the Atari ST might have been Z800 based. Again, probably wouldn’t necessarily have saved them in the end, but it would have been… interesting.

The 68000 was and is a great chip, and really should have been the basis for the IBM PC, but not to be.

This is the standard alternate universe line and, sure, a 68000 based PC would have been great, but the resulting machine would have been gobsmackingly more expensive, and the chip wasn’t even really available in quantity yet when IBM went shopping. If the base price of the PC had been $5,000 instead of $1,595 and it came out a year later I’m definitely thinking we’d all be using something else today. Very possibly Macintoshes, or machines descended from Atari ST-style knockoffs.