VOGONS


Reply 20 of 29, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++
pan069 wrote on 2023-06-08, 21:39:
H3nrik V! wrote on 2023-06-08, 19:53:
dionb wrote on 2023-06-08, 16:28:

Some did, but back then they were the exception, not the rule.

It wasn't until 1994 or so that it became commonplace.

Agreed, AFAIR, my own first 486 had serial and parallel ports on the VESA "super I/O board" which also had the IDE controller ...

286's were made well into the 90's. By that time they were cost reduced and integrated and mostly used for business purposes. A 20 or 25 Mhz 286 was good enough in those days to run WordPerfect and do a bit of Lotus 123 on a monochrome display.

I think what wiped out the 286 as a base level DOS system was AMD finally winning in court and releasing the floodgates on it's 386es, then it was wall to wall 386 integrated boards.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 21 of 29, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
dionb wrote on 2023-06-08, 14:22:

486 and serial/parallel cables? Onboard I/O was very much exception rather than rule (except on OEM LPX boards, which didn't need them because already on backplate).

VLB era 486 onward almost always had floppy parallel and serial cables. IDE was hit or miss

But both my micronics vlb and PCCHIPS pci 486 came with 1 serial, parallel, floppy and ide cables plus a manual and even a little packet of screws plus a couple jumpers.
Odder still the cache chips came uninstalled packaged separately (with the first board) and I had to set the jumpers and install the cache.

My guess is before 1994 the stuff that came with your board matched what was on the board, even in the dark ages it wasn’t uncommon for serial or parallel to be optional on a header but it was by no means a standard thing as folks still wanted everything discrete

Reply 22 of 29, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
BitWrangler wrote on 2023-06-09, 13:50:

I think what wiped out the 286 as a base level DOS system was AMD finally winning in court and releasing the floodgates on its 386es, then it was wall to wall 386 integrated boards.

The 286 could have limped past 1992 but several major things killed it.

1. Cost, for a 286 to continue it would have had to get further integrated like a 186 or MediaGX as the 386sx was virtually the same price . The 286 already had a power consumption advantage which embedded it into the subcompact/ notebook laptop space into 1994, further focus on this could have embedded it into a niche terminal, pc104, embedded and handheld market longer for power, size and cost reasons.

2. Clock speed, the 286 was definitely beloved especially overseas but it did not continue getting refined into higher speed silicon. In an alternate universe Harris’s cutting edge process which did save power but was several years behind “others” would have gotten updated in 1988 to release a 33mhz part, this would have made Intels smear campaign against 286 chips less effective and forced them to get “good sx” chips out the door faster in quantity
As it actually Turned out Intel was given the chance to scale up and flood the early 386sx market with slow parts as others stagnated selling the same slow 286 parts.

3. Emulation, Dos extenders, software support.
Intel felt it could only survive if older chips were made unsellable, to accomplish they started with the 286 smear campaign in the late 80’s and heavily supported the development of free dos extenders for 386 chips. Microsoft and IBM abandoned support relatively quickly along a pure cost basis with full support from Intel, even though the 286 install base was massive in 1993. If a free and easy to use extender and programming toolset would have become available for 286 early that alone would have stifled the main reason the 286 went from being on sale one year to virtually unsupported the next. 286 extenders were expensive and had far less development and ease of use effort put into them.
Further the 286 (a von Newman ) actually could easily emulate 386 operations, especially on 640k dos software. Historically there were small emulators that worked extremely well but almost nobody knew they existed , speed was less important than compatibility back then and having an emulator even right in the bios with a work around for detection would have added further legitimacy even if it was a cludge.
It could also simplify cross portability if it was in a quality easy to use toolset.

The 286 sales chart is quite unusual as it literally shut off like a light switch.

Nothing like that happened before or since so it means a major shift across, corporate software, hardware vendors and consumers all in an extremely short period of time

Reply 23 of 29, by rasz_pl

User metadata
Rank l33t
Rank
l33t

Nobody wanted 16bit segmented mode, 386 finally introduced sane flat memory programming paradigm. Coupled with x86 virtual mode you could finally do some interesting stuff like virtual machines seamlessly (except for Intel screwing up not trapping POPF) emulating hardware in software. Windows 3.0 with its Enhanced Mode was a success, and a follow up Windows 3.1 practically killed any competing attempts.

286 was terrible to program for and its a good thing it died.

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 24 of 29, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
rasz_pl wrote on 2023-06-09, 18:29:

Nobody wanted 16bit segmented mode,

286 was terrible to program for and its a good thing it died.

It’s worth noting even those that did want to support segmentation and 286 could not reasonably do so because of the very high cost and low quality of 286 dos extenders compared to FREE for 32 bit.

Only those willing to stay in 640k or had their own tools could afford to support 286

This stopped a lot of mom and pops despite the significantly larger install base of 286 systems in the early 90’s

This is why you actually saw a lot of 286 windows “games” in the twilight years of the 286 because it was easier to support a 286 in windows than dos (even though it made for a slow experience)

Intel rightly judged that the best way to kill the 286 was buy making development tools cost too much to support it vrs 32bit. Strangling software support.

Reply 25 of 29, by rasz_pl

User metadata
Rank l33t
Rank
l33t
rmay635703 wrote on 2023-06-10, 03:18:

It’s worth noting even those that did want to support segmentation and 286 could not reasonably do so because of the very high cost and low quality of 286 dos extenders compared to FREE for 32 bit.

There were free 286 extenders, Borland shipped one with its compilers starting in 1992.
Free DOS/4GW (W is for Watcom) bundle was in 1993.
https://virtuallyfun.com/2011/08/02/some-history-on-dpmi/

rmay635703 wrote on 2023-06-10, 03:18:

Only those willing to stay in 640k or had their own tools could afford to support 286

pre dos extenders we had EMS/XMS

rmay635703 wrote on 2023-06-10, 03:18:

This is why you actually saw a lot of 286 windows “games” in the twilight years of the 286 because it was easier to support a 286 in windows than dos (even though it made for a slow experience)

I dont think thats accurate. REMOVED contains ~1100 win3 games, but if you search by year in file list REMOVED you get more or less
1989 0
1990 10
1991 30
1992 60
1993 100
1994 200
1995 300
1996 260
1997 100
1998 50
1999 30

Peak is right before win95. Imo main reason behind the flood of win3 games was market analysis. Computers were expensive, most computers were in offices, most office computers in early nineties ran windows3, people with jobs are more likely to pay for software.

rmay635703 wrote on 2023-06-10, 03:18:

Intel rightly judged that the best way to kill the 286 was buy making development tools cost too much to support it vrs 32bit. Strangling software support.

I dont understand that argument, how could legacy dev tools cost more than brand new shiny 32bit versions?

Last edited by DosFreak on 2023-12-05, 00:17. Edited 2 times in total.

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 27 of 29, by rasz_pl

User metadata
Rank l33t
Rank
l33t
kolderman wrote on 2023-06-10, 11:10:

> pre dos extenders we had EMS/XMS

Well except for 286 EMS meant an actual memory board and XMS meant resetting the cpu every context switch.

plenty 286 chipsets with ems support, 286 XMS HIMEM.SYS uses loadall

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 28 of 29, by rmay635703

User metadata
Rank Oldbie
Rank
Oldbie
rasz_pl wrote on 2023-06-10, 10:52:
There were free 286 extenders, Borland shipped one with its compilers starting in 1992. Free DOS/4GW (W is for Watcom) bundle w […]
Show full quote
rmay635703 wrote on 2023-06-10, 03:18:

It’s worth noting even those that did want to support segmentation and 286 could not reasonably do so because of the very high cost and low quality of 286 dos extenders compared to FREE for 32 bit.

There were free 286 extenders, Borland shipped one with its compilers starting in 1992.
Free DOS/4GW (W is for Watcom) bundle was in 1993.
https://virtuallyfun.com/2011/08/02/some-history-on-dpmi/

rmay635703 wrote on 2023-06-10, 03:18:

Only those willing to stay in 640k or had their own tools could afford to support 286

pre dos extenders we had EMS/XMS

rmay635703 wrote on 2023-06-10, 03:18:

This is why you actually saw a lot of 286 windows “games” in the twilight years of the 286 because it was easier to support a 286 in windows than dos (even though it made for a slow experience)

I dont think thats accurate. REMOVED contains ~1100 win3 games, but if you search by year in file list REMOVED you get more or less
1989 0
1990 10
1991 30
1992 60
1993 100
1994 200
1995 300
1996 260
1997 100
1998 50
1999 30

Peak is right before win95. Imo main reason behind the flood of win3 games was market analysis. Computers were expensive, most computers were in offices, most office computers in early nineties ran windows3, people with jobs are more likely to pay for software.

rmay635703 wrote on 2023-06-10, 03:18:

Intel rightly judged that the best way to kill the 286 was buy making development tools cost too much to support it vrs 32bit. Strangling software support.

I dont understand that argument, how could legacy dev tools cost more than brand new shiny 32bit versions?

Too little too late.

Those game numbers seem a little “odd”, as I appear to have more on my personal systems than you list.

Various individuals who programmed at the time have mentioned cost issues With 286 extended memory distributables. One of my relatives had a school programming class that allowed students free copies of various 32 bit dev tools including distributables like dos4g despite the fact no one at the school owned a 386. On the other hand If a student wanted a take home 16 bit dev or distributable for a 286 program they had to pay for the softwares student version as the school could not get an agreement from those companies.

Check Wikipedia

Dos4g could be freely distributed with a target 32 bit program

Versus 286 dos extenders were indeed multiple thousands of dollars for a base license and were also on a per instance basis and not freely distributable with the target software until VERY late.

Why this matters is because most small game developers and shareware folks started late 80’s and early 90’s many times at school. These folks would not have a 286 dev laying around and would definitely target the latest and greatest since the tools were free to distribute and freely available to them at school.

Last edited by DosFreak on 2023-12-05, 00:17. Edited 1 time in total.

Reply 29 of 29, by rasz_pl

User metadata
Rank l33t
Rank
l33t
rmay635703 wrote on 2023-06-12, 15:08:

Those game numbers seem a little “odd”, as I appear to have more on my personal systems than you list.

are you saying you have more than 1100 pirated win3 games on your hard drive? 😀 There was more shovelware, Softkey 2000 Shareware Games lists ~900 https://www.pixelships.com/sdig/directories.html

rmay635703 wrote on 2023-06-12, 15:08:

Various individuals who programmed at the time have mentioned cost issues With 286 extended memory distributables. One of my relatives had a school programming class that allowed students free copies of various 32 bit dev tools including distributables like dos4g despite the fact no one at the school owned a 386. On the other hand If a student wanted a take home 16 bit dev or distributable for a 286 program they had to pay for the softwares student version as the school could not get an agreement from those companies.

One school not having free student software versus factual data on what was on the market. Borland shipped 286 extender with its products in 1992, so did Zortech (DOS/16M). $90-200 MS QuickC 1.0 for Windows was another solution.

rmay635703 wrote on 2023-06-12, 15:08:

Dos4g could be freely distributed with a target 32 bit program

W, dos4gw bundled with Watcom, just like aforementioned Borland (https://en.wikipedia.org/wiki/Talk%3ADOS_exte … _C++_and_Pascal) and Zortech

rmay635703 wrote on 2023-06-12, 15:08:

Versus 286 dos extenders were indeed multiple thousands of dollars for a base license and were also on a per instance basis and not freely distributable with the target software until VERY late.

~1991-2, 1-2 years before Watcom

rmay635703 wrote on 2023-06-12, 15:08:

Why this matters is because most small game developers and shareware folks started late 80’s and early 90’s many times at school. These folks would not have a 286 dev laying around and would definitely target the latest and greatest since the tools were free to distribute and freely available to them at school.

Watcom wasnt free either.

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction