mkarcher wrote on Yesterday, 16:34:

Grzyb wrote on Yesterday, 08:59:

286 with 512 KB RAM? In 1988? Seriously???

That was a quite common entry-level AT configration in Germany at that time.

OK, just had a look into "Byte" magazines from 1988 - 286 boxes with only 512 KB RAM were still sold even in the USA.
Hard to believe it, considering that in 1987, all 286 IBM PS/2s already had 1 MB - but I guess some people wanted something cheaper, with the option to upgrade it later on.

Anyway, the fact that OS/2 was designed for 286, looks even more of a mistake...

Floppy controllers started to appear integrated on the mainbaord at that time.

Yes, but I would expect stuff integrated on the mobo only in brand-name PCs, eg. Commodore PC 10-III/20-III.
In generic clones, I would rather expect a multi-I/O card = FDC+LPT+COM+GAME+RTC.
At least that's what I recall from late XT clones...

Grzyb wrote on Yesterday, 17:36:

Hard to believe it, considering that in 1987, all 286 IBM PS/2s already had 1 MB - but I guess some people wanted something cheaper, with the option to upgrade it later on.

Anyway, the fact that OS/2 was designed for 286, looks even more of a mistake...

I think it's wrong to equate a processor type with a system configuration. I guess most people that bought the entry-level AT machines with MDA/Hercules and 512K of RAM were not looking for multitasking or modern operating system, they just wanted a system that supported 1.2MB instead of 360KB floppies. 360KB already started to feel "small" at that time. You started to have an operating system disk to boot, an application disk with the software and a third disk to store your data on. Even in a two-drive system, this means swapping disks got part of the daily routine. Note that this 286 system we are talking about did not include a hard drive! This system is somehow like a "Turbo PC" - still no hard drive, but more performance and bigger disk capacity.

Sure, there is no technical reason to not have 1.2MB drives in a PC-class 8088 system, and I guess most happy customers of 512KB/noHDD/Hercules/8MHz systems would have been fine with a 10MHz Turbo-XT as well, if it had 1.2MB drives. But the XT BIOS does not support multi-rate floppy controllers, and floppy controller cards with their own BIOS were not common. On the other hand, integrated AT chipsets started to get common, and AT boards with native HD drive support got a commodity item.

On the other hand, there were the higher-end 286 systems like the PS/2 series that started at 1MB RAM and could be upgraded to 4MB at least, possibly even 8MB. Those are the system OS/2 was designed for. In retrospect, we know that the 286 class as affordable PC system was quite short-lived, especially due to 386SX systems replacing that position on the market. While it didn't matter for native OS/2 applications, the 386SX with its built-in 8086 virtualization capability that enabled EMM386 and Windows/386 (the origin of Windows 3.0 enhanced 386 mode) was clearly the better processor to multitask legacy DOS applications. Harris published a paper that for 286 software, a 286 system is typically more performant than a 386sx system at the same clock rate. While that paper clearly has obviously written for marketing purposes (Harris had a 286 license, and was AFAIK the only vendor selling 20MHz and 25MHz 286 processors; Harris had no 386 license), and the performance benefits of 286 vs. 386sx is likely overblown, looking at the execution time tables of the 286 and 386 processors, there likely is some truth that the 286 would actually execute 16-bit OS/2 code faster than a similar 386SX system. If the 386SX wouldn't have taken the 286 market share, I would expect we would have seen 4MB 286 systems as mid-range systems in 1991, the perfect target for 16-bit OS/2.

Grzyb wrote on Yesterday, 17:36:

Floppy controllers started to appear integrated on the mainbaord at that time.

Yes, but I would expect stuff integrated on the mobo only in brand-name PCs, eg. Commodore PC 10-III/20-III.

Well, I thought the same, but I just recently help to get two dumpster-rescued retro systems up and running that were of the very age we are currently talking about - and both had a mainboard with an on-board floppy controller. One system was ASEM branded, and the other system was SWEDA branded. At least for me, that doesn't sound like the big brands. The SWEDA system is a 10MHz AT clone (with 512KB of 1MB populated on the main board), and the only issue it had was a broken WD 16-bit hard drive controller. The ST-225 drive in that system was still working fine with no bad sectors. We got a replacement HDD controller (a WD 1003-WAH) on which the floppy interface can not be disabled or moved to the secondary I/O address without also moving the hard drive part to the secondary I/O address, and thus we had to experiment which of the (luckily very few) mainboard jumpers was used to disable the mainboard floppy controller.

Just as a side note: The ASEM branded system is a 9.54MHz Turbo-XT based on the FE2010A XT-on-a-Chip, and someone installed an original IBM CGA into that system. The only issue with that system was a shorted tantalum on the 12V line on that very CGA card. The ST-225 in that system also worked fine with no bad sectors...

mkarcher wrote on Yesterday, 16:34:

I wonder who could make use of a 386 processor, but considered Hercules graphics sufficient. Classic Hercules cards are 8-bit cards and do not even manage to run without extra wait-states in a 4.77 MHz PC/XT.

Hi, I assume that DESQView+QEMM, DESQView 386 or Windows/386 or CEMM could have utilized the the 386 that early.

PC-MOS/386 also existed by 1987, but it was rather niche maybe. It just comes to mind because my father had it..

Concurrent DOS and Xenix had 386 versions, too but I have little experience here.
By late 80s, AutoCAD had 386 versions available, I think.

The DOS Extra Magazine does list Windows 2.03 and Windows/386, at very least.

But to be honest, I assume that the average user simply wanted an 386 in first place because it had offered noticeable better performance. 😟
The enhanced MMU and V86 were next on the wish list, maybe.

286 with 512 KB RAM? In 1988? Seriously??? What a beautiful machine for the brand-new OS/2 1.1, with Presentation Manager! 🤣

Hey, the Magazine wasn't called "DOS Extra" for nothing! 😃
It it was about OS/2, it'd been called "BS/2 Extra", after all! 😉

Seriously, though. As-is the models shown wouldn't have run OS/2 Presentation Manager anyway because IBM refused to support Hercules graphics.
Microsoft by contrast did support Hercules in Windows/386, for example.

Speaking of OS/2 1.1, it also ran on 386 systems, but there were differences between IBM and MS releases.
And tzere were different editions. Standard, Extended etc. Some might have needed 286 or 386 to run properly.
The HPFS 386 filesystem driver needed an 80386, for example.

The whole LOADALL vs LOADALL386 differences might have mattered, too.
AFAIK the average 80386 chip didn't have normal LOADALL instruction of the 80286 anymore (for security reasons), so the AT 386 BIOS emulates it for software requiring it.

However, if an OS is not running in Real Mode but Protected-Mode that emulation may or may not work?
An AT with an 386 CPU upgrade doesn't have LOADALL emulation, either, maybe.
Because, I assume, the AT BIOS is unware of the situation and treats the 80386 like an 80286.

PS: My apologies for being a bit off-topic. I didn't mean to de-rail the thread here through my replies.
It's just that I got a little bit carried away. I hope the OP doesn't mind.

mkarcher wrote on Yesterday, 18:59:

I guess most people that bought the entry-level AT machines with MDA/Hercules and 512K of RAM were not looking for multitasking or modern operating system, they just wanted a system that supported 1.2MB instead of 360KB floppies. 360KB already started to feel "small" at that time. You started to have an operating system disk to boot, an application disk with the software and a third disk to store your data on. Even in a two-drive system, this means swapping disks got part of the daily routine. Note that this 286 system we are talking about did not include a hard drive!

I'm looking at 1988 price lists, and it seems that XT with 20 MB HDD costs about the same as AT without HDD.

I would've definitely choosen the XT: 360 KB + 20 MB is way more convenient than 1.2 MB, no disk swapping at all!
...unless I was planning to upgrade it later - the XT was already pretty much at dead end, while the 286 was still very promising.

^But an traditional 8088 PC/XT @4,77 MHz was painfully slow (I have a few XTs here).
DIR command alone was a stress test (V20/V30 helped a bit).
The characters crawled the screen like you were calling a mailbox at 300 or 1200 Baud.

In terms of user experience, a slow 6 MHz AT must felt about same as a super fast 16 MHz PC/XT.
In practice, even a C64 seemed to respond quicker than an original 4,77 MHz PC.

If I had to opt between suffering financially and mentally, I would have chosen the first.
There are things that are more important than money in life, I think. Suffering isn't worth the money.

Also, in terms of productivity, an 4,77 MHz PC can hold up an entire workflow.
Using a mechanical typewriter or a real card file was the better alternative sometimes.

So unless the PC was used as an electronic typewriter (word processor) or for hobby use,
I would have installed a CPU accelerator not long after.
Such as Orchid Tiny Turbo 286 or an Microsoft MACH card (fast 8086 or 80286).

Because, if you pay lots of money for a PC it must at least be usable also.
If it's not fully functional, then there's no real gain and the investment was a failure.
So it's better to pay extra money to make it run as intended, even it hurts financially a little bit.

Edit: Again, sorry for being a bit off-topic. Maybe should go back on main topic.

Jo22 wrote on Yesterday, 21:23:

^But an traditional 8088 PC/XT @4,77 MHz was painfully slow

We're talking about 1988 here - 4.77 MHz XTs are no longer available.
Note that those Marlow systems from your magazine are 10 MHz - both the 8088 and 286 models.
The 286 is more than 2x faster, of course - but the Turbo XT is also bearable with 80s software.
Also, speed is not just about CPU, but also about disk I/O - there's plenty of applications that run faster on an XT with HDD, than on 286 with floppy!

If you needed a computer for a well-defined purpose, without plans to use it for anything else - a good XT was perfectly fine!

Otherwise... better buy a poor AT - with the constantly dropping prices, adding more RAM and HDD was sure to become affordable sooner or later.

Edit: Again, sorry for being a bit off-topic. Maybe should go back on main topic.

I wouldn't worry - looks like the topic got exhausted anyway...

BaronSFel001 wrote on 2025-09-09, 13:58:

This is the bigger market picture I see: as of 1991 Apogee still supported CGA not just as a matter of principle; their catalog for that year still included Kroz and the FAST engine games while, as I already mentioned, Softdisk's subscription model accommodated an even lower common denominator. Come to notice it, Goodbye, Galaxy! was the last Apogee game to support CGA, and only in that special edition (not unlike how MicroProse released special editions of F-117A Stealth Fighter 2.0 & Gunship 2000 that same year for those who still only had EGA capability). Invasion of the Vorticons and Todd Replogle's games based on id technology (Dark Ages & Duke Nukem) required EGA, so I can buy Apogee considered potential sales lost for users of lower-end systems. Yet John Carmack did the work, as confirmed by how impressed he was by the composite CGA conversion a few years ago, and he's not a man known for engaging in afterthought projects - such projects existed for id, but typically got farmed out to Jason Blochowiak instead of involving the main team.

I guess actually the irony is that the initial Commander Keen game was based on an engine the original id guys wrote while working at Softdisk in their spare time. Because the game engine did not support CGA, the Softdisk management did not want to use it. So the id guys took their ideas to Apogee who were contacting them seeing their games, and who then published a game from the engine, which only supported EGA. But then Softdisk, finding out what they did, confronted the team about using company resources, and ended up preventing a revolt of sorts of people quitting, and decided to contract with the team that was forming id that they would deliver a number of games for their publications. One of those games was Keen Dreams, which supported CGA. So I do think that in the effort of delivering those games, Carmack decided to extend Keen to support CGA for a Softdisk release, maybe because he thought it was technically possible anyway? So Keen went from being rejected for not supporting CGA, to writing a CGA port anyway because they had to get another game out, to get out of a mess they created with their former employer.
https://web.archive.org/web/20160402151326/ht … om/keenhistory/

So since the Keen engine gained CGA support with Keen Dreams, I guess the subsequent releases also had it just simply out of principle that the engine supported it.