Protected mode on the 286 was seriously flawed. It couldn’t run most existing DOS real mode applications without resetting the CPU, and it didn’t make it easy to access the new features. You now had a 16MB address space (up from 1MB on the 8086), but you still had to access it through 64kB segments. Protected mode on 286 actually made this worse because it added overhead when modifying segment registers, so accessing large memory areas got slower.
OS/2 1.x was designed for the 286 and couldn’t escape these limitations. In theory it was a decent improvement on MS-DOS, but in practice there wasn’t enough value to counter the lack of compatibility and the higher price.
pjmlp 2 hours ago [-]
The root cause is that at the time they did an Apple, as they though everyone would run OS/2 or UNIX without any consideration for "legacy" MS-DOS applications, that is why 80286 design never considered this use case worth supporting.
Naturally as history has proven, this was a big mistake.
JdeBP 44 minutes ago [-]
As noted elsethread, there were no "legacy" (an abuse of that word) MS-DOS applications when the 80286 was designed. Intel wasn't really "doing an Apple".
At the time most people were expecting 80286es to become a second higher-end market where people ran Xenix (or perhaps a new MP/M for the 80286), separate from a low-end 8086 market with whatever tinpot non-multiuser non-virtual-memory operating systems were going to come out for it.
There was no CP/M for 8086/8088 CPUs in 1981 which is why Tim Paterson came up with QDOS for Seattle Computer Products. Microsoft's Paul Allen knew about QDOS and acquired it for IBM to use for the IBM PC.
PC-DOS 1.0 was released with the IBM PC in Aug 1981.
CP/M for IBM PC didn't show up until spring 1982.
80286 release date? 1982 Feb 01. Six months after IBM PC/PC-DOS 1.0 was introduced.
By the time the IBM PC's popularity took off, the 80286 design was etched in silicon already.
twoodfin 11 hours ago [-]
It’s kind of wild that neither IBM (who got beat to the hardware punch by Compaq) nor Microsoft (who targeted the NT effort at everything except x86) initially grokked how revolutionary the 80386 was: A “decent enough” 32-bit architecture with a huge preexisting ecosystem that would be able to ride the rocket ship of commodity PC scale.
glhaynes 10 hours ago [-]
On both counts, I think it's actually IBM who didn't get the importance. Both of the 386 to OS/2 and how important quick-to-market hardware was (even if just for brand prestige) versus Compaq.
Microsoft always got it, and I feel certain the first release of NT (3.1) sold many times as many copies for x86 as it did for other architectures; and it was targeted for it as much as for any other arch.
It was actually Microsoft that saw early that OS/2 needed to exploit the 386 but IBM dragged their feet on it. A strategy similar to Windows/286 vs. Windows/386 would've made a lot of sense IMO. And probably helped IBM sell more 32-bit Micro Channel hardware early on!
icedchai 10 hours ago [-]
Microsoft kind of did though? Windows/386 was released in late 1987. It could run multiple DOS apps using the "virtual 8086" mode. That was pretty revolutionary at the time. I think I knew more people using DESQview back then, though.
twoodfin 9 hours ago [-]
Sure, but their mainstream OS didn’t support the bulk of the 80386’s capabilities until late 1995.
That’s kind of insane if you think about it.
icedchai 7 hours ago [-]
True, though Windows 95 kinda ran like crap on the 386. We were well into the Pentium era by then.
msh 4 hours ago [-]
Not just kinda, it was close to unusable without a 486.
joshmarinacci 12 hours ago [-]
I notice some unusual spelling errors. Was this translated from another language or scanned with ai?
cr3ative 12 hours ago [-]
Probably just scanned in with classic OCR.
Rendered at 09:02:20 GMT+0000 (Coordinated Universal Time) with Vercel.
OS/2 1.x was designed for the 286 and couldn’t escape these limitations. In theory it was a decent improvement on MS-DOS, but in practice there wasn’t enough value to counter the lack of compatibility and the higher price.
Naturally as history has proven, this was a big mistake.
At the time most people were expecting 80286es to become a second higher-end market where people ran Xenix (or perhaps a new MP/M for the 80286), separate from a low-end 8086 market with whatever tinpot non-multiuser non-virtual-memory operating systems were going to come out for it.
There was no CP/M for 8086/8088 CPUs in 1981 which is why Tim Paterson came up with QDOS for Seattle Computer Products. Microsoft's Paul Allen knew about QDOS and acquired it for IBM to use for the IBM PC.
PC-DOS 1.0 was released with the IBM PC in Aug 1981.
CP/M for IBM PC didn't show up until spring 1982.
80286 release date? 1982 Feb 01. Six months after IBM PC/PC-DOS 1.0 was introduced.
By the time the IBM PC's popularity took off, the 80286 design was etched in silicon already.
Microsoft always got it, and I feel certain the first release of NT (3.1) sold many times as many copies for x86 as it did for other architectures; and it was targeted for it as much as for any other arch.
It was actually Microsoft that saw early that OS/2 needed to exploit the 386 but IBM dragged their feet on it. A strategy similar to Windows/286 vs. Windows/386 would've made a lot of sense IMO. And probably helped IBM sell more 32-bit Micro Channel hardware early on!
That’s kind of insane if you think about it.