- VCF South West - June 14 - 16, Davidson-Gundy Alumni Center at University of Texas at Dallas
- VCF West - Aug 2 - 3, Computer History Museum, Mountain View, CA
- VCF Midwest - Sept 7 - 8 2024, Schaumburg, IL
- VCF SoCal - Mid February 2025, Location TBD, Southern CA
- VCF East - April 2025, Infoage Museum, Wall NJ
-
Please review our updated Terms and Rules here
Why Intel did not release 386SX until 1988?
- Thread starter yuhong
- Start date
SpidersWeb
Veteran Member
286 didn't really have problems, it just didn't have 386 features.
Release a cheap 386SX processor and you'll hurt the throughput on those almost new 286 factories and brand new 386DX factories.
So you would've wanted to leave it as long as possible, but to get it out just before a competitor stole that 'in-between' market.
I wasn't there, just the 386SX being late makes sense to me.
Someone else will hopefully know a bit more.
Release a cheap 386SX processor and you'll hurt the throughput on those almost new 286 factories and brand new 386DX factories.
So you would've wanted to leave it as long as possible, but to get it out just before a competitor stole that 'in-between' market.
I wasn't there, just the 386SX being late makes sense to me.
Someone else will hopefully know a bit more.
Chromedome45
Veteran Member
This was a case of Intel eating it's young. It was trying to kill off the 286 as fast as it could. And make you pay more for a "32" bit cpu. All about money!!! Just my 2Cents worth
Chuck(G)
25k Member
I do know that Intel was getting very frustrated with the slow pace of x86 software development. It was basically a game of chicken. 16-bit MS-DOS applications in the popular area reigned supreme; any extra memory was used for RAM disk or cache.8088 systems were still widespread in use, as were 80286 systems, so no developer wanted to cross the line and say that a 386 was needed to run software. DOS extenders were out, but they made use of the 286 protected mode (segmented) and 32-bit instructions were left to languish. There were a few 32-bit applications, but they were mostly high-priced stuff, such as AutoCAD that were not used by the general population.
As one 386 engineer said (I can't remember when), "We gave you a fantastic 32-bit platform and you piss it all away running real-mode DOS".
The 386SX was a stroke of genius. It substantially lowered the entry price to 32-bit computing for the masses and began the end of the 8088 PC's dominance. By 1992, it was safe to write 32-bit code and expect it to sell.
As one 386 engineer said (I can't remember when), "We gave you a fantastic 32-bit platform and you piss it all away running real-mode DOS".
The 386SX was a stroke of genius. It substantially lowered the entry price to 32-bit computing for the masses and began the end of the 8088 PC's dominance. By 1992, it was safe to write 32-bit code and expect it to sell.
Of course, after that the problem was on the software side, particularly the MS OS/2 2.0 fiasco. But by the "problems" of the 286, I am referring to things like lack of support for switching back to real mode and lack of virtual 8086 mode.
Last edited:
Chuck(G)
25k Member
It wasn't long before the 286 "LOADALL" instruction was leaked (I sometimes wonder if it was a leak), so the issue of switching back to real mode became a non-issue. Microsoft employed it in Windows. Virtual 8086 mode was a non-issue for most people.
krebizfan
Veteran Member
Intel had to increase the speed of the 386 to much better than the 286 before releasing a 386sx. With Intel's policy of selling the 386sx slower than the 386DX at the same time, when Intel was only able to make 12MHz 386 that would require making a 386sx that would be slower than Intel's own 286, let alone the improved competition. The other challenge that needed to be met was for the development of megabit memory chips to permit systems to break the 16 MB limit. Clock speed and increased memory gave reasons to purchase the 386dx instead of having cheap 386sx models consume most of the high end sales.
Though the sx was not delayed by much, it took Intel about a year to make the 8088 from the 8086 compared to year and a half for the 386sx from the original 386.
Though the sx was not delayed by much, it took Intel about a year to make the 8088 from the 8086 compared to year and a half for the 386sx from the original 386.
LOADALL do not allow switching back to real mode. And yea:
http://www.rcollins.org/articles/loadall/tspec_a3_doc.html
http://www.rcollins.org/articles/loadall/tspec_a3_doc.html
Chuck(G)
25k Member
No, but the big problem for DOS users was accessing upper memory. Going through the controlled-reboot was very slow, as you mentioned. But if you didn't need the execution model of 286 PM (and who really did?), LOADALL could solve a lot of problems.
Caluser2000
Banned
Let's not forget Intel was tying up AMD with regard to producing 386 cpus. AMD eventualing one a lawssuit allowing them to produce 386s up in the early 1990s. AMDs 286s were on par with Intels 386sxs, apart from 386 specific instructions, for the most part anyway.
Last edited:
Chuck(G)
25k Member
"one a lawsuit"? Are you using a speech-to-text tool?
But your post points up a key fact--in 1988, almost nobody really needed a 32-bit instruction set in their PC. Phar Lap, for instance, continued well into the mid-to-late 90s licensing their 16-bit DOS extender. I found some use for 32-bit registers and instructions in some code I wrote back then, but it was important that I not make a 386 a requirement for my code. So I wrote code that tested for 386 presence and used it in key routines where it would make a difference. That was very typical for the time.
How long was it before Windows got rid of "thunking" from 32-bit code to its own 16-bit DLLs?
I think it was Win9x that did that, and took the infamous Win16Mutex while doing it to simulate a cooperatively multitasking system. OS/2 2.x did it too but didn't need the mutex since it was based on 1.x."one a lawsuit"? Are you using a speech-to-text tool?
But your post points up a key fact--in 1988, almost nobody really needed a 32-bit instruction set in their PC. Phar Lap, for instance, continued well into the mid-to-late 90s licensing their 16-bit DOS extender. I found some use for 32-bit registers and instructions in some code I wrote back then, but it was important that I not make a 386 a requirement for my code. So I wrote code that tested for 386 presence and used it in key routines where it would make a difference. That was very typical for the time.
How long was it before Windows got rid of "thunking" from 32-bit code to its own 16-bit DLLs?
Caluser2000
Banned
Depends on the market segment your talking about. For the most of the 90s 32 bit consumer hardware market, despite the faults, Win 9x was good enough. It wasn't until XP was released it(NT) really took off for consumers and prior to that it was NT4 for mass corporate/business adoption.
Chuck(G)
25k Member
Win2K was widely used--I still use it. Much lower resource demand than XP.
Remember that the 386 was released years before Win95. Before Win95 there was Win32S under Windows 3.1. My first web browser used Win32S.
Remember that the 386 was released years before Win95. Before Win95 there was Win32S under Windows 3.1. My first web browser used Win32S.
Caluser2000
Banned
W2k never had the mass consumer market XP received. It was more a business corperate/orientated OS.
Caluser2000
Banned
Yes it was realeased earlier than win95. IBM could've pushed harder if they wanted. To run smoothly OS/2 still needed more resources than the average PC had at the time. Though they did do some shifty things it wasn't entirely MSs fault. I've got OS/2 v2 and v2.1 documentation mentioning problems with particular IBM computer models. These could run dos/windows 3.x without any issues at all.
As a Corperate whole IBM, despite the marketing, was beginning to pull the plug on OS/2 around the time ver 4 was released. If it wasn't for a dedicated team of OS/2 supporters and users it would have been killed off completely. There must be still some core users/niche or eComstation wouldn't exist today. .
dpatten
Experienced Member
Technically they still haven't. If you run the 32 bit version of windows 7, you will have no trouble at all installing 16 bit windows applications.
Chuck(G)
25k Member
As an OS/2 developer, as much as I detested Microsoft's duplicity, it made sense. IBM under Cannavino was becoming moribund. For that matter, Apple nearly went bankrupt during the same period. The problem was that there was no vision out there. Microsoft took advantage of the situation.
If you've fooled much with OS/2 2.x and tried to add to or modify your system configuration, you'll remember how awkward it was. Long after OS/2 2 was being distributed on CD-ROM, the CSDs were still being passed out on floppy--and the size of some of those correction sets was staggering. No, as nice as a platform that OS/2 was, there was a good reason that it was ditched by both Microsoft and IBM.