• Please review our updated Terms and Rules here

Does the 386 design predate the 286?

cr1901

cr1901

Veteran Member
Joined
Dec 28, 2011
Messages
817
Location
NJ
http://www.redhill.net.au/c/c-1.html

On the above webpage, if you go down to the 286-8 portion of the page, you will see this blurb:
RedHill said:
Oddly enough, the 286 was actually designed some time after the much more powerful 386. While the 386 design was clearly superior, it was just too large and complex and expensive to manufacture at first, and the 286 came into being as a temporary measure — a temporary measure that sold in millions and was in volume production for over ten years!
Click to expand...
Can someone who is familiar with the designs of the early x86 families comment on whether the above is accurate? Is there a more reliable source confirming the above? That would be VERY interesting to me if it were true.

However, two things bother me about this page that makes me suspect what they are saying:
  1. Look at their description of the 286-6...
    Illustration: a very rare 286 indeed. We can't remember ever seeing a 286-8, though we may well have done at some time, but we are quite certain we have never seen a 6MHz 286 like the one at right, nor indeed heard of one outside a one-word mention in Adrian Offerman's Chiplist. In fact, the picture raises more questions than it answers: first, it is the only 286 we can remember seeing in PGA (Pin Grid Array) packaging, though this was to become common with 386 parts and all but universal through the 486 and Pentium eras. Second, if it ever found its way into an AT style system, what speed did the I/O bus run at?
    Click to expand...
    I guess they never opened up a PC AT then... although "what speed did the I/O bus run at?" is an interesting question I can't answer offhand :p.
  2. Another picture shows what seems to be a DIP 286... well according to Chuck(G) in this thread, there is no such beast that exists... so what is that picture then?
 
I've never read any suggestion that the 386 design was completed before the 286, of course that doesn't mean it wasn't on the design board and it quickly became apparent that it couldn't be done with the technology available at the time.

Re IO bus. 8-bit ISA slots were always run (by IBM at least) with a cycle time of about 700-800ns, achieved by adding wait-states, to ensure compatibility with (what were at the time expensive) cards designed for the original, 4.77MHz machines. These wait states can though be removed by hardware asserting the B8 line.
 
the 286 was physically produced 2-3 years prior to the 386 being physically produced, but I have read from another source that the 386 design was slightly before the 286 design (or during the 286 design).
cost is probably a smaller factor compared to the physical readiness to manufacture the 386.. i would bet waiting on the 386 assembly was too far away and they were in a race to be the next company to release a new chip that was still within tangible performance.

8Mhz 286 was a big seller for a long time.. i read every PC magazine i could get my hands on.

i'm at a loss on why the author says 8mhz 286 chip is rare.. the AT began at 6Mhz in 1984 and quickly raised to 8Mhz and stayed at 8Mhz until discontinued in 1987. -- Even the 286-6 was sold in PCs through 1987 (including the XT 286). -- I think the author's first PC ever just happened to be a clone, a used 286 10mhz clone PC.. or like me, he jumped from the PC XT to a 10mhz 286.
My wild guess is for the home market, a home user, the 286-8mhz probably was a rare sight, even just as rare as seeing a PC in the home at all. Commercial uses were big into the PC AT at 8Mhz.
 
cr1901 said:
http://www.redhill.net.au/c/c-1.html
However, two things bother me about this page that makes me suspect what they are saying:
  1. Look at their description of the 286-6...
    I guess they never opened up a PC AT then... although "what speed did the I/O bus run at?" is an interesting question I can't answer offhand :p.
Click to expand...

That made me raise an eyebrow as well. I instantly thought of the original IBM AT.

But I think what might've been a factor is the .com.au domain address. I can't speak for Australia, but I wonder if they had a similar experience NZ.

Down here 8 bit machines dominated the early 1980's, then in the late 80's it was low priced 8088 clones. It would've been pretty rare for someone to purchase an IBM AT in 1984/1985.
 
386 was probably being designed before the 286 'release' but its design is not earlier than 286.
The 80286 processor was anounced March-1982.
The IBM-PC/AT was anounced August 1984. Most people bought ATs in the period 84-89.
The 80386 processor was anounced October 1985. Being only three years later, it is obvious the design was much earlier, but designing a micro does not mean to be able to produce it.
In August 1986 Compaq introduced the first 386 computer. IBM introduced, one year later, the PS/2 Model 80. No one was able to purchase a 386 before mid 1986, and most were sold from 1987 on.
However, in 1989 (the same year 486 was released) intel realized that people were still buying 286 machines!, so intel introduced the famous advertising campaing with the red 'X' over the 286 on a full white page, meaning people should stop buying 286s and buy at least the 386sx.
It was 1989 when intel introduced this campaign so ATs were selling very well for 6 years, since 1984 to 1989.
 
Last edited:
People were ignoring those ads still buying 286s in 1990 ;) That's when I got mine new, my first x86 machine, a 286/16. Comparing it with freinds 386SX ,similar clock speed, didn't seem terribly different and in some cases the 286 seemed better. Zenith/Heathkit were certainly producing 286 machines with the cpu a few years after that, with the option to replace the cpu daughterboard with a 386sx one. A few other OEMs also oftered that option. With my 286/16 being a generic clone box it did get upgraded, as folk passed on their hand me downs, over the next 8 years or so.

I actually skipped the 386 cpu altogether but did try a new 386 class modo with 486DLC chip out a year or so after I got my 286.Ended up getting my money back.I wasn't a gamer and the 800 or so dollars for the upgrade just didn't seem worth it to me at the time.
 
Last edited:
This is real interesting stuff. Sometime during 1989 I bought a 386 bare bones setup from some warehouse around O'Hare Airport in Chicago (saw an ad in Computer Shopper). The price was about $700 at the time and I believe the chip was a 386SX-20. It came with the motherboard, CPU, power supply, and a fairly decent tower case. The government agency that I was working for at that time was in love with the Wyse 286-10 desktops with the peach CRT's. The only 386's floating around at the time were of the laptop variety. About once a year a HQ team would show up with a load of new gear. Desktop PC's in our area were a premium and the 286's were the standard for the longest time. The next batch of upgrades were the Pentiums. I'm thinking that 386 was more or less a niche type PC, but was a real step forward in performance for its time. You just didn't see a lot of them in the business world, but the Mom & Pop computer outlets were full of the them. Personally, I like the 386 and have one that I recently acquired (bonus) with a 5160 purchase. Of course I'm speaking of my own experiences in my geographic specific area, and the 386 may have been the horse in other areas.
 
The 386 design might well have been on the books, but the 286 early steppings were concurrent with the 186 pre-release, which was very closely on the heels of the 5150 being available for market. It was also initially a 6MHz LCC chip, which makes me think that 6MHz was probably the initial limit of the process (we ran initial samples at 4 MHz). Both early pre-release chips were very buggy, the 286 more so. You have to understand that back then, prototype silicon was done, debugged, then released to selected customers after changes, who then found their own bugs. I suspect that the 286 and 186 both hail back to 1978-79 designs.

The development process back then could be pretty long. From a strategic standpoint, a 386 would have represented a market challenge to the then-ongoing 432. When that began to founder, I think the 386 was introduced as a stopgap.

Most of this is conjecture, admittedly.
 
My first x86 machine was an 80386DX-33. I bought the parts from a local computer store that has reinvented itself as necessary, and so is still in business. I installed SCO Unix on the 386, to more-or-less mirror what I was using at work. Until that time, my main computer at home had been my PDP-11/73 running RSX-11M-PLUS.

I never had an MSDOS box at home, although I had a couple Windows 3.1 systems when I was using that technology at work.

I never owned a 286 system until I bought a used one around 1995 for testing some software I wrote.
 
Red Hill is located in Australia and focused on products that appeared there. Early production CPUs with their extremely high prices for low clock speeds were more commonly sold in the US.

I am fairly sure that Intel announced a 32-bit CPU in 1980 or 1981. How similar it was to the 386 we all know is hard to say; as it was, Intel had to recall much of the production from 1985, 1986, and 1987 because early 386 chips did not correctly handle 32-bit instructions. I think the very early 386 also did not have a working Virtual 8086 mode.

Jan 1985 was the release date of the 386-12. I don't know how much earlier Engineering Sample 386s were available. Somewhere, I read that the 286 was designed and entered production in about a year.

I bought my first x86 system with a 286-10 in 1988; the entire system including EGA monitor cost less than just a bare 386 chip.
 
Last edited:
That 1981 32-bit CPU was the iAPX-432 as I mentioned. Horrifically expensive, even for that time, the CPU was actually 2 ICs, with more (expensive) chips, such as the MMU to be introduced later. Very advanced architecture (i.e. nothing like the 8086), and, depressingly slow on benchmarks.

Intel abandoned it right about the time that the 386 was introduced. It was probably a difficult decision, as Intel started working on the thing in---1975(!), while the 8080A was still shaking up the world.

There are those who say that the 432 got a bum rap because the compilers used for benchmarking didn't generate optimal code. Personally, I think that Intel saw the 68K and changed their mind about selling the thing.

John Dvorak has an article on the 432, while containing a lot of Dvorak, is basically accurate.
 
Agent Orange said:
I'm thinking that 386 was more or less a niche type PC, but was a real step forward in performance for its time. You just didn't see a lot of them in the business world, but the Mom & Pop computer outlets were full of the them. Personally, I like the 386 and have one that I recently acquired (bonus) with a 5160 purchase. Of course I'm speaking of my own experiences in my geographic specific area, and the 386 may have been the horse in other areas.
Click to expand...
You may well be right on the niche thing. I got given a 386DX25 system- full length mobo incl FPU and 8 megs of ram dated from '88. Must've cost a small fortune at the time. Was used for satilite traking using a modem connected to goodness knows where. Not long ago I aquired an early-mid '90s AMD 386DX40 mobo along with a VLB 486 mobo. Looking back 386SXs certainly seemed to be a stop gap, with a vast marketing champian from Intel, also MS/IBM(OS/2) as well as others who wanted to promote there wares. Of course it's all history now but interesting non the less.
 
inakito said:
Chuck(G), the iAPX-432 history sounds a lot like the Itanium!.
Click to expand...

Except, if I'm not mistaken, the Itanium platform is still very much in use in the enterprise server market. It's actually not a bad chip.

Xeon probably will kill it off, but it ain't dead yet.
 
Last edited:
iAPX-432 was an odd unit but not what I was remembering. Intel talked up the idea of 32-bit memory addressing; the iAPX-432 was designed around 64kB segments. The 432 had 32-bit integers and 32-bit bus but the overhead of extra data meant that multiple bus accesses for every instruction. The concept of 300 bit instructions (taking at least 10 bus transfers) doesn't make sense.

In some ways, the 80286 looks like a stripped down iAPX-432 with 8086 compatibility. 80286 has the same maximum of 16MB of RAM in 64kB segments, virtual memory reduced to a mere 2GB instead of the iAPX-432's terabyte, and a simplified protection system.
 
krebizfan said:
as it was, Intel had to recall much of the production from 1985, 1986, and 1987 because early 386 chips did not correctly handle 32-bit instructions. I think the very early 386 also did not have a working Virtual 8086 mode.
Click to expand...
Source? I believe you, just curious to know more about it.
 
Have a look at this http://www.cpu-collection.de/?l0=co&l1=Intel&l2=i386+DX
There's a wiki link about the 32bit issue http://en.wikipedia.org/wiki/Intel_80386

"Intel originally intended for the 80386 to debut at 16 MHz. However, due to poor yields, it was instead introduced at 12 MHz.

Early in production, Intel discovered a marginal circuit that could cause a system to return incorrect results from 32-bit multiply operations. Not all of the processors already manufactured were affected, so Intel tested its inventory. Processors that were found to be bug-free were marked with a double-sigma (ΣΣ), and affected processors were marked "16 BIT S/W ONLY". These latter processors were sold as good parts, since at the time 32 bit capability was not relevant for most users. Such chips are now extremely rare."

The press and competing marketing boys would've had a feild day. That may well have been another reason for 386s not being adopted a lot earlier.
 
Last edited:
cr1901 said:
Source? I believe you, just curious to know more about it.
Click to expand...

Sep 21, 1987 issue of InfoWorld page 6 which points to some 386s labeled as being for 16-bit only and an Intel plan to replace failed chips in the first half of 1988. Other references exist but Google Books online copy of the issue should make it easy to verify.
 
krebizfan said:
In some ways, the 80286 looks like a stripped down iAPX-432 with 8086 compatibility. 80286 has the same maximum of 16MB of RAM in 64kB segments, virtual memory reduced to a mere 2GB instead of the iAPX-432's terabyte, and a simplified protection system.
Click to expand...

In other ways, the 432 was way, way ahead of the curve when it came to architecture, and somewhat limited by current state-of-the-art constraints. Regardless, trying to compare it to a conventional CISC processor such as a 386, is hopeless. The 432 is an object-digesting architecture, with all the bells and whistles thereof. Heck, it didn't even have a general-purpose register file. I think one familiar with other CPU architectures could absorb most of the 80386 in about a day of study; I don't think the same person could achieve the same level of facility on the 432 in a week.
 
You must log in or register to reply here.
Back
Top