• Please review our updated Terms and Rules here

80386 chipsets

Those are two different questions.
Since a 386 is a superset of the 286, it should be compatible with 286 hardware, not 8088 hardware, if you want to run all software.
Namely, some software will try to auto-detect what system they are running on, and when they detect a 286 or higher, they may make various other assumptions about the hardware as well (all 286-based PC-compatibles are AT-compatible, which is very different from being PC/XT-compatible).

If you just want to run 'some' software, which has no knowledge of any system beyond the original PC/XT, you could cheat a little.
An AT has these things that PC/XT do not have:
- Second 8259A PIT
- Second 8237A DMA controller
- MC146818 CMOS realtime clock/32 kHz timer
- AT keyboard interface

Hey Scali,

Don't you mean second 8259A PIC (and not PIT)? The interrupt controllers are two so you can have more than 8 interrupts (achieved via cascading). But there should still be only one 8253 PIT, correct?

Also I think that on a 386 motherboard (maybe all but early AT compatibles) had some of those chips: 8253, 8259 and/or 8237 condensed in a single chip on board? Makes no difference to programming though.

Regards,
Vlad.
 
My wife used to have a 386 PC when she was still living at her parents, and once I had to fix it. The motherboard was very interesting, as it had the 486 option onboard, besides the 386 socket there was a 486 socket, both at 33 MHz. Now that PC has long gone...
 
Don't you mean second 8259A PIC (and not PIT)? The interrupt controllers are two so you can have more than 8 interrupts (achieved via cascading). But there should still be only one 8253 PIT, correct?

Yes, you're right, that was a typo.
It does have two timers, but the second one is the MC14618, so not an 8253 PIT.

Also I think that on a 386 motherboard (maybe all but early AT compatibles) had some of those chips: 8253, 8259 and/or 8237 condensed in a single chip on board? Makes no difference to programming though.

Probably. Later 8088s also had many chips integrated into a single chip, same with later ATs.
I am not sure what the earliest 386 motherboards looked like, but by the time 386 became mainstream, they were probably all highly integrated already.
Brand name PCs tended to have things like printer, serial, floppy and IDE integrated on the motherboard as well, some even graphics.

This should be the first 386, the Compaq Deskpro 386: http://www.pcjs.org/devices/pc/bios/compaq/deskpro386/
It looks like it doesn't have an integrated chipset yet. You can spot the 8237s, MC14618 and other chips.
 
Last edited:
ALR made the first commercially available 386. Companies sometimes make announcements, and because they're so "big", they're given credit in these instances. People will tell you the Commodore SX-64 was the first color portable. It only was announced prior to the actual release of the TIPPC.
 
FWIW as far as I can tell the Intel iSBC 386AT was the first 386 motherboard available (only to OEMs). According to infoworld 10 Nov 86 p72 they were "released over 2 months ago" (before September). The ALR Access and Compaq Deskpro 386 were released in November. So you could 'roll your own' 386 before compaq/ALR's systems, provided you built about a 1000, an expensive hobby:)

I am assuming ALR and Compaq must have known about this board (or a prototype) as they were working quite closely with Intel at the time and they all wanted to maintain at least some compatibility.

It has no chipset as such, but it does have what I think is known as the AT/32 bus, those very long slots. Mylex also used a compatible slot. Intended for memory but supposedly there is a SCSI controller out there somewhere. Shame this didn't catch on, as it's essentially the same as the Vesa local bus that showed up in ~1992. Also no cache, just interleaved memory. Its probably the most basic 386 motherboard ever, doesn't even work with a 387. Here's mine:
$_57.jpg

Note: CPU is inserted incorrectly. BTW if anyone has a manual or the Intel bios setup, Pls pls pls send it to me.

I am researching early 386 chipsets at the moment. From what i can tell the C&T CS8230 is the first chipset for the 386. - InfoWorld Oct 13, 1986 p8. I think C&T worked on the principle of cloning what already existed, and they were a pioneer of chipsets. From what I know "Chipsets" didn't really exist even for the 286 before '86, that is, in terms of (v)LSI packages.

The Compaq Deskpro 386, had two very different motherboard architectures between the early and 20Mhz version IIRC. I don't know much about the ALR Access.

"386 system w/a 286 chipset" Possibly, but since so much is going to need to be custom built, i would guess nothing that would be called Northbridge in today's language. Maybe some components of a 286 chipset, with an array of custom logic. A 386 system w/a 286 chipset is probably possible with the 386SX, but by the time this came along in 88, (first seen in the Compaq 386S, i believe) SX chipsets were about. I also have the Compaq 386S if you want me to check any components for you.

Andrew

Edit:

InfoWorld May 25, 1987 p56 - Advert for ZyMOS POACH chipset. mentions an add-on chip for their 286 chip set, that turns it into a 386 chip set. Its marketing so I don't really know how accurate this is. It's 'not currently available' so its all a bit iffy....
 
Last edited:
InfoWorld May 25, 1987 p56 - Advert for ZyMOS POACH chipset. mentions an add-on chip for their 286 chip set, that turns it into a 386 chip set. Its marketing so I don't really know how accurate this is. It's 'not currently available' so its all a bit iffy....

According to the Intel sourcebook "Introduction to the 80386" the 386 supports "Dynamic bus sizing". (See section 5-7) According to such an external device can assert a "BS16" signal which instructs the CPU to only use the low 16 bits of the data bus and automatically perform two bus cycles if the data transfer is 32 bit.

Given that it would sort of seem to me that if that's indeed the case you could in principle wedge a 286 "chipset" (assuming here that "chipset" in this sense means a collection of peripherals like the keyboard, interrupt controllers, timers, and ISA bus drivers) into a 80386 with 32 bit RAM by simply adding a circuit that asserts BS16 whenever the CPU's address bus is in the non-RAM areas of the memory map (IE, for the 384k between 640k and 1024k) and/or when doing port I/O. I'm sure that's oversimplifying things a bit (I'm probably crucially glossing over something like DMA) but in principle at least it seems like it shouldn't be that big of a deal to turn a 286 chipset into a 386 one. (Or to design a chipset that works with either the "full" 386 or a 386SX.)
 
well by 86 or 87 808s were rapidly moving away from discrete logic towards something in the way of chipsets, always proprietary I guess. Even the initial Tandy 1000 had custom asics. Not in all cases was it just a consolidation of glue, but I can't say off the top of my head.
 
From what I know "Chipsets" didn't really exist even for the 286 before '86, that is, in terms of (v)LSI packages.

Well, they were rare, that's for sure.
I have three 8088-based systems from 1987, an IBM 5160, a Philips P3105 and a Commodore PC20-III.
The first two use the standard MCS-85 chips, so no real integration.
The Commodore however uses a Faraday FE2010 chip, which integrates most stuff: http://www.datasheetarchive.com/dlmain/Datasheets-8/DSA-151234.pdf
According to the date in that document, it is from 22 November 1985.
That is certainly an early chipset, perhaps even THE first (Commodore also used these on their Bridgeboards for Amiga).
I assume that around that time, similar solutions for 286 would also have been available, but at first they were more exception than rule I guess.

By the way, the Faraday FE2010 is not 100% equal to the MCS-85 chipset. Because of some minor timing differences, 8088 MPH cannot run on it.
 
in principle at least it seems like it shouldn't be that big of a deal to turn a 286 chipset into a 386 one. (Or to design a chipset that works with either the "full" 386 or a 386SX.)

Here's a 386 board with the Zymos 286 chipset after Intel bought them out:

http://www.ebay.co.uk/itm/VINTAGE-3...791879?hash=item21058919c7:g:uK8AAOSwFGNWRPiS

It would have to have been built in '88 or later as it can deal with the 25Mhz CPU. The AUSTEK A38152 is a cache controller. I think its the same as the Intel 385 cache controller. I would buy the board myself but its a bit pricy for me. ~$60 extra for shipping and EU import tax. You Americans are soooo lucky for vintage stuff.


Here is the datasheet for the Intel 82230 Zymos bought design. which I think is the POACH. Perhaps someone who knows more about motherboard design could work out to what extent its features may actually be being used in a 386:
http://pdf.datasheetarchive.com/indexerfiles/Scans-001/Scans-0027218.pdf


From what I can see a 386 built with a 286 chipset is a bit of a matter of semantics. If you take this board above, as I understand it the Zymos chipset is kinda acting as the 'southbridge' if you can apply this term to these old boards. The Austek cache controller is dealing with the RAM. That combined with some of the other chips I would guess forms the 'northbridge'.

So yes it's a 386 with a 286 chipset, but its been heavily modified to the point that the 286 chipset is now part of a bigger custom chipset. the 286 chipset has so many custom parts that IMHO its no longer a 286 chipset.

I don't really know either way, it is, but it isn't.

Well, they were rare, that's for sure...
According to the date in that document, it is from 22 November 1985.

Think I have beat you by a month:p Found this:

InfoWorld Mar 9, 1987 p13 - C&T introduces 12.5Mhz of its 5 chip set first released in October 1985. I *think* this is the CS8220. the CS8220 datasheet I found doesn't have a date.

Andrew
 
Here's a 386 board with the Zymos 286 chipset after Intel bought them out:

http://www.ebay.co.uk/itm/VINTAGE-3...791879?hash=item21058919c7:g:uK8AAOSwFGNWRPiS

It would have to have been built in '88 or later as it can deal with the 25Mhz CPU. The AUSTEK A38152 is a cache controller. I think its the same as the Intel 385 cache controller. I would buy the board myself but its a bit pricy for me. ~$60 extra for shipping and EU import tax. You Americans are soooo lucky for vintage stuff.

Are we? We are short a lot of Amiga and Atari stuff over here now it seems.
 
Are we? We are short a lot of Amiga and Atari stuff over here now it seems.

Most importantly, the Atari, Amiga and C64 hardware you do have there are NTSC, where a lot of software (especially games/demos) were aimed at PAL machines because the European market was bigger.
On the other hand, we never got the PCjr :)
 
Most importantly, the Atari, Amiga and C64 hardware you do have there are NTSC, where a lot of software (especially games/demos) were aimed at PAL machines because the European market was bigger.
On the other hand, we never got the PCjr :)

Yep... less of a problem on the Atari STs and Amigas if you have an RGB monitor, moreso on the C-64. That's one reason I switched out the VIC IIe in my 128D for a PAL one. :)
 
I'm not that sure about us Americaners being thatyt lucky either. There's loads of stuff I want that wawn't sold here.granted it may not be available in the UK either, but that's actually the whole point, evwery country has stuff the other doesn't.

Did the Atari STs see any level of saturation in Japan?

I happen to like the Peanut. Just don't want to hwng out with wll day every day
 
Well, I suppose my focus is mainly on early PC to early Pentium PC stuff. Most of the items I see are for sale in the USA. Simple things like a Full AT case can be very difficult to find sometimes. Not that many on eBay.com, but much rarer on eBay.co.uk. There were two sellers recently selling a memory card for that Intel motherboard I mentioned above, both USA, both £50 when you factor in shipping.

I was never into the Amiga, they always seem to cost more than I'm willing to pay:). I had an Atari ST but got rid of it years ago, as I did with a C64. I suppose we all want what we individually see as rare. The C64 and Atari ST, 10-15 years ago, were everywhere in the UK. Items like early 386 motherboards, EISA systems, even IBM AT's and yes particularly the PCjr you rarely see about in the UK. We do have lots of Amstrad PC1640 though!

I don't know how you get on with import tax into the USA, here it seems to be totally random amounts.
 
I'm not that sure about us Americaners being thatyt lucky either. There's loads of stuff I want that wawn't sold here.granted it may not be available in the UK either, but that's actually the whole point, evwery country has stuff the other doesn't.

Did the Atari STs see any level of saturation in Japan?

I happen to like the Peanut. Just don't want to hwng out with wll day every day


The Atari ST computers were and are pretty much unknown in Japan. A few were owned by musicians who imported these themselves, but the Atari ST was neither officially available in Japan nor demanded. The lack of Japanese character input abilities and Japanese software made these rather uninteresting and hard to use for the typical Japanese computer user. If you find anyone with an Atari ST in Japan they will proudly show off this "obscure" machine to you.
 
What I said there was my answer to what I thought was a reasonable interpretation of this part of your question:



Said question was followed immediately by the mention of the '030 so I logically assumed you were asking if there was a version of the 80386 that had "memory management", which I interpreted as "paging support", as an optional external component, IE, like the 68020, and the answer to *that* is what I said, which is "no, it's always been built-in except for this weird rarely-used 386SX variant". Apparently that wasn't the question you were asking, but in that case what *was* the question? If by "memory management chip" you meant something like a VLSI DRAM controller (vs. a RAM controller made out of discrete logic or PALs), well... that was the very wrong way to ask.

Regarding the Mac II and it's possession of an HMMU, well, so what? First, the HMMU is not a full-featured paging unit. (So far as I'm aware it's essentially limited to remapping the Mac's memory map into... what, two and a half? static arrangements.) And secondly, there were a number of 680x0 machines using both the 68020 and the MMU-less 68EC030-variant of the 68030 that lacked any kind of mapping/paging hardware so it's not a given that any particular Motorola-based machine has a "memory management chip" comparable to the support built into the 80386.

(Some 68020-based machines that had MMUs didn't even use the standard Motorola 68851; Sun-3 workstations were built around their own discrete-logic unit that was simpler but faster than the Motorola one, so it's hard to really say things like the "Mac IIx was something of a standard" and have it mean anything. Motorola-based computers were *all* over the map.)

In fact, I think the 68851 came a few years later than the 68020. BTW, I think the HMMU deals with mapping 24-bit addressing to 32-bit addressing.
 
Back
Top