• Please review our updated Terms and Rules here

I wonder if the physical size of the AT-bus (i.e. ISA) card was a mistake...

R

ropersonline

Experienced Member
Joined
Jun 17, 2011
Messages
153
It is a relatively rarely-encountered...
Ahem.

"It is a truth greybeardishly acknowledged that a single AT ISA card in possession of maximum dimensions must be the cause of strife."

No, but seriously:
It is a relatively rarely-encountered but potentially significant annoyance to owners of PC/XT-class computers, that when IBM came out with the PC/AT, the ISA cards for those machines were allowed to be taller, by an extra 3/4 of an inch.
This means that some ISA cards—admittedly a minority—that otherwise would fit and work in a PC/XT will not fit in those machines (or clones with like case dimensions). Or at least they won't allow you to put the case cover back on.

I've done some measuring and reading, and I've noticed something:
Many sources, so far as they show the physical dimensions of ISA cards at all, do not mention the card bracket/slot cover, and especially not its size.
The amount by which the narrow bottom end of the slot cover bracket overhangs from the bottom of the edge connector is 3/4 of an inch.
The amount by which an AT card is allowed to be taller than a PC/XT-class ISA card? Also 3/4 of an inch.

So now I'm really wondering: Was the larger size of the AT ISA card a mistake?
Did someone measure from the wrong starting point and/or not communicate their measurements properly until it was too late?

Does anyone have any thought on this – or maybe even know for sure, because you were there?
 
This means that some ISA cards—admittedly a minority—that otherwise would fit and work in a PC/XT will not fit in those machines (or clones with like case dimensions). Or at least they won't allow you to put the case cover back on.
Click to expand...
Maybe exactly that was the purpose? Allowing to make cards that would not fit into a PC/XT, because the card (or its software) was not designed to run on any pre-AT?

It's like quite a few 8-bit PC/XT cards have an overhanging section after the 8-bit connector, which makes it impossible to install them into a 16-bit AT slot. Same thing, just the other way around.

Also, it should be noted that - strictly speaking - the PC and PC/XT don't even have an ISA bus. ISA was first defined as part of the PC/AT.
 
I suspect, that for several cards, the extra real estate was needed. Note also, that IBM wasn't the first to "extend" the height of the 8 bit cards; Compaq was an early adapter of taller cards. I noticed this when I got a couple of the Compaq serial cards and they wouldn't fit in a 5150 slot.
 
The super tall cards from a Compaq portable won’t even fit in their own deskpro. I picked up an OG deskpro with no VDU and noticed that when I tried one from my early portable. There are later, shorter cards with the same features that fit both machines, if you swap brackets.
 
Chuck(G) said:
Well, IBM had to use up their stock of 5160 cases somehow.
Click to expand...
I have not seen any evidence to support that. I cannot even find a trade magazine of the time suggesting it. I guess that long ago, someone on a BBS suggested it, it sounded plausible, and slowly became fact.
 
Something like the XT/286 was effectively inevitable. If IBM didn't make something in that form factor, a cloner would. Improved design integration meant you didn't need the full vastness of the AT board form-factor as designs evolved. If you're a Taiwanese manufacturer primarily selling motherboards to second-tier systems integrators, you may not have the market-shaping power to say "here's a custom small form factor design", so "let's build it to fit the XT cases that glut the market at $29 each" is the obvious answer.

What surprises me is that there wasn't another dimension to this: when people sat down to design their "2nd generation" 286 boards, that seemed like a good opportunity to begin integrating more peripherals. Keep the full-AT form factor, and use the extra space to build on a floppy controller and serial/parallel/game ports. Hard disc controllers might have been a controversy point (people might have wanted ESDI, SCSI, IDE, or MFM/RLL depending on market position and exact era) but the rest could be soldered in place without much risk. You can then mark the board up $50 knowing that people don't have to buy an add-on $75 super-I/O card. Then the next "shrink" would have been to an XT/Baby-AT size and maybe you only offer 3-5 slots instead of 7-8 because you've already consumed many of the needs for extra slots.

It seems like on-motherboard I/O was almost exclusively for big-name OEM systems until the late 486 era.
 
The PCJr had a good number of peripherals included on the motherboard. It flopped. I think that was part of what kept integrated devices out of most of the mass produced budget motherboards. The need for a special case would have limited the market as well.

Doesn't mean there weren't attempts. The Advance86 was one of the earlier efforts to have an I/O panel for a desktop system. The lower pizza box with the motherboard had parallel, serial, video, joystick, and cassette ports while the upper box had drive bays and expansion slots. See OLD-COMPUTERS.COM Museum ~ Advance Technology Advance 86
 
Board shrinking had to wait for LSI chpsets, which the 5170 did not use. Heck, they weren't available for the early 80386 systems (I have one full-sized AT 80386 board with no memory on the planar--it's all on a separate plug-in board. Those LSI chipsets made a huge difference--my Faraday ATease used one of the early ones and had serial (with RS-422 drives) and parallel on board.
When the chipsets did make an appearance, you could find them on XT clones as well.
 
I always thought of LSI as something that had already happened, and the integrated PC chipsets as VLSI. Off by one?
 
It's like UHF--in 1940, UHF was anything above 100 MHz.

Sooner or later, you run out of superlatives. Our stock of pejoratives is similarly limited.
 
I particularly love those mid-late 80's vintage boards that sometimes have literally dozens of PAL/GALs on them, as a half-step between regular 74-series logic and ASICs. I can speak from my own limited experience that just having GALs in your designer toolbox might be enough to eliminate the need for that extra 3/4" of card headroom.
 
Chuck(G) said:
I suspect, that for several cards, the extra real estate was needed. Note also, that IBM wasn't the first to "extend" the height of the 8 bit cards; Compaq was an early adapter of taller cards. I noticed this when I got a couple of the Compaq serial cards and they wouldn't fit in a 5150 slot.
Click to expand...

Zenith also used extended height cards in their Z-150 desktop and Z-160 luggable computers. Both of these machines are based upon ISA backplanes instead of a traditional motherboard. The ISA backplane contained 8 slots. The 1st 3 of these were extended height. These slots were used for the CPU card, video and floppy drive controller. Depending on the particular model these cards may contain other features combined with these base features such as serial ports, parallel ports or system memory.

The first generation of cards for these systems were jam packed with socketed chips and needed the extra space. The combination floppy controller/serial board in my Z-161 has 32 socket chips. 23 74 series logic chips, 6 SN7516 bus transceivers, 2 8250 uarts and the NEC floppy controller chip. As time went on the extra space was used to combine functionality. For instance the CGA and floppy/serial functionality were combined into a single card. The base system memory was combined with the CPU card instead of being on a separate card in a regular sized isa slot.
 
You must log in or register to reply here.
Back
Top