More than two hard disks in DOS(os2museum.com) |
More than two hard disks in DOS(os2museum.com) |
Thats a hardware limit:
Early mainboards only had a single IDE / parallel ATA port. Each port has two pins for drive select, so you had a maximum of two addressable drives, the master and slave drive.
With a secondary ATA port you got another set of master/slave, pushing the limit to 4 drives.
That's where the "primary master" text comes from that showed up on the screen during booting.
Most people I knew with computers prior to 1992 or so either booted from floppy or had less than 40mb hard drives. They were expensive. By the time I got more into the hardware (1994 or so), dual IDE was common (4 devices) and PATA transition was pretty seemless. The only reason I'm even aware of the difference is I worked at iomega for a while, and the IDE zip drive was IDE and not PATA.
Around 2001, I had a motherboard with dual PATA and another PATA that was via onboard raid controller. I had 4hdds, a cd burner and an ide zip at that time. The drives I had first used were the first IBM Deskstar drives... fast, but died very prematurely... the second died before I could RMA the first. I had switched from OS/2 to Windows 2000 (not ME) around that time. Then came SATA, and no more rounding pata cables.
On the TRS-80, they just ordered all the drives jumped with all 4 positions, then pulled out the other 3 unused pins in the connector on the cable.
... like in the PC AT, PC XT[1] or the Compaq DeskPro 386[2] that the article discusses didn't have those ports at all.
Those were instead on ISA expansion cards, just like the floppy controller that would often share a card with the UART controller for the serial interface.
[1] https://theretroweb.com/motherboards/s/ibm-xt-type-5160-64-2...,
[2] https://theretroweb.com/motherboards/s/compaq-deskpro-386-20...
Before that, when you installed a hard disk you had to go into the BIOS to specify the geometry of the drive. 46 types were already defined, to match individual drives on the market. "Type 47" allowed -- required -- manually specifying the drive geometry in terms of cylinders, heads and sectors. So for a short while some traditional MFM or RLL drives would be informally classed as Type 47 because their geometry and capacity differed from earlier drives.
My 486 and earlier systems have all I/O provided by ISA cards, other than the 5-pin DIN keyboard port which was standard since the original PC.
P-ATA IDE didn't arrive until the late 80's.
I'm honestly not sure I had a machine with more than 2 fixed disks until well into the days of Windows 7 and SATA. The exception would be logical disks such as Stacker or similar compressed volumes - but I wasn't using them until later either.
If I recall correctly before SATA we had IDE which only had two devices (primary & secondary) per controller, and usually only two controllers on a motherboard. Given the physical size of disks even you'd probably just have a boot disk, maybe a data disk and then perhaps two optical drives. So it's absolutely believable that nobody found the bug simply because nobody had a machine configured that way.
Sure, you could have SCSI for more disks. But if you did, then you were probably doing something that required a lot of CPU grunt - at which point you might just leave the PC behind and go to a UNIX workstation anyway.
OK, now I'm starting to get flashbacks to just how bad SCSI support was on the PC, and it's stripping the the rose-tint from my glasses. Time to go!
Nobody expected this silly machine to be relevant and affecting computing 44 years later!
It came with 16KB of RAM! 640KB would have been 40 times as much - that's the equivalent of a modern laptop (which comes with checks Apple 16GB of RAM) going up to 640GB of RAM.
The original machine had support for two floppies and a tape drive - the first hard drives were in the 5MB for $2000 in 1982 range. That's about $6,700 today.
Even the writings of the day assumed that the IBM PC would last "for a time" like all previous machines had, newer ones would come out on new chips that were completely different. Nobody really expected backwards compatibility and Windows to eat the world.
- The total address space is 1MB, and that's a CPU architectural limit (which is only "broken" many years later, and in a rather unsatisfying way).
- You need somewhere to map the RAM, and because of CP/M quasi-compatibility, that needs to start at address 0.
- The CPU starts executing code at 1MB minus a few bytes, so your system ROM must go right at the top of address space.
- You need memory windows for the system ROM, options ROMs and 2x framebuffers.
- Bank switching adds extra complexity and 74xx logic chips on a motherboard which is already very busy.
Given these constraints, the 640K limit for RAM, with framebuffers and ROMs mapped at 0xA0000, is the only thing that makes sense.
640GB of RAM doesn't sound unreasonable to me.
I had the original IBM PC with two 5.25" floppy drives, and I think that was all the room there was on the disk controller. Dad bought a 10MB Hardcard to expand it; that went in an ISA slot, if I remember correctly. The disk controller might have been in an ISA slot, too.
I think that pre-AT era would have constrained DOS <5.0 more than the IDE/SATA/SCSI eras.
Not long afterwards I ended up on Coherent OS, fun times.
At one point I did have two hardcards plugged into my Amstrad 8086 machine which felt pretty decadent. (Or maybe it was a hardcard plus the internal hard drive?) In total it wasn't even 100MB of storage. https://en.wikipedia.org/wiki/Hardcard
Plus, as better disks appeared, older disks became much cheaper (or even free in some cases). It was pretty simple to collect a few of older, smaller, drives and stick them all in one machine just to give it more space.
I had to save up to buy floppies in the 80s!
https://en.wikipedia.org/wiki/Expanded_memory#Expanded_Memor...
(And IIR, Bill Gates had plenty of company in denouncing EMS while announcing official support for it.)
Most people first experienced it long after it was seen as a crippled chip, which is not what it originally was.
I blame Intel by having the boot address at FFFF0. That guarantees you need ROM at the top of memory rather than loading boot code and sticking frame buffers in low memory, and starting user RAM at like 0x20000.
Details here: https://theretroweb.com/motherboards/s/dell-system-486p
But, yes, most bog standard machines would've had a separate "SuperIO" card containing serial, parallel, and IDE interfaces until the mid 90s.
That power - phenomenal, especially in the early Internet age.
This kind of thing can even be a problem trying to virtualize DOS stuff: https://www.os2museum.com/wp/idle-dr-dos/
https://en.wikipedia.org/wiki/HP_200LX
30-40 hours but it worked out to basically "it was always good to go".