About EEVblog

Check Also

EEVblog #1334 – Mystery Dumpster Teardown

Mystery dumpster teardown time! With the most amazing mechanical mains power switch you’ll ever see! ...


  1. I think that’s an MFM hard disk interface, isn’t it? I don’t think there’s any way to connect an SSD to that.

  2. IBM used a “buckling spring” key design in their “Model M”keyboards. The legendary keyboards have their own Wikipedia page. http://en.wikipedia.org/wiki/Model_M_keyboard Unicomp still makes them for around $95 USD.

  3. I’m not sure that has a standard IDE controller so you might not be able to use an SSD there unfortunately.

    As for the keyboard, I still use IBM model M boards at home and at work. Love them 🙂

    • Pretty sure it’s a standard IDE. I kind of recall getting the Daewoo drive working at one stage.

      • Fair enough. I suppose I’m thinking of one of my old desktop PCs of a similar age. I think the drives went through a seperate controller of some sort. Obviously a laptop would have to have it all on board.

        Man, my memory is really scratchy about all the different things you had to do to get systems of that vintage up and running. Thank goodness we don’t have to tell BIOS cyl, head, sector numbers anymore. And good riddance to manually setting IRQs.

  4. That hard drive is a standard IDE PATA drive. It’ll work just fine if you replace it with a PATA drive (under 540MB due to BIOS limitations back then). Of course, PATA SSDs are VASTLY overpriced. You probably will have better luck using a PATA to CompactFlash adapter. Again, smaller than 540MB due to BIOS limitations. (And no, you can’t use a USB-PATA adapter – those things require the drive be able to be put in LBA mode, and LBA mode came into play I think for drives larger than 8.4GB.

    I’m guessing the PLL board is the central clock chip for the entire system – for the CPU, graphics and other chips. It’s a separate module so IBM could sell different models at different speeds by replacing the PLL module.

    The interesting thing about the 80x86SX chips were that the FPU versions (80×87) were actually full fledged chips – i.e., the 80387 chip was actually an 80386 with an addon FPU on die. Plugging it in basically isolated the old CPU and everything ran off the new CPU. The IIT chips though, if I remember, were fast, but they were not completely compatible – the FPUs didn’t produce the same results as the Intel chips which did cause a few problems back in the day.

    The primary battery is probably for the RTC and CMOS SRAM memory. The rechargable battery is for battery change protection – what happened was when the battery ran low, you shut the lid, the system suspended, and then you could swap the battery. The rechargable kept power going to the memory and the CPU registers for about 5 minutes or so, enough to swap batteries.

    This computer is also one of the last where the keyboard was at the edge closest to the user. The Apple PowerBook 100, released around the same time, pushed the keyboard to the back (where it’s been since on every laptop released since then with the mouse in front of it the keyboard).

    • Oh, and base 10 units. the 540MB, 8.4GB and 136GB BIOS limitations are in base 10. I think the binary equivalents would be 512MiB, 8GiB and 128GiB.

    • That’s the 80486SX / DX you’re thinking of. The SX had no FPU. The upgrade 80487 “FPU” was a full 486 with FPU that fitted in an FPU socket but actually took over as the full processor.

  5. Brings back old memories. Before math co-processors became ubiquitous, the floating point math for Microsoft basic had an overflow error. I discovered it originally on the old TRS-80 and I periodically checked on IBM PCs and they still had the error. Simple program to demonstrate.

    10 A = 1E38

    20 B = SQR(A)

    30 C = B*B

    What you would find is that the assignment in line 30 would fail with an overflow error. But obviously that line shouldn’t overflow since the actual maximum value is about 1.7e38 and the assignment in line 30 should be only 1e38.

    The underlying error is that when the floating point exponents are added (because of the multiplication) the result would overflow by 1. However, the significand when multiplied, would result in a denormalized value that needed to be shifted 1 bit and then the exponent would be decremented causing it to once again be within range. But they didn’t bother to notice the near overflow and instead reported it as soon as the overflow happened with the exponent was added.

    Thankfully, that error does not exist in the floating point circuitry.

  6. Traders do pay direct and oblique costs.

Leave a Reply

Your email address will not be published. Required fields are marked *