Watch on YouTube:

1986 IBM 5162 PC XT 286 Restoration Update! - Part 3


A roundup of my findings so far and the decisions I’ve decided to take going forward with this rare and unusual IBM 5162 machine, just before I put it all back together. We’re now running reliably at 10MHz, but I reckon I’ll build this up as a rock-solid 8MHz EGA gaming PC, perhaps with the option to switch clock speeds further down the line.

I’ve been on all sorts of adventures trying to get EPROMs to work so I thought I’d share my findings in case they’re useful for anyone else. Oh, and there’s a very generous donation from Has Beard, Plays Games too!


In parts 1 and 2 I checked out this rare and interesting 286 machine called the IBM 5162, discovered the story behind it, built an adapter to use my Model F XT keyboard with it, and also had a few attempts at upgrading it beyond its measly 6MHz 286 CPU - with varying degrees of success.

I finished on a bit of a cliffhanger last time and I think I’ve kept everyone waiting long enough, so I think it’s about time to sum up what I’ve been up to since that video went out so I can finally put it all back together in the next instalment.

Which should hopefully be sooner rather than later.

So last time I removed the tantalum capacitors and I explained that after 30 years they can wear out and that when they do, they tend to explode. Completely coincidentally, LGR released a video a few days after that that shows what they look like when this happens - and I mentioned at the time that when they do go pop they tend to create a short circuit which can damage various other things, so they’re well worth replacing as a preventative measure while the motherboard is out of the case.

In those intervening weeks between part 2 and now my tantalum caps finally arrived, and as the three legged caps that IBM originally used aren’t easily available these days, I replaced them with the more standard 2 legged version. The inner hole is positive and the two outers are negative, and on the caps themselves there should either be a small + sign to denote the positive leg or, if all else fails, it will be the longer of the two.

So with all that in mind I used a multimeter to check that I had continuity between both outer legs and also ground, and I was careful to make sure that the one cap at C66 that’s rated at 33uF rather than the 10uF of the others got replaced with the correct value.

But lets’ face it, recaps are pretty dull and there are already a million videos out there about them already, so I’m sure the main thing regular viewers want to hear about is my ongoing adventures in CPU performance.

I found previously that, once a few slower components were replaced - including the BIOS which has an artificial clock speed restriction built in by IBM - this machine which was designed to run at 6MHz would boot at 8MHz and even 10MHz (but not 12, that would be silly), but I found that adding an AdLib card would make things unstable at the higher speed, with regular lockups ruining the fun - in my case, Wolfenstein 3D.

Well, I thought that, after replacing the SIMMs with much lower latency versions due to this system not supporting wait states, these lockups might be down to the chips on the motherboard itself which make up the top 128K of our 640K of RAM. That comes in the form of four 256K 4464s (or 41464 in this case), and two 256K 41256s for parity, and the originals were all rated at 150ns. So I’ve now replaced those for some chips with a much lower latency of 100ns.

Unfortunately this didn’t seem to make much difference to the lockup situation with the AdLib card in place. But for some reason - and I’m still not sure what possessed me to try it - I put the 8-bit AdLib card into a 16-bit ISA slot and… Well. It all just works fine at the higher clock speed now. I’m guessing there’s a logic chip or something that handles the 8-bit slots that isn’t happy at 10MHz. With the way the slots are supposedly wired up this doesn’t really make sense but, hey, the proof is in the pudding… I guess?

In my research I was quite surprised to find that ISA doesn’t even have a proper final, published standard. There’s a draft - known as the IEEE P996 specification - but it was never formally finalised. Which probably explains some of the weirdness around it, including clock speed. You see, it’s generally accepted that ISA runs at 8MHz, with machines utilising faster system clocks adding wait states, where the CPU basically waits on every cycle for the slower components to catch up. But some IBM clones did indeed clock their ISA bus at 16 and even an eye-watering 20MHz, so you’d think that 10 wouldn’t be a problem.

On that note, as previously mentioned I also had to change the BIOS to get around an artificial speed restriction that IBM put in to stop people doing just this instead of buying more expensive machines. I initially went with a generic AMI 286 BIOS and it all worked great. But to keep things slightly more authentic I wanted to try out a leaked internal BIOS from 1989 that IBM engineers were using to upgrade their own 5162 machines to squeeze a bit more life out of them.

Well, I tried burning it to various EPROM chips and it didn’t work. So I dumped the original BIOS and wrote that back to the original EPROMs and that didn’t work either. In a more interesting twist, I thought I’d try dumping the AMI BIOS and burning that to some of the chips and it worked on all of them! So it seems neither the original nor modified IBM BIOS work when I burn them using my particular set up, even though they verify and read back just fine. A bit of a conundrum, and even VCF forum users couldn’t work it out, and all the back and forth massively delayed this video, so I’ve given up on that for now and I’ll just stick with the AMI BIOS.

I mean, I made fancy labels for the chips and everything.

Oh, and to top it all off, one batch of EPROMs I got were fake - or at least, rebranded. So always be sure to check the device ID in your EPROM burning software.

The only good thing to come out of all this is that what I have here is a later AMI 286 BIOS from 1990 that doesn’t seem to be available online anywhere, so I’ve uploaded it to for anyone wanting to make their own EPROMs and give it a try in their 286 system. It’s quite nice and modern and has some built-in diagnostics, and it should be generally very compatible so I’ll link that down below.

It’s a bit of a shame that I couldn’t use this opportunity to show the original BIOS configuration process, which on the early IBM BIOSes involves writing a setup disk which I pre-emptively did using a USB device called a Greaseweazel and the original floppy drive from this machine. So I guess I’ll have to save all of that for another time.

You’ll remember that in part 1 I made an adapter to use this XT Model F keyboard with this AT machine, and there has been a rather nice development on that front. Byron from the excellently-named channel Has Beard, Plays Games sent me this rather lovely bolt modded Model M to use with this 5162 - and if you’re not familiar with this mod, it basically involves replacing all of the original internal plastic rivets with much stronger bolts, making this excellent keyboard as strong on the inside at it is on the outside. There’s a series covering this process over on his channel, so be sure to check out the links to that in the usual places and I just wanted to thank Byron for his incredible generosity with this donation and I’m certainly going to get a lot of use out of it on the channel - so thanks, that’s very kind of you.

As far as the PC itself is concerned, I’m going to err on the side of caution and build it all back up as an 8MHz machine - which is still an impressive 30% real world speed increase over the original spec - and I may at some point in future look at building some kind of device to switch clock speeds by switching between the different crystals but this thing has driven me mad for long enough now and I just want to get it back together and get on with using it. Hopefully at least documenting my findings will be useful to anyone playing with one of these in future.

On that note, I’ve decided that I’d like to build the 5162 up as an EGA-era gaming machine. I already have a really nice 486 setup that I can use for those early 90s VGA games so I think this should nicely fill the gap between that and my IBM 5150. As per the 5150 I can also use this with my 5153 CGA monitor here - most EGA games don’t use the extended resolutions that would require an EGA monitor which is something I’ve covered in a previous video - but when I went to power it up it was dead, and it seems the power supply needs some attention so that’s the next thing on the list.

Yup, it’s just never ending with this hobby, but that’s why we love it.

Big thanks to my patrons and channel supporters whose names you can see on screen as I speak - they make projects like this possible. Thanks for watching, I hope you’ve enjoyed this update, I’m now going to go and fix a monitor and reassemble a PC so look forward to seeing those soon and I’ll see you next time.

Support the channel!
Become a Member:

Relevant Links:
Tantalum Capacitors:
VCFed Thread On EPROM Struggles:
VCFed Thread On IBM Internal Leaked BIOS:
AMI 1990 286 BIOS Dump:
EPROM Device ID List:

Further Viewing:
Part 1:
Part 2:
Has Beard, Plays Games IBM Model M Bolt Mod Part 1:
LGR Packard Bell Video:

If you liked this video please consider subscribing to ctrl.alt.rees on YouTube!