Repair of a Cirrus Logic VLB Graphics Card

Fellow vintage computer collector, Gammitin, challenged me to fix four VLB graphics cards, as well as upgrade the RAM on them. If I could do it, I could keep one for my 486 PC build. Can I do it? Let’s find out.

The Cards

These are the four cards:

They are Cirrus Logic GD-5429 based VLB cards. If you have never seen VLB before, it stands “Vesa Local Bus”, a predecessor to the PCI bus that can be found on some 386 and 486 motherboards. A 32bit bus that typically runs at 33MHz. It uses the same connector as a PCI card, but the connector sits behind a regular ISA connector. So, you can have ISA and VLB on the same card.

With these four cards, the ISA bus handles the video BIOS and the VLB bus handles the data. The GD-5429 is only 16bit, but it is still a very fast DOS VGA chip for the era.

The only difference between the two types is the package for the RAM chips. There are two RAM banks on each. 1MB soldered on and sockets for an extra 1MB.

Testing

I’m going to cut a long story short here. Three of these cards worked immediately. You need to make sure the motherboard is set to 0 wait state for the VLB bus, or it won’t boot. I also had compatibility issues when anything else was in the VLB bus. But they worked great.

The RAM on all of these is basically standard DRAM, so on the left two cards, two 512KB 8bit SOJs fill the sockets. On the right cards, eight 128KB 4bit chips are used. This is one of the left-hand cards after the upgrade.

Diagnostics

So, it is just the bottom right card from the first image to fix. The first I checked was the BIOS chip. It was an entirely different type to the rest, and it was completely blank. I swapped it with the card above and flashed an AT27C256 with the correct BIOS, which you can see in the photo above.

I popped this into my test 486 VLB motherboard along with a POST card to see what happens.

The four dashes there meant that the card was completely blocking the machine’s own BIOS from executing. Even without the card, you would get some codes showing instead. I tried a few extra things out:

• Initially tested without the RAM upgrade
• Swapping the default RAM bank to the socketed one
• Put the card into a non-VLB ISA slot

The last one of these let the computer boot (but error codes and noises about the lack of VGA chip happened). Which indicates that it is the VLB bus itself that is locking up.

There are five buffer chips between the Cirrus Logic chip and the VLB interface. My current working theory is that one of these is bad, especially since they are 74F-series, which are more prone to failure.

The thermal camera doesn’t give any indication of obvious failures.

Side note: this is why there is a weight sitting on the 486DX2/66 CPU, it acts as a crude heat sink whilst I’m doing thermal testing.

The chips here were a 74F32 and four 74F245 ICs. Which all came out with minimal effort.

I know from experience that my BackBit Chip Tester Pro v2 doesn’t test 74F245 chips properly, so I used my T48 programmer.

The 74F32 passed, but every 74F245 chip fails the tests in the same way. I used a reference 74F245 as well to confirm that this was the problem, and that chip passed.

Replacements

After a 2-3 working day wait, replacement chips arrived. There was a 0 ohm resistor providing ground to the right-hand chip. I removed that and used some thin patch wire instead. This meant that I could attach that ground and use sockets for the chips without it looking too messy.

With everything in place, it was time to test again. I used the same rig as before, but the VGA output was connected to my OSSC. This helps debug if there are any issues, also the VGA port for my test bench monitor was already in-use, and I didn’t want to re-route it.

Success! Although I really need to dust that monitor.

The image looks a little out of shape because either the monitor or OSSC is stretching the image. Not something I’m concerned about at this point. The main thing is, this dead card is working again!

With that, three of the cards are off to Gammitin!