Well, I have two dead ISA VGA cards as well. I have not yet the time to check on them and also have only a few ideas:

* measuring all resistors to see if they match their markings
* ESR measurements of all capacitors, to see if they degraded
* Test the diodes for correct forward voltage drop
* Test the oscillator if it produces the correct frequency or if it is broken
* Inspect all solder joints and retouch them if they seem iffy
* Test continuity between traces / pins with a meter

The next culprits might be the DRAM ICs, but they are probably hard to test. Maybe this needs a scope and someone with experience. The other ICs are also then to be tested.

Have you tried removing and reseating the other socketed chips a few times each aswell? Not that I've ever seen bad socket connections resulting in those symptoms, but it's easy enough, and might be worth a try at least 😀

That kind of looks like bad a DRAM chip, since the machine boots and you get things happen the ROMs and ISA side of things works.
What happens when you run some VGA game ?

I never thought of trying games.

I don't actually have any VGA games to test at present, as I'm currently working with an XT. The only other PC I have with an 8-bit ISA slot is my 286, so I could try a VGA game on that later.

But CGA games I've tested all appear pretty much like this:

The text in games is readable unlike DOS text mode, but the screen has those white vertical stripes which flicker multi coloured speckles when there's movement and activity on screen. I caught some of the speckles in that image.

EGA gaming seems to fair better, no flickering on it's vertical stripes:

Edited to add: I tried the EGA version of Mini-Putt also, which has slight flickering in dark green vertical stripes.

Beware: I have little knowledge of the inner workings of graphics cards, but in general, the 74LS series ICs can (still) be found for little money
in the average electronics shop or radio store "around the corner". If one of these ICs is defective, it could explain the odd behavior (partially, at least).
So if you're lucky, the main chip is still intact. Perhaps it's just a slight damage on the RAMDAC or one of the DRAM chips, not sure.
Maybe it is worth replacing some of the ICs with sockets and see if installling new parts could cure the card.. 😀
It's just a suggestion, of course. Personally, I wouldn't throw such a card away before giving it a try.
It's an ET4000, after all. A legend. That beeig said, ISA cards are generally becoming rarer these times.. 😐

Edit: Found some utilities for the original ET4000.
Maybe they come in handy for testing and do work with the AX model, too. 😀

We need someone to design a nice VGA core on a FPGA...

root42 wrote:

We need someone to design a nice VGA core on a FPGA...

Absolutely! Though an ATMega-based VGA or CGA/Hercules card would be a feat, too.
Even if they would do initially support the basic video modes only, it would be fine.

Edit: "Fine" in the senses of a start. In the past years, a lot of VGA/CGA emulation code (C++) was written
that *could* be transfered somehow into a a microntroller. Flea86 comes to mind, which ran a complete
XT-class machine -with VGA CRTC!- on an 8051 derivative.

Edit: Just to make it clear: I have nothing against FPGAs.
I just mentioned the ATMegas because they are common and cheap (less than 5€ each).
For a proper, authentic VGA replica, an FPGA might be way better, actually.
I just thought of the perspective of someone who is used to build stuff on a veroboard,
rather than designing sophisticated PCBs on a PC. 😁

The pictures definitely suggest the fault being in some data line(s) of VRAM bus. Straight vertical lines at regular spacings. Address issue would manifest in repeating graphics.

MCU wouldn't make a very good video card, there's no CPU time to handle both data writes and displaying the image. FPGA approach is definitely the most sensible thing and something I'd like to do in future but my hands are currently full with other projects. Zero wait states ISA and VLB GFX cards are in my todo list, to get maximum performance possible from the given bus.

Ah, I see. I was thinking of a high-level approach. Just like some "compatibles" in the 1980s where emulating CGA on the BIOS level,
without the underlying Motorola 6845 hardware. Just the graphics modes, the bit planes/video memory layout and some dummy registers.

Emulating the whole CRTC in real time would be way too much for such a simple MCU, of course. 😅