Hmm not sure if that old CGA card will work proper in a newer AT. I have an Everex Edge II (8bit MD/CGA) that works in XT, AT and even my 386 but it is from 1990+/-
There are no switches on the XL24 to lock CGA mode so it must use auto-switching and does appear to run the horizontal too fast..

AgustinCordes wrote on 2024-07-20, 23:35:

Hi all!

I need assistance to confirm if an original IBM CGA card is supposed to work in a 286 clone motherboard.

The original CGA is supposed to work in any computer with ISA slots. There are videos on YouTube about an original CGA in a 486- or Pentium-class mainboard at up to 16MHz bus clock, and it still worked. The bus clock is not a severe issue for the CGA, as it generates wait states on any access. The CGA doesn't count bus clocks for the memory access timing, but works on a bus-clock independent 14 MHz clock, so it is quite insensitive to "modern computers".

AgustinCordes wrote on 2024-07-20, 23:35:

This is a VLSI in good shape that boots fine with those common Winbond cards and an ATI Wonder XL24. However, when I plug in the CGA, nothing happens -- not one beep.

If you have tools to confirm it, check the 14.31818MHz signal on the ISA connector. If that clock is missing, the CGA card will lock up the bus on the first video memory access, which would fit your symptoms.

AgustinCordes wrote on 2024-07-20, 23:35:

BIOS is set to COL 40.

That's a correct setting. The AT BIOS and all compatibles enable the integrated CGA support if you set the BIOS to COL 40 or COL 80.

AgustinCordes wrote on 2024-07-20, 23:35:

The 5153 seems to be in working condition too, but the XL24 may be too much for it, or maybe I need to configure it with VINSTALL. See the video -- the horizontal refresh is wrecked, but I suspect it's a combination between the monitor and the card.

The horizontal sync is working perfectly in your video. The vertical oscillator in the monitor is out of sync. The monitor has a V-HOLD adjustment on the back side (no need to open it) to coarsly tune the vertical oscillator. Likely adjusting the knob will make it work with the XL24.

AgustinCordes wrote on 2024-07-20, 23:35:

Is it possible it [that is: the CGA card] was only meant to work with original IBM hardware?

No, not at all. Before IBM went proprietary with their PS/2 line of machines, there was no intention to make IBM cards limited to work with IBM computers only.

Looks like the IBM is missing the character rom u33...when comparing other pictures.

Horun wrote on 2024-07-21, 21:07:

Looks like the IBM is missing the character rom u33...when comparing other pictures.

Yes, I don't see any pictures of this card online that show that socket unpopulated, and if the characters are in ROM, that is why the display is initializing but there is nothing showing on the screen. There are no characters to display without the ROM installed.

Horun wrote on 2024-07-21, 21:07:

Looks like the IBM is missing the character rom u33...when comparing other pictures.

That is likely the least important chip that is missing. The card on the photo also shows U5 (74S174) and U31 (74S08) unpopulated. B e careful: The "S" chips are faster than "LS" chips. LS is not a suitable replacement. 74Fxxx or 74HCTxxx will likely do as well. The missing character ROM will create a completely white screen (or whatever the foreground color of the characters is) in text mode, but will not prevent booting. I didn't check the IBM CGA schematics what the logic chips are used for, but the missing logic chips might cause a "not even beep" symptom if they are involved in the pixel/character clock generation ("wait for blank" will block if there is no pixel clock), or if they are involved in the bus timing/waitstate generation process.

I have ran my old-style IBM CGA card with a Tualatin Pentium 3 without major issues:

Re: What retro activity did you get up to today?

As for the stability issues with trying to boot into FreeDOS that I mentioned in the original post, it turns out the Apollo 133T on the mobo can't do 133FSB properly. I have a micro-ATX Apollo 133T mobo that doesn't have this issue at all, and the CGA card works great in that.

Is it still possible to use composite mode?

Anonymous Coward wrote on 2024-07-23, 01:56:

Is it still possible to use composite mode?

Mostly yes. As the whole CGA card does, also composite mode relies on the 14.31818MHz signal on the ISA OSC pin, but as the NTSC color carrier is directly derived from it, the composite color circuit is most sensitive to inaccuracies of this frequency. The frequency is available on all ISA systems except maybe some odd 80s clones. But there might be an issue: On the original PC/XT, the frequency of that signal was fine-tunable by adjusting a varicap on the mainboard, and you could use it to make the composite colors "look right". Modern computers don't provide a way to fine-tune this frequency, so if the composite color doesn't work well enough, bad luck.

I’m not sure if the old full CGA card I had was an ibm but it had an issue that the back part of the card hit ICs on the 386 I tested it in so it wouldn’t seat

rmay635703 wrote on 2024-07-23, 20:11:

I’m not sure if the old full CGA card I had was an ibm but it had an issue that the back part of the card hit ICs on the 386 I tested it in so it wouldn’t seat

I also can't tell you whether your card was original, but the issue you talk about applies to the original card as well as some clone cards. The photo in the OP shows that there is no problem in the AT clone board, though (at least as long as the SIPP sockets won't get populated).

Thanks all for the replies! Sorry for taking long to come back, I had notifications disabled.

mkarcher wrote on 2024-07-21, 18:44:

The horizontal sync is working perfectly in your video. The vertical oscillator in the monitor is out of sync. The monitor has a V-HOLD adjustment on the back side (no need to open it) to coarsly tune the vertical oscillator. Likely adjusting the knob will make it work with the XL24.

I'm happy to report it was the V. Hold indeed. I feel like a total dork 🫠

The XL24 is performing admirably!

rmay635703 wrote on 2024-07-23, 20:11:

I’m not sure if the old full CGA card I had was an ibm but it had an issue that the back part of the card hit ICs on the 386 I tested it in so it wouldn’t seat

I was careful to ensure the card is plugged correctly and no parts of it are in contact with the motherboard.

The only suspicious thing that I notice, other than the missing ROM, is this wear around the lanes in the back. At least one seems to be in 'danger' zone.

mkarcher wrote on 2024-07-22, 16:50:

That is likely the least important chip that is missing. The card on the photo also shows U5 (74S174) and U31 (74S08) unpopulated. B e careful: The "S" chips are faster than "LS" chips. LS is not a suitable replacement. 74Fxxx or 74HCTxxx will likely do as well. The missing character ROM will create a completely white screen (or whatever the foreground color of the characters is) in text mode, but will not prevent booting. I didn't check the IBM CGA schematics what the logic chips are used for, but the missing logic chips might cause a "not even beep" symptom if they are involved in the pixel/character clock generation ("wait for blank" will block if there is no pixel clock), or if they are involved in the bus timing/waitstate generation process.

So the missing ROM could well be the culprit here? I'll ask the guy who sold the card to me if by any chance he has it.

In fact, now that you mention it, two chips seem to be missing 🤔

Sorry, I can seem to edit posts.

Two chips were mentioned in the thread as being missing, but it's actually three. Here's a new picture.

I wonder if suitable replacements can be found these days.

AgustinCordes wrote on 2024-07-24, 18:49:

I wonder if suitable replacements can be found these days.

I guess you missed this

mkarcher wrote on 2024-07-22, 16:50:

Horun wrote on 2024-07-21, 21:07:

Looks like the IBM is missing the character rom u33...when comparing other pictures.

That is likely the least important chip that is missing. The card on the photo also shows U5 (74S174) and U31 (74S08) unpopulated. B e careful: The "S" chips are faster than "LS" chips. LS is not a suitable replacement. 74Fxxx or 74HCTxxx will likely do as well. The missing character ROM will create a completely white screen (or whatever the foreground color of the characters is) in text mode, but will not prevent booting. I didn't check the IBM CGA schematics what the logic chips are used for, but the missing logic chips might cause a "not even beep" symptom if they are involved in the pixel/character clock generation ("wait for blank" will block if there is no pixel clock), or if they are involved in the bus timing/waitstate generation process.

edit: wrt character rom
https://minuszerodegrees.net/rom/rom.htm <- do a search for character , there's only one line, it has 2 links, one with info one which 'chip' to use and one with the character data.

weedeewee wrote on 2024-07-24, 19:05:

AgustinCordes wrote on 2024-07-24, 18:49:

I wonder if suitable replacements can be found these days.

I guess you missed this

You're right, I missed that bit. So 3 chips are missing:

U5 and U31, logical? I can find these locally.

U33, character ROM set. Way harder to get, at least around these parts. Sigh.

weedeewee wrote on 2024-07-24, 19:05:

edit: wrt character rom
https://minuszerodegrees.net/rom/rom.htm <- do a search for character , there's only one line, it has 2 links, one with info one which 'chip' to use and one with the character data.

Just saw your edit, extremely helpful, thanks.

AgustinCordes wrote on 2024-07-24, 19:24:

weedeewee wrote on 2024-07-24, 19:05:

edit: wrt character rom
https://minuszerodegrees.net/rom/rom.htm <- do a search for character , there's only one line, it has 2 links, one with info one which 'chip' to use and one with the character data.

Just saw your edit, extremely helpful, thanks.

https://store.go4retro.com/2364-adapter/ the link from the info link on minuszerodegrees an adapter for a 27xx or 27cxx eprom

read the reviews 😀