The BIOS is a 256 kBit mask ROM. You should be able to read it with any 27c256 setting - but you need to disable device ID verification.

The UM82c4010 on the picture in the VGA museum is obviously pin compatible to an ICS2494, which is a manufacturer programmable clock generator. But there are a lot of different editions of that chip with different programmed clocks. If you install an ICS2494 with a wrong set of clocks, some or all video modes will run at wrong refresh rates. The super VGA documentation collection "VGADOC" states about the ICS2494: "special versions such as the [...] UMC UM9502 exist", which might indicate that this chip uses a custom set of clocks that's not in any kind of standard ICS2494.

mkarcher wrote on 2022-11-29, 20:10:

The BIOS is a 256 kBit mask ROM. You should be able to read it with any 27c256 setting - but you need to disable device ID verification.

The UM82c4010 on the picture in the VGA museum is obviously pin compatible to an ICS2494, which is a manufacturer programmable clock generator. But there are a lot of different editions of that chip with different programmed clocks. If you install an ICS2494 with a wrong set of clocks, some or all video modes will run at wrong refresh rates. The super VGA documentation collection "VGADOC" states about the ICS2494: "special versions such as the [...] UMC UM9502 exist", which might indicate that this chip uses a custom set of clocks that's not in any kind of standard ICS2494.

I need an exact match to replace this 9502. That sucks.
Now I have to change the title with "Do you have a spare UM9502?"

Thanks for the explanation.

if that picture is of how it works at this point then your chip works, mounting backwards usually fries totally and there would be no output without proper clocks
if anything I would look for another ram

rasz_pl wrote on 2022-11-30, 03:47:

if that picture is of how it works at this point then your chip works, mounting backwards usually fries totally and there would be no output without proper clocks
if anything I would look for another ram

It is of how it worked before I made the mistake.

After it's no output at all.

Some ram chips had cracked solder joints, I guess it was probably that too.

After some research I found that this video card wasn't great, budget model.
I'm not crying about it.

Edit: I just put it back in to see if a miracle, but still dead. I see that the crystal has one lead with 3.04V, the other is 0.00V.
Maybe it says something to you.

You can still try a ICS2494 from a different card. Many of those chips have the two standard VGA clocks (25 MHz and 28 MHz) as clock 0 and 1, so standard modes would work at the correct refresh rates, even if the clock chip is not as expected by the BIOS.

mkarcher wrote on 2022-11-30, 14:14:

You can still try a ICS2494 from a different card. Many of those chips have the two standard VGA clocks (25 MHz and 28 MHz) as clock 0 and 1, so standard modes would work at the correct refresh rates, even if the clock chip is not as expected by the BIOS.

If you know where I can find one 😀
I'll keep it there trying to locate one, maybe in time...

It's easy to help if you have the knowledge, thanks to you too.

I'm going through old stuff as I'm closing projects that have no more clues for a repair.
This card hit a wall and I can't move forward in any way. The PAL is busted and there's nobody witha spare to trace it from.

I'd like to give a last check to learn something.
1. check clocks, I do that on the VGA connector or from the chip? or is it from PAL? ---> I'll report
2. desolder RAM, if card is found by monitor (I guess it just checks for +V and ground and taht's it) maybe I just killed the ram? ---> I have a vid card with that ram.
3. UMC UM70C188, the RAMDAC. No idea if it could have died. I presume it is custom for the other UMC chips, non replaceable.

Any input is appreciated.
I'm not betting a dime it is still good 😀 I can spare 30 mins on it to just bin it for spares.

Edit:
https://theretroweb.com/expansioncards?itemsP … Ids%5B0%5D=7735

TRW has a page for it 😀
There is documentation available.

I also found this https://github.com/moogway82/ics2494-replacement

Could be a worthy project.

Sometimes there are simple solutions that are so simple nobody thinks of them. In this case the solution is ... MILFs!
Re: 286 Headland HT12 Mainboard Replica mentioning that MILF clock chip brought back feipoa clock replacement topic, I havent looked there in a while, looked again and Re: Project: Full Can Clock Oscillator Replacement
https://github.com/moogway82/ics2494-replacement
edit: ah, you edited your post with same find while I was writing my reply 😀

rasz_pl wrote on 2026-10-25, 13:33:

Sometimes there are simple solutions that are so simple nobody thinks of them. In this case the solution is ... MILFs! Re: 286 H […]
Show full quote

Sometimes there are simple solutions that are so simple nobody thinks of them. In this case the solution is ... MILFs!
Re: 286 Headland HT12 Mainboard Replica mentioning that MILF clock chip brought back feipoa clock replacement topic, I havent looked there in a while, looked again and Re: Project: Full Can Clock Oscillator Replacement
https://github.com/moogway82/ics2494-replacement
edit: ah, you edited your post with same find while I was writing my reply 😀

Yep, but thanks for the answer!
In 4 years a lot happened. I'd like to make that replacement, or feipoa's. Stuff worth having, being universal ain't bad 😉

I compared my pin diagram guessing and it all makes sense, matching the pins.
The datasheet is really helping.

I have UMC UM85C408AF ISA VGA card with this chip:

What do you need to check? I can try to dump the BIOS, if you need it (I have XGecu T48). Unfortunately I don't have suitable oscilloscope, so I can't measure any clock or data signals.

analog_programmer wrote on 2026-10-25, 13:57:

I have UMC UM85C408AF ISA VGA card with this chip:

The attachment UMC UM85C408AF VGA.jpg is no longer available

What do you need to check? I can try to dump the BIOS, if you need it (I have XGecu T48). Unfortunately I don't have suitable oscilloscope, so I can't measure any clock or data signals.

When you have time if you want to dump the BIOS I'd be glad to compare it with mine.

I probed all the leads on the PLL:

1. 3.4v -- 20. 5V (4.9)
2. 0.379 -- 19. 5V
3. 0V -- 18. 5V
4. 0V -- 17. 0
5. 0V -- 16. 0V
6. 5V -- 15. 5V
7. 0V -- 14. 0V
8. 0V -- 13. 5V
9. 5V -- 12. 0V
10. 0V -- 11. 0V

OSC
LEAD 1 3.419V (connected to pll #1)
LEAD 2 0.379V (CONNECTED TO pll#2)

UMC70C188 (or 178 on others)
(readings are in volts)

1. 0.12v
2. 0
3. 0
4. 1.84
5. 0
6. 0
7. 0
8. 0
9. 0
10. 0.13
11. 0.12
12. 0
13. 5v
14. 0

15. 5v
16. 0
17. 3.6
18. 4.6
19. 4.6
20. 4.6
21. 4.6
22. 3.6
23. 3.6
24. 3.7
25. 5
26. 1.29
27. 1.29
28. 5

The 14.xx crustal has issues, I can't read it and when I read it the MHz vary a lot, from 1 to 45.
It was working but I have many and I'll replace it.

The card is picked up by the motherboard and won't beep missing vid card.

Question: if the ram died, would it behave like this?

Maybe I was hasty again believing there is some life here, could be a dead core...

For a test, could I use a 4-pin crystal, the ones you find in big squares on motherboards?
The one replaced by the PLL from feipoa or atheatos.

Nexxen wrote on 2026-10-25, 16:51:

The 14.xx crustal has issues, I can't read it and when I read it the MHz vary a lot, from 1 to 45.

with x10 probe, right? crystals are very fiddly to read, probe capacity crushes oscillations

Nexxen wrote on 2026-10-25, 16:51:

The card is picked up by the motherboard and won't beep missing vid card.
Question: if the ram died, would it behave like this?

I imagine VGA bios would run a VGA ram test, but thats not a given. Afair some main mobo bioses do run VGA ram test.
Same for running clock - a good vga bios would check if VGA refreshes are firing, lazy one wont.

>For a test, could I use a 4-pin crystal, the ones you find in big squares on motherboards?

yep, question is which clock is the chip requesting?
Your post motivated me to expand https://theretroweb.com/chips/6606 and document all the clocks for all the variants. Quickly looking at my notes UMC UM9502 and um82c4010 matches ICS2494-332 but I forgot to note how I got that 😐 I got nerd sniped on the way and moved to Acer motherboards 😀 hmm searching
... there we go https://theretroweb.com/expansioncards/s/gain … d-cardex-vl-1av those two used on same card.

edit: nohmm you would need to inject two clocks, Mclock for ram, this one is very forgiving anything in 30-50MHz should work, and video one but no idea what it might want, 30MHz? early cards used 25.174 28.332 36 MHz

No problems, Nexxen, tomorrow I will dig-up my card from the storage to dump its BIOS and attach it here, since I see, that there are no any BIOSes uploaded on TRW site for this ISA VGA cards.

Meanwhile, for testing you can try to set another frequencies by JP1 and JP2 near the UM85C408AF UM9502 chip according to setting printed on the back side of the card from this picture: https://theretroweb.com/expansioncard/image/i … e0263678473.jpg

By the way, this card is not very good as an option for DOS gaming, since it lacks any VBE implemented in its BIOS. VESA modes are supported through DOS TSR/"driver" only. But probably it's worth to try saving it from scrapping, if the problem is caused by something, that could be fixed easily.

EDIT: I mean UM9502 chip, not UM85C408AF.

rasz_pl wrote on 2026-10-25, 19:00:

with x10 probe, right? crystals are very fiddly to read, probe capacity crushes oscillations […]
Show full quote

Nexxen wrote on 2026-10-25, 16:51:

The 14.xx crustal has issues, I can't read it and when I read it the MHz vary a lot, from 1 to 45.

with x10 probe, right? crystals are very fiddly to read, probe capacity crushes oscillations

Nexxen wrote on 2026-10-25, 16:51:

The card is picked up by the motherboard and won't beep missing vid card.
Question: if the ram died, would it behave like this?

I imagine VGA bios would run a VGA ram test, but thats not a given. Afair some main mobo bioses do run VGA ram test.
Same for running clock - a good vga bios would check if VGA refreshes are firing, lazy one wont.

>For a test, could I use a 4-pin crystal, the ones you find in big squares on motherboards?

yep, question is which clock is the chip requesting?
Your post motivated me to expand https://theretroweb.com/chips/6606 and document all the clocks for all the variants. Quickly looking at my notes UMC UM9502 and um82c4010 matches ICS2494-332 but I forgot to note how I got that 😐 I got nerd sniped on the way and moved to Acer motherboards 😀 hmm searching
... there we go https://theretroweb.com/expansioncards/s/gain … d-cardex-vl-1av those two used on same card.

edit: nohmm you would need to inject two clocks, Mclock for ram, this one is very forgiving anything in 30-50MHz should work, and video one but no idea what it might want, 30MHz? early cards used 25.174 28.332 36 MHz

Wow, nice job here.
Thanks a lot!

Mclock I have all frequencies up to 50: no issue with a 30-32.
Vclcok I'll go with 25, but I have a lot of crystals to try.

I'm trying to understand if the "chipset" is selecting, via 2 jumpers (JP1 and JP2) the right frequency.
70ns ram chips = 14.318MHz, like the crystal already on.
My ignorance.

analog_programmer wrote on 2026-10-25, 19:21:

No problems, Nexxen, tomorrow I will dig-up my card from the storage to dump its BIOS and attach it here, since I see, that there are no any BIOSes uploaded on TRW site for this ISA VGA cards.

Meanwhile, for testing you can try to set another frequencies by JP1 and JP2 near the UM85C408AF chip according to setting printed on the back side of the card from this picture: https://theretroweb.com/expansioncard/image/i … e0263678473.jpg

By the way, this card is not very good as an option for DOS gaming, since it lacks any VBE implemented in its BIOS. VESA modes are supported through DOS TSR/"driver" only. But probably it's worth to try saving it from scrapping, if the problem is caused by something, that could be fixed easily.

Take your time, but really thanks!

---------------

When removing the PLL:

socket pins:
all = 0v but:
12 - 0.15/16V
13 - 5v
15 - 5v
17 - JP1
19 - 0.15v
20 - 5v

Nexxen wrote on 2026-10-25, 23:09:

I'm trying to understand if the "chipset" is selecting, via 2 jumpers (JP1 and JP2) the right frequency.

yes, 2 for Memclock, 4 for video. But its not as simple as setting tow logic levels, VGA chip also needs to operate STROBE to latch settings into PLL
dosdays had an article in June about the replacement project and some experimentation https://www.dosdays.co.uk/topics/retro_review … m85c408_pt2.php
Interesting bit:
"I swapped out the UM9502 on the UMC card with the ICS2494 from the ET4000 - it POSTed just fine but never displayed anything on-screen. Subsequent starts then produced the Long-Short-Short-Short POST beep code which means "display card issue"."

But sadly although he had an oscilloscope on the bench he didnt measure HSync nor MemClock (with ICS 9294-237 replacing UM9502 on UMC card) when the card somehow POSTed. He only measured them in their own cards. He measured chip pins while both sat in ET4000 card.
ICS 9294-237 has very low MEMclocks, but very high VCLK, especially the first three basic ones:
VCLK 0: 50.350
VCLK 1: 56.644
VCLK 2: 65.000
VCLK 3: 72.000
VCLK 4: 80.000
VCLK 5: 89.800
VCLK 6: 63.000
VCLK 7: 75.000
VCLK 8: 25.175
VCLK 9: 28.322
VCLK A: 31.500
VCLK B: 36.000
VCLK C: 40.000
VCLK D: 44.900
VCLK E: 50.000
VCLK F: 65.000
MCLK 0: 40.000
MCLK 1: 41.612
MCLK 2: 44.744
MCLK 3: 50.000

Great thing is he DID post "The table below shows the VCLK and MCLK frequencies while the machine was outputting graphics in different modes." which should be great. In theory we should be able to corelate matching frequencies to UM9502. Lets see
1 First thing I see he switched VCLK (P#19) and MCLK (P#12) for ET4000, happens, good thing those clocks are so distinct it was easy to spot 😀
2 Second thing is he picked unfortunate resolutions, almost all are running on same 640x400 clocks 🙁 with clock and line doubling turned on/off, making his "The fact that VCLK or MCLK don't change when different graphics modes are selected indicates we either have a faulty clock generator or a faulty chipset." conclusion wrong. Frequencies dont change because all those resolutions use same clocking 😀
3 UM9502 frequencies are low, VCLK 28.4 MHz and MCLK 50 MHz. No wonder ICS 9294-237 didnt work in UMC card!

28.4 MHz probably meant to be 28.322 = VCLK 1, which for ICS 9294-237 is 56.644 and in dosdays table measured at 56 MHz. Things match.

For text modes to work you will need hmm, text mode is 720 x 400 at 70Hz soooo 28.322 MHz, so pretty much same VCLK 1.

tldr Stick closest you got to ~28.4 MHz into pin 19, and ~50 MHz (more lenient, probably anything 33-50 will do) into pin 12

found it 😮 dosdays article was super helpful. He was torturing https://www.dosdays.co.uk/topics/images/my_vi … ga_card_413.jpg which is https://theretroweb.com/expansioncards/s/unknown-h52pt-2024 with barely faintly visible ICS ICS2494 so I googled the model number and bam https://www.ebay.com/itm/324695209123 and another seller form Poland so more believable 😀 https://www.ebay.com/itm/147485156755 both ICS2494-253!
ICS2494-253 is dedicated for NCR 77C22E, direct replacement ST49c214-6. Table:
ICS2494 -253
ST49c214 -6
VCLK 0: 25.175
VCLK 1: 28.322
VCLK 2: 40.000
VCLK 3: 65.000
VCLK 4: 44.900
VCLK 5: 50.000
VCLK 6: 130.000
VCLK 7: 75.000
VCLK 8: 25.175
VCLK 9: 28.322
VCLK A: EXTRN
VCLK B: EXTRN
VCLK C: 60.000
VCLK D: 80.000
VCLK E: EXTRN
VCLK F: EXTRN
MCLK 0: 32.900
MCLK 1: 35.600
MCLK 2: 43.900
MCLK 3: 49.100

Yep, that makes sense, much lower early frequencies. Imo this is your table.

EDIT: but there is caveat. I missed the PAL! 🙁 You mentioned PAL at the start but linked to a picture of card without one so Im confused.
All ICS2494-253 using cards do have the PAL, no idea whats the purpose, maybe to map frequencies by manipulating VCLK on the fly so ICS2494-253 output matches UM9502? Definitely to manipulate clock itself, PALs arent suitable for implementing PLLs, can only do basic division by 2, 4, 8. Even multiplication wouldnt work so deffo used to map VCLK states and maybe handle Strobe I mentioned earlier. You will have to trace Strobe pin if its conencted at all