analog_programmer wrote on Yesterday, 11:17:

Nexxen wrote on Yesterday, 11:08:

This is the BIOS I found on VGA museum IIRC.

Thanks! This one is identical to my dump (minus 8 KB of empty space at the end). The BIOS from your card contains second OpROM starting at address 0x6800 to the end of the file (I have no idea for what is this one), otherwise it's identical to mine and to the one downloaded from the VGA Museum site. The checksums are correct for both OpROMs in the dump from your card.

So, it seems like the problem is not in the videocard's BIOS chip.

2nd OpRom? I'm intrigued too by this.
Is there any reason for it to be like that? My chip is a OTP if I'm not wrong. At least, the Vcc for Programming is missing and the lead is n/c.
It must have been a factory thing.

This lousy card is a real mystery. 😀

Look at the picture:

Your ROM dump starts with the VGA BIOS part, which is 24 KB in length. Next there is a 2 KB part with "free space" ("FFs"). And finally at address 0x6800 there are also two "magic bytes" for OpROM start "55 AA" and the third byte at 0x6802 is "0c", which means length of 12 sectors (512 bytes each) or 6 KB summary for this OpROM. The last byte for this "second" 6 KB OpROM, which is also the last byte for the entire ROM dump (32 KB = 26 KB VGA BIOS OpROM + 6 KB unknown OpROM), contains correct checksum hex-value for this "second OpROM". I have no idea what is the purpose of this 6 KB "second OpROM", but I don't think it masses with the VGA BIOS i.e. the "first OpROM".

Probably the VRAM is bad, or there's some problem with PLL's frequency generation.

analog_programmer wrote on Yesterday, 13:06:

Look at the picture: […]
Show full quote

Look at the picture:

The attachment Nexxen's_VBIOS.jpg is no longer available

Your ROM dump starts with the VGA BIOS part, which is 24 KB in length. Next there is a 2 KB part with "free space" ("FFs"). And finally at address 0x6800 there are also two "magic bytes" for OpROM start "55 AA" and the third byte at 0x6802 is "0c", which means length of 12 sectors (512 bytes each) or 6 KB summary for this OpROM. The last byte for this "second" 6 KB OpROM, which is also the last byte for the entire ROM dump (32 KB = 26 KB VGA BIOS OpROM + 6 KB unknown OpROM), contains correct checksum hex-value for this "second OpROM". I have no idea what is the purpose of this 6 KB "second OpROM", but I don't think it masses with the VGA BIOS i.e. the "first OpROM".

Probably the VRAM is bad, or there's some problem with PLL's frequency generation.

What a sleuth! 😀
One ram chip is hotter than the other 3, they are lukewarm, the last one is hot.
I'll desolder the ram jumpers and set it for 2 chips (256KB). I can desolder the involved ram chip anyway.

Fortunately the design is easy enough to just inject clock.
At least I can get it off my to-do bucket.

Nexxen wrote on Yesterday, 14:20:

What a sleuth! 😀

BIOS tinkering/hacking is part of my retro computer hobby 😀 I wish I knew better x86 assembly language, so I could read the instructions directly from the hex-code , but I'm already too old to learn that too.

Nexxen wrote on Yesterday, 14:20:

One ram chip is hotter than the other 3, they are lukewarm, the last one is hot. I'll desolder the ram jumpers and set it for 2 […]
Show full quote

One ram chip is hotter than the other 3, they are lukewarm, the last one is hot.
I'll desolder the ram jumpers and set it for 2 chips (256KB). I can desolder the involved ram chip anyway.

Fortunately the design is easy enough to just inject clock.
At least I can get it off my to-do bucket.

This hotter VRAM chip is a clue. If your card is with four VRAM chips, than they are connected to UM85C408AF chip with four 4-bit data lines. On my card the two chips in the second pair are socketed, so I could try to use the card without them and see if it works with only 256 KB of VRAM. I'll check this out and report later...

As for the defective PLL chip (UM9502 is PLL, right?) case, you may just use another compatible chip with some DIY small PCB-adapter, if pinout positions don't match.

analog_programmer wrote on Yesterday, 14:48:

This hotter VRAM chip is a clue. If your card is with four VRAM chips, than they are connected to UM85C408AF chip with four 4-bit data lines. On my card the two chips in the second pair are socketed, so I could try to use the card without them and see if it works with only 256 KB of VRAM. I'll check this out and report later...

I hope so, other cards work on just 256KB.

analog_programmer wrote on Yesterday, 14:48:

As for the defective PLL chip (UM9502 is PLL, right?) case, you may just use another compatible chip with some DIY small PCB-adapter, if pinout positions don't match.

Yep, I'll probably do that. I'd like to buy an adjustable device. There are a few projects out in the wild and they are must in this retro hobby 😀

Thanks for the help so far.

Nexxen wrote on Yesterday, 15:23:

I hope so, other cards work on just 256KB.

I have good news, my friend 😀 Just checked out and my card still work in DOS with the two socketed VRAM chips removed i.e. with only 256 KB of VRAM:

So, you can play for tests with only two chips connected with 4-bit data lines.

Nexxen wrote on Yesterday, 15:23:

Thanks for the help so far.

You're helpful and responsive forum's member, so I am too.

analog_programmer wrote on Yesterday, 15:38:

I have good news, my friend :) Just checked out and my card still work in DOS with the two socketed VRAM chips removed i.e. with […]
Show full quote

Nexxen wrote on Yesterday, 15:23:

I hope so, other cards work on just 256KB.

I have good news, my friend 😀 Just checked out and my card still work in DOS with the two socketed VRAM chips removed i.e. with only 256 KB of VRAM:

The attachment VRAM_chips_removed.JPG is no longer available

So, you can play for tests with only two chips connected with 4-bit data lines.

Nexxen wrote on Yesterday, 15:23:

Thanks for the help so far.

You're helpful and responsive forum's member, so I am too.

GREAT news!
I'll desolder them fast and retest. I'd be great if it worked with no artifacts 😀
Later tonight I'll be on it.

analog_programmer wrote on Yesterday, 14:48:

BIOS tinkering/hacking is part of my retro computer hobby 😀 I wish I knew better x86 assembly language, so I could read the instructions directly from the hex-code

Nobody reads machine code from hex anymore, people did it in the 80ties on ZX Spectrums and C64s because too much time, no internet, no tooling and 'tism.
... so anyway CBh is a retf aka this option rom just returns and does nothing. It looks like garbage. Judging by rest of the card it might be marginal EPROM dies (sides of a wafer) they had to check before programming 😀 thus did a control write at the end and then wrapped in a dummy header.
Garbage does look like burst of calibration data, or maybe even it was a factory defective EPROM to begin with and they salvaged what they could.

analog_programmer wrote on Yesterday, 14:48:

, but I'm already too old to learn that too.

Sure you can and its never to late! Grab free Ghirda (~same UI and even key bindings as IDA), couple YT tutorials and my collection of disassembled Bioses as a template.
x86 assembler really isnt all that bad as people nostalgic for 68K paint it. For me its miles better than what Berkeley came up with RISC-V (no flags, stupid compression, jumbled bits in data formats, minus bit, of the mess).

Nexxen wrote on Yesterday, 16:10:

GREAT news!
I'll desolder them fast and retest. I'd be great if it worked with no artifacts 😀
Later tonight I'll be on it.

Good luck! I'm looking forward to see the results of your memory chip tests. Now I have a brand new idea for memory upgrade to 1 MB (according to VGA museum this UM85C408 chip suppots 1 MB of VRAM).

rasz_pl wrote on Yesterday, 16:10:

Nobody reads machine code from hex anymore, people did it in the 80ties on ZX Spectrums and C64s because too much time, no internet, no tooling and 'tism.
... so anyway CBh is a retf aka this option rom just returns and does nothing. It looks like garbage. Judging by rest of the card it might be marginal EPROM dies (sides of a wafer) they had to check before programming 😀 thus did a control write at the end and then wrapped in a dummy header.
Garbage does look like burst of calibration data, or maybe even it was a factory defective EPROM to begin with and they salvaged what they could.

It is strange to me, because those 6 KB of "garbage code" really look like a valid OpROM - the size and the checksum are correct. I didn't bothered to disassembe this code yet, since it seems redundant to me.

rasz_pl wrote on Yesterday, 16:10:

Sure you can and its never to late! Grab free Ghirda (~same UI and even key bindings as IDA), couple YT tutorials and my collection of disassembled Bioses as a template.
x86 assembler really isnt all that bad as people nostalgic for 68K paint it. For me its miles better than what Berkeley came up with RISC-V (no flags, stupid compression, jumbled bits in data formats, minus bit, of the mess).

I use Ghidra and some old DOS disassemblers like SOURCER and HIEW (honestly, I don't like IDA Free), so I can read some parts in assembly code, but it is time consuming for me. There are still people around, who can see some assembly instructions directly in the hex-code, or rather they're not ordinary people 😀 Unfortunately I'm too far from the level of knowledge of such machine-code/assembly gurus.

Removed the last 2 ram chips and still the same result.

Hsync is 31.9KHz
Vsync is 66Hz

I measured my other ISA and PCI and both have 70Hz and 31KHz. 66 could be the issue?
My screen is flashing the power LED, indicating some error.
Edit: yes, being Vclk below the minimum of 25.175MHz something isn't correct.

Edit: I found http://martin.hinner.info/vga/timing.html
Yes, frequencies are not correct after entering BIOS or F1 to continue to boot.

https://en.wikipedia.org/wiki/VGA_connector

You definitely need two ram chips minimum for correct operation. imo would be safest to remove last one, put two 20 pin sockets and two good working standard 256x4
damn I have sclerosis, I even posted in earlier thread about weird UMC ram before Video RAM chips for a UMC TK-85C408VGA ISA Video Card?

31.9KHz/ 66Hz = 490 lines, weird

>Yes, frequencies are not correct after entering BIOS or F1 to continue to boot.

are they different before doing that? do you have Ami graphical bios on that mobo? noo, even that Ami bios used text mode 3.
ok I tested, boot is 720x400 and almost immediately AMI changes to 640x350, both 70Hz. Both require 28.322 MHz video clock. Hmm how does it do correct Horizontal clock but bad refresh? maybe its using 14.4MHz clock too?
how about you try feeding same 32MHz to both inputs?

Nexxen wrote on Yesterday, 23:23:

Removed the last 2 ram chips and still the same result.

Do you mean "the second" pair of VRAM chips (like I showed with the socketed chips on my card) and not the last two remaining chips on the card?

Nexxen wrote on Yesterday, 23:23:

Hsync is 31.9KHz Vsync is 66Hz […]
Show full quote

Hsync is 31.9KHz
Vsync is 66Hz

I measured my other ISA and PCI and both have 70Hz and 31KHz. 66 could be the issue?
My screen is flashing the power LED, indicating some error.
Edit: yes, being Vclk below the minimum of 25.175MHz something isn't correct.

Edit: I found http://martin.hinner.info/vga/timing.html
Yes, frequencies are not correct after entering BIOS or F1 to continue to boot.

https://en.wikipedia.org/wiki/VGA_connector

So, you're back to PLL related problem, right?

rasz_pl wrote on Today, 01:10:

You definitely need two ram chips minimum for correct operation. imo would be safest to remove last one, put two 20 pin sockets and two good working standard 256x4
damn I have sclerosis, I even posted in earlier thread about weird UMC ram before Video RAM chips for a UMC TK-85C408VGA ISA Video Card?

You're absolutely right.

I have a question for you, since you're much more into electronic spare components/parts, than I am. Are there any suitable 512 x8 memory chips (SOJ or DIP), that could be routed to these four 4-bit data lines (with some kind of DIY PCB adapter), so we could get 1 MB of RAM in total for this card?