analog_programmer wrote on 2024-04-07, 08:14:

I'm wondering if I can swap these four 256 K x 4 chips with four 512 K x 8 chips with some disabled data lines and used as 512 K x 4 for a total of 1 MB VRAM (that idea came to me from the six chips version of the card where are used some "low-grade" chips with reduced number for data lines).

I assume you got the idea of using 512K x 8 chips from this card on VGA museum: https://www.vgamuseum.info/images/palcal/umc/ … _umc_top_hq.jpg (mentions 256K x 8 on the PCB) or another card I can not currently find at the moment where it says 512K x 8 on the PCB. Count the pins on those chips: Those "x8" chips have 28 pins, not 26 pins, so they most likely dont have a compatible pinout to the SOJ26 x4 chips used on other cards.

mkarcher wrote on 2024-04-07, 08:27:

I assume you got the idea of using 512K x 8 chips from this card on VGA museum: https://www.vgamuseum.info/images/palcal/umc/ … _umc_top_hq.jpg (mentions 256K x 8 on the PCB) or another card I can not currently find at the moment where it says 512K x 8 on the PCB. Count the pins on those chips: Those "x8" chips have 28 pins, not 26 pins, so they most likely dont have a compatible pinout to the SOJ26 x4 chips used on other cards.

Yeah, this idea came from some pictures from different interent resources and our discussion on old RAM chips yesterday. But first I have to make some research in old chip's packages if there are any DIP-20 compatibles... and this will be another day. You know the slogan about "stupid things" and I don't want to try unconfirmed or potentially impossible projects anymore 😉

P.S. mkarcher, you're right. It seems like there are no 512 K x 8 chips with less than 28 pins. So, my UMC card has no chance to be upgraded with more than 512 kB VRAM. I don't want to mess with some interposer PCBs to adapt pin-incompatible chips.

mkarcher wrote on 2024-04-06, 17:41:

I started to get quite confident, that these chips are B-class (or even worse) 256K x 4 chips, purposefully put into an unusual case to prevent these chips entering the "quality chip" market. I noticed that on the UMC 85C405AF and 85C408AF cards, there seems to be a pattern of having 3 413256J chips or 2 414256J chips, and a jumper field routing 8 data lines from the UMC VGA chip to 12 different positions in the RAM array. This looks like picking "working" or "nearly working" bits. In the 414256J, all 4 bits should be fine, in the 413256, only three of them. See Re: [SOLVED] First attempt at "let's make some 30-pin FP memories"... turned out to "let's not make stupid things again" for details.

hah I love how UMC or whoever was buying UMC chips was targeting bottom of the barrel market with this PCB designed for partially defective chips like Sinclairs ZX Spectrum using defective 64K x 1 as 32K x 1 in upper/lower arrangement 😀 or Chuck Peddles company Re: Strange SIMM ram package

btw does anyone recognize the logo on those supposedly partially working ram chips on UMC cards?

dJOS wrote on 2024-04-06, 22:24:

Cheers, some pics would be helpful.

As promised, the pictures of my card. It's a bit dirty but I'll try to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

pado wrote on 2024-04-07, 21:17:

dJOS wrote on 2024-04-06, 22:24:

Cheers, some pics would be helpful.

As promised, the pictures of my card. It's a bit dirty but I'll ty to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

This picture clearly shows that the 85C408AF does support 1M of video memory organized as 8 chips of 256K x 4. From looking at the 512K cards, I know that those cards use 16 dedicated data lines, making it a single bank of 256K x 16. It would be interesting to trace the traces on that card to find out whether the UMC 85C408 uses a 32-bit memory bus, and accesses the memory as 256K x 32, or whether the chip accesses the memory as two banks of 256K x 16 each.

Anyway, in either configuration, you need traces not present on a 512KB-only card to either select the second bank or connect the missing 16 data bits. Just adding RAM to a PCB not design to accomodate for 1MB video memory is not going to work.

The fact that there are pictures out there showing 1MB as 2 chips of 512K x 8 seems to indicate that the UMC chip is likely able to operate at 1MB RAM with a bus width of 16 bits. This of course does not mean that it might also be able to operate at a bus width of 32 bits. Counting traces might be enough to disprove the 32-bit bus hypothesis, though.

really curious now if adding 512KB translates into any performance gains

rasz_pl wrote on 2024-04-08, 15:17:

really curious now if adding 512KB translates into any performance gains

It shouldnt make any difference, unless adding more ram also increases the memory bus width.

pado wrote on 2024-04-07, 21:17:

dJOS wrote on 2024-04-06, 22:24:

Cheers, some pics would be helpful.

As promised, the pictures of my card. It's a bit dirty but I'll try to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

Cheers, your cards PCB is identical to mine, even down to the footprint labels.

rasz_pl wrote on 2024-04-08, 15:17:

really curious now if adding 512KB translates into any performance gains

It's not about performance, but for higher resolutions/more colors support.

Sometimes is. For example, using RAM interleaving techniques.
But on this particular card/chipset that is not the case and your note checks out.

pado wrote on 2024-04-07, 21:17:

As promised, the pictures of my card. It's a bit dirty but I'll try to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

Does anyone know what the JP5 jumper does?

mkarcher wrote on 2024-04-07, 21:33:

pado wrote on 2024-04-07, 21:17:

dJOS wrote on 2024-04-06, 22:24:

Cheers, some pics would be helpful.

As promised, the pictures of my card. It's a bit dirty but I'll ty to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

It would be interesting to trace the traces on that card to find out whether the UMC 85C408 uses a 32-bit memory bus, and accesses the memory as 256K x 32, or whether the chip accesses the memory as two banks of 256K x 16 each.

Knowing only the very basics of electronics, I don't know how to check this. Do you know an easy way to do this?

pado wrote on 2024-04-09, 23:17:

pado wrote on 2024-04-07, 21:17:

As promised, the pictures of my card. It's a bit dirty but I'll try to clean it as soon as I can.
It uses the same chip model as suggested by @pshipkov, but with a different speed.

Does anyone know what the JP5 jumper does?

It seems to be related to installed ram size, my card didnt work until I removed it.