These vram IC added is EDO hence not working, you need FPM VRAM SOJ.

This one you need:
https://www.ebay.ca/itm/123111558436?epid=151 … 7oAAOSwZz9a6ifB

Cheers,

Looking at the pinouts for the two, I think they're different enough that could be why it's not working.

The main thing I noticed is that the OKI is "Hyper page mode" or EDO, while the Hitachi is just regular FPM. The picture with all the RAM soldered shows NEC chips, those could usually be UPD424260 I think? But the NEC UPD424260 and the M514265BSL do have UCAS/LCAS lines, for 8-bit mode data transfer. So the pinout's a little different from the Hitachi, which only has one CAS line.
Looking here: http://www.amoretro.de/2012/09/ati-graphics-u … rafikkarte.html - that's actually the UDP42S4170, which is mostly the same but a bit earlier. The 424170 only has the one CAS line on the SOJ-40 chip, so it's probably the best chip if the Hitachi chips aren't available.

Cheers guys!

In regards to the ram mentioned above, being in the UK, the closest I can buy for a reasonable price would be either of these:

https://www.ebay.co.uk/itm/2-Pcs-DRAM-Memory- … 0-127635-2958-0

Or

https://www.ebay.co.uk/itm/HM514260DJ6-262-14 … 7-127634-2357-0

Neither are perfectly as recommended by you guys but seem very similar. Are either any good for me? Thanks again.

I just thought of something, those OKI chips might work if you tie the UCAS & LCAS pins together on the SoJ. Maybe. What I'm thinking is, if the card only has one of the two CAS lines that the OKI chips need, maybe it's only getting 8 bits out of 16 bits.

The two that you've linked are FPM, so they're worth a go.

I Think they might have the same problem as the OKI chips. In that they also have the UCAS & LCAS line differences from the original Hitachi chips, which just have the one CAS line. You can test that for yourself non destructively by checking with a multimeter (or visually) whether the SoJ socket's pins 28 & 29 are tied together on the PCB. If those two pins are connected then any of the 256k x 16 SOJ 40 FPM chips should be fine.

It seems like the end numbers, 4170 vs 4260 mean something different in terms of possible arrangement which might explain the different pinout on SoJ that you'd think were standard. The 4170 chips don't seem very available though.

------
Edit, well this was a rabbit hole. Looked up datasheets that mention UWE & LWE since that would be similar to your original chips. Found the "(HM514270,HM51s4270)" seem to be both available and have the same pinout.

Thx @Thermalwrong oddly though reading the datasheet it says that the 'S' variation designates fast page ability and that particular version doesn't seem to hit up any results on eBay, but the non-S varient does. I'm assuming I'd need the S varient. By the way I'll check those cas lines as suggested in a bit.

I would buy some cheapo PCI Mach64 and steal the chips from that. Where I am you can get them for 5€ range all days long and most have 4 chips to harvest, often 2 in a socket already.

In general, cards with sockets should not need any modding to apply additional memory.
They made the effort to install those sockets, why would they omit some two bit two cent component?

Then, look up the fine print on the existing ones and get that - meaning chips of the same geometry, like 256k x 16 in this case.
Didn't check, but the numbers on those Hitachi ones read ...8 which probably indicates 80 ns access time, meaning rather basic ones for the era.
Since better ones in addition won't make that thing any faster, it might allow for cheaper parts.

Sourcing parts: There's an obvious marketing bias between
"semiconductor xyz cryptic technobabble"
for engineering and
"RAM upgrade for your prized vintage ATI Mach32"
in the classic computer department.

So, if on a budget, it can sometimes be a better deal to look for a lot of 10 or so in a more "business to business" region of the world, like maybe Ali. At least, if you're willing to sell the other ones in the "RAM upgrade for..." department afterwards.
I mean, these parts were manufactured by the gazillion. Dont' know how "legacy" they are by now, but in my gut feeling, anything above a buck per piece is basically theft.

Found some for 3.5 sterling in a two minute search:
https://www.ebay.co.uk/itm/222449233845

Or five for ten from good old Shenzen, but those are EDO again:
https://www.ebay.co.uk/itm/173752765061

Edit: hooray, 100 posts, I've made it from "Newbie" to "member". 😁

Right now the problem is wrong memory.

There are two kinds of SOJ40 pinouts (separate CAS lines and separate Write strobles) and both are mutually incompatible, plus there's EDO vs FPM. This means 4 combinations, and only one of those works (Separate CAS + FPM), but luckily this is the most common memory and shouldn't be difficult to find. But you must verify pinout etc. from datasheet prior if buying bare chips.
I prefer to salvage the chips from existing old PCI video cards (nearly all the cheapo Cirrus, mach64, S3 Trio cards), they already have the right stuff on them.

Thermalwrong wrote on 2022-03-04, 00:01:

I just thought of something, those OKI chips might work if you tie the UCAS & LCAS pins together on the SoJ. Maybe. What I'm thi […]
Show full quote

I just thought of something, those OKI chips might work if you tie the UCAS & LCAS pins together on the SoJ. Maybe. What I'm thinking is, if the card only has one of the two CAS lines that the OKI chips need, maybe it's only getting 8 bits out of 16 bits.

The two that you've linked are FPM, so they're worth a go.

I Think they might have the same problem as the OKI chips. In that they also have the UCAS & LCAS line differences from the original Hitachi chips, which just have the one CAS line. You can test that for yourself non destructively by checking with a multimeter (or visually) whether the SoJ socket's pins 28 & 29 are tied together on the PCB. If those two pins are connected then any of the 256k x 16 SOJ 40 FPM chips should be fine.

It seems like the end numbers, 4170 vs 4260 mean something different in terms of possible arrangement which might explain the different pinout on SoJ that you'd think were standard. The 4170 chips don't seem very available though.

------
Edit, well this was a rabbit hole. Looked up datasheets that mention UWE & LWE since that would be similar to your original chips. Found the "(HM514270,HM51s4270)" seem to be both available and have the same pinout.

Hi Thermalwrong, I buzzed out pins 28&29 and they were not connected, so I linked them (easily reversible if I need to) and tested, but unfortunately still did not work with the OKI chips. I've been told that pins 12 & 13 may need linking too? On HM514170AJ they are LWE & UWE but on M514265BSL they are NC & WE.

Tiido wrote on 2022-03-04, 10:50:

There are two kinds of SOJ40 pinouts (separate CAS lines and separate Write strobles) and both are mutually incompatible

OK, wasn't aware of that. Still not rocket science, but better check for that.
You'll find a data sheet for pretty much anything.

The original Hitachi chips:
https://4donline.ihs.com/images/VipMasterIC/I … 6BF9D772C3B8264

The OKI ones I suggested:
https://ardent-tool.com/datasheets/OKI_MSM514260.pdf
Yup, they appear to be of the other kind, sorry.

You'll figure it out!

Also thank you Cutoon & Tiido, I am also checking your ideas too, much appreciated 😀