My experience in this area is that if you have an assortment of 10 different PCMCIA ram/flash/sram cards and an assortment of 10 different devices that have PCMCIA, then after a year of messing around, you will have found 3 cards that work in 3 of the devices, but only in that device. For some reason other cards like modems, NICs, interface adaptors are all far more interchangeable. The fun thing about the flash disk versions, is often that if by the chance that the Gods of Chaos weren't paying attention that you can actually get a single card recognized in two different devices... it needs to be formatted differently for each.

For the early 90s the standards were all over the place, changing, updating, not sure if there were some later ones that were backwards incompatible, also that for RAM type uses there was only meant to be an 8bit standard and the 16bit standards weren't standard. Then you have the 3.3v vs 5V thing which ports handle different.

Info was very hard to come by on exact specifics in the 1990s, I think you had to suck it up and pay the $500 or whatever your equipment vendor wanted for the manufacturer approved solution... if they actually did, most left it hanging like "Probably maybe you can potentially use this slot at sometime for something vaguely useful, anyway it's there."

Anyway, just saying there may be difficulty in getting a definitive answer of what works in this machine, or finding it if you know the specifics.

BitWrangler wrote on 2026-01-24, 16:44:

My experience in this area is that if you have an assortment of 10 different PCMCIA ram/flash/sram cards and an assortment of 10 different devices that have PCMCIA,

These modules aren't PCMCIA, it's JEIDA of unknown version. Unlike PCMCIA they have 88 pins. There is a wiring diagram for this connector in the Toshiba manual, but it doesn't make it clear whether it requires 3V and/or FPM: it has both Vcc and B3V.

The card stays cool in while connected, so it's definitely not overvolted. I'd rather blame EDO vs. FPM or timings. The seller has lots of the cards, so he might provide a replacement if it's really a dead chip. I wonder if I can identify this without sending the cards back and forth.

Oh man, like weirder is better 🤣

Anyhoo, according to the system board pics off here...
https://zelandeth.org/oldtech/toshiba-t1950ct/index.htm

What is on the motherboard is the HM514400BL S7 which should be 1mx4bit 70ns FPM according to https://www.alldatasheet.com/html-pdf/84402/H … HM514400BL.html

So, seems you need FPM not EDO. Though that doesn't tell us what that module you have actually is unless you want to peel it open and eyeball chips.

BitWrangler wrote on 2026-01-24, 20:14:

So, seems you need FPM not EDO. Though that doesn't tell us what that module you have actually is unless you want to peel it open and eyeball chips.

Yeah, as long as there is a chance to get a replacement, I'll wait with peeling. 😀 But if it's EDO, I'll peel it to convert to FPM anyway.

Meanwhile I've run like a hundred CheckIt RAM tests, and what I observe is that there are 3 64K regions which are failing, and it's always either bit 1 or 3. Unless CheckIt reports it really weird way, this clearly means a failing chip, not an EDO/FPS issue. In the latter case it would have been random failures.

BitWrangler wrote on 2026-01-24, 20:14:

So, seems you need FPM not EDO. Though that doesn't tell us what that module you have actually is unless you want to peel it open and eyeball chips.

Meanwhile I've got a replacement from the seller. The replacement fails too, but with a different bit. The first card failed with bits 1 and 3, while the second one with bit 7.

The seller didn't want the first card back, so I peeled it open. Inside there are level shifters and usual 60ns FPM chips. Should work actually except if the require some specific RAS/CAS timings.

I have one more mystery with the second card. It always fails when cold started. Sometimes even POST detects the problem after 2nd MiB. But if I let it run some hours, the errors disappear. Sounds like problems with PSU, but then I'd expect problems with the on-board RAM too, which is not the case.