I also have this matched set of four 30-pin SIMMs. I don't have anything that can use this memory, so I don't know how to read their capacity from the writing on the chips. I would like to label them, if someone can help identify them for me.

One more, another EDO stick. I has 8 chips on both sides.

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width.
#2 1 MegaByte each with parity.
#3 8 MegaByte, again doubting it is EDO.

Seems more like plain Fast Page mode dram.

I see. FPM and EDO both came in 72-pin varieties and can fit into the same slot. I guess these were pulled from a 386 or 486, if they're individual sticks. I've had these in a bin for a long time, and can't remember their origin.

Thank you for helping to identify these.

The requirement of installing 72-pin SIMMs in pairs is due to the Pentium's 64 bit data bus (each SIMM is 32 bits), so it's not related to it being EDO. In fact, the same restriction applies if you use FPM. On the other hand for a 486 1 SIMM is enough, but I understand the confusion since EDO RAM is usually associated with Pentium boards.

Not sure about the 4 chips on the back in your first post, I suppose it could be for parity. Usually for 72-pin SIMMs 1, 4, 16 MB modules are one-sided while 2, 8, 32 MB are double-sided.

Regarding the 30-pin SIMMs, the best idea is to look up the datasheet. In your case, each chip is 1M x 1 bit, so in total the module has 1MB (8 chips = 1M x 8 bit) + 1 chip for parity.

weedeewee wrote on 2022-10-29, 05:06:

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width. #2 1 […]
Show full quote

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width.
#2 1 MegaByte each with parity.
#3 8 MegaByte, again doubting it is EDO.

Seems more like plain Fast Page mode dram.

Is there a way to visually tell FPM from EDO ?

My EDO sticks are labled but I do have a bunch of unlabled 72pin simms, if there is a visual indicator that you can use to tell them apart that would be awesome.

TrashPanda wrote on 2022-10-29, 05:36:

weedeewee wrote on 2022-10-29, 05:06:

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width. #2 1 […]
Show full quote

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width.
#2 1 MegaByte each with parity.
#3 8 MegaByte, again doubting it is EDO.

Seems more like plain Fast Page mode dram.

Is there a way to visually tell FPM from EDO ?

My EDO sticks are labled but I do have a bunch of unlabled 72pin simms, if there is a visual indicator that you can use to tell them apart that would be awesome.

Some discission here Q: EDO vs. FPM - how to tell which is which

TrashPanda wrote on 2022-10-29, 05:36:

weedeewee wrote on 2022-10-29, 05:06:

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width. #2 1 […]
Show full quote

#1 are you sure it's edo ? and yes, parity, also 486 only needs one module, pentium mostly needs 2 due to memory bus width.
#2 1 MegaByte each with parity.
#3 8 MegaByte, again doubting it is EDO.

Seems more like plain Fast Page mode dram.

Is there a way to visually tell FPM from EDO ?

My EDO sticks are labled but I do have a bunch of unlabled 72pin simms, if there is a visual indicator that you can use to tell them apart that would be awesome.

it's a crap shoot if the datasheet isn't available or found, but going by the chip id and looking at other manufacturers chips with a mostly similar id gives good results, though not always correct?
that being said, having looked a bit further, information gleaned from the interwebz seems to indicate both are likely edo, though I still have my doubts until they get tested, with the first one less doubt and the third one a bit more doubt. 😉

Be careful with the first module. I can't find datasheets at the usual place for those chips. I don't see any brand manufacturer name. I have some modules with these properties, and they often turn out to be of low quality, i.e. the memory is unreliable. Before trusting them in a system, I highly recommend you to give that system a thorough run of memtest86. No-name "Made in Taiwan" memory modules from the 90s sometimes seem to be built from factory rejected chips that have been rebadged.
You are right that it seems like the 8 identical chips on the "front" side are data chips storing 4 bits each to provide the required 32 data bits. The four chips on the back are used for 1 bit each to add the four required parity bit for a fully conformant 36-bit PS/2 modules. The number "400" as part of the chip code indicates that the chips have "4 mega" of something. Either they have 4 mega addresses with 4 bits at each address, which would be called 4M x 4 (16 megabits total) or they have 4 mega bits with 4 bits at each address, so just 1 mega addresses, which would be called 1M x 4 (4 megabits total). The package of the chips is SO24/26 (it has 24 pins, in a package big enough to fit 26 pins). Standard 1M x 4 chips use SO20/16, whereas standard 4M x 4 chips use SO24/26, so it is likely that the chips were produced as 4M x 4. This doesn't necessarily mean the module makes use of all 4M addresses of these chips, but maybe the module just uses 1M of them, because 1M of the 4M addresses happen to be "good enough" in a otherwise defective chip. It's unlikely that the chips are used only partially though, because modules pulling that trick usually have solder jumpers per chip to select which part is going to be used.
If we go with the (most likely) theory that the chips on the front are 4M x 4 chips, the chips on the back should be 4M x 1 chips (1 bit per each of the 4 mega addresses). 4M x 1 chips also typically are delivered in SO20/26, not SO24/26, so this looks suspect. I'd guess the chips on the back are produced as 4M x 4, and only one of the bits (the most reliable one) is used. As there are no solder jumpers to select which bit is going to be used, they need to have built these modules in large quantities, and produced four kinds of PCBs.
So I'd guess the module will be 16MB (4 mega addresses with 4 bytes per address) with parity.
These modules also show why some of IT professionals reject the terms "single-sided" and "double-sided". This modules obviously has chips on both sides, i.e. it is "double-sided", but it only provides one bank of memory, which makes those module "single-sided" in terms of colloquial memory naming. "double-sided" modules (or "dual-rank modules" as the preferred technical term) have 32 data bits on each side. In modules like your #1, if there was supposed to be a dual-rank variant, they would have made a bigger PCB, and placed the parity chips on the front side, typically rotated 90 degrees above the data chips, and provided space to solder the same amount of 12 chips on the back side. This would allow them to equip only the first side to make a 16MB single-rank module or a 32MB dual-rank module. As you see, in the case of partially populated dual-rank PCBs, the terms single-sided and dual-sided perfectly align with the technical property of being single-rank or dual-rank.

I think the 3rd set are FPM, there's a datasheet for the NN514400 chips here https://datasheetspdf.com/datasheet/NN514400A.html

First one is annoying. I think the logo might be 'QC', that at least turns up some results with sticks using chips with that logo. Still no datasheets. I can see the logo might instead be something like ERC, but I still can't find anything.

For 1 and 3 it looks like the presence detect resistors are all empty. At least one standard ( https://www.ele.uri.edu/iced/protosys/hardwar … arpoint-8MB.pdf ) says that would make them 8MB/60ns. Although it looks like IBM had other meanings for the PD pins.

snufkin wrote on 2022-10-29, 11:53:

For 1 and 3 it looks like the presence detect resistors are all empty.

Most consumer mainboards do not care about presence detect. It's big brand hardware like HP or IBM that cares about PD, even more so in printers like HP LaserJet printers. Because consumer HW doesn't care about PD, also a lot of consumer memory modules don't have PD resistors installed. So empty PD pads are mostly meaningless.

Ok, that makes sense about the PD pins.

This is going to sound a bit weird, but I think the second digit of the part number is the bit width. So 43C400 are 3 bit chips, and the 42C400 are 2 bit. That gives the 32 bit data lines, and the 4 parity pins (2 either side of the notch) aren't connected. Don't know how they'd select which data pins to use. It would be interesting to see the PCB without the chips in place, and to check if maybe the parity pins are routed on an internal layer and I'm wrong.

Found a photo of a stick claiming to be 32MB with 16 of 44C400 QC chips. Which I think means they must be arranged as 4 bit * 4M addresses in 2 ranks of 8 chips. I think that suggests that the chips on this odd EDO stick are also 4M addresses, but arranged as a single rank of 8 chips @ 3 bit and 4 chips @ 2 bit. So 16MB, same as your guess, but without parity.