You can probably work most of those out. E.g., that first Goldstar one, looking up the part number gets to this datasheet:
https://pdf1.alldatasheet.com/datasheet-pdf/v … GM71C4400D.html
Which says it's a 1Mbit x 4bit chip. 2 of those gets us 1Mbit x 8bit, or 1MB. The other chip'll be a 1bit parity (can't find a datasheet, but there's this random page https://fthain.github.io/dram/ that lists it as 1Mx1). So that's a 1MB module.

Then it's just (ha) a case of going through all the others.

BLockOUT wrote on 2021-05-16, 22:07:

Hey guys i hope you can help me with this. I got a little box full of 30pin simm modules at that time there were no stickers and […]
Show full quote

Hey guys i hope you can help me with this. I got a little box full of 30pin simm modules
at that time there were no stickers and nothing written on the pcb that could tell very simple the capacity of the module.
My problem comes when i know there must be some 4mb modules and the rest is 1mb, but its hard to tell which are 4mb.
Not only that, the modules most of them have 3 chips, and two of those chips are the same and the 3rd chip is different.

here is a list of what i have:

simm1: (3chips) : (2) goldstar gm71c4400aj70 , (1) goldstar gm71c1000j70
simm2: (3chips) : (2) toshiba tc514400asj-70 , (1) toshiba tc511000ajl-70
simm3: (3chips) : (2) HY514400A (1) HY531000AJ-70
simm4: (3chips) : (2) MM51440-07 (1) 81c1000a-70
simm5: (9chips) : (9) 81c1000a-70
simm6: (3chips) : (3) eagle 9226 EM4103NJ-7
simm7: (3chips) : (2) HY514400j-70 (1) HY531000J-70
simm8: (3chips) : (2) LG6A4400ak-70L (1) gsm mpe 41000j7
simm9: (3chips) : (2) TL210407 (1) panasonic MN41c1000asj-07
simm10: (9chips) : (9) 41000BL 0062e50 21 27103
simm11: (9chips) : (9) km41c1000aj-7
simm12: (9chips) : (9) km41c1000bj-8

And the simm module that I think its 4mb but im not 100% sure is this one:

simm13: (3chips) : (2) toshiba TC5117400CSJ-60 (1) BP41C4000B-6

can anyone help me to figure out which modules are 4mb 30pin simms?

The trick is watching your captials.

SIMM sizes are given in MegaBytes (MB), chip sizes are given in Megabits (Mb). You're mixing them up using "mb" (millitbit?). Look up the chip documentation, that gives you chip size in Mb, add up the size of all equal chips (the odd one is invariably parity) and then divide by 8. If there are 9 identical chips, 8 are data, ninth is parity.

So the first SIMM has two 4Mb chips, make 8Mb total. Divide by 8 and you get 1MB. The third chip is a 1Mb chip, which is spot on for parity for a 1MB module (parity is 1/8 of the module size). Repeat for the rest of the SIMMs and you know what you have. For what it's worth, I think you have two 4MB SIMMs there, but I can't be sure about eagle 9226 EM4103NJ-7 because that's a(n obscure) SIMM part number, not a chip code for the chips on it.

the vast majority of memory chips can be easily identified from the numbers. After decades of experience, its second nature. This got a lot more complex with Dimms, and isnt as reliable.

in general, if you see 4000 or 4400, its going to be 4mb. 1000 will be 1mb. and so forth.

#1 thru #4 - 4MB - this one is a bit tricky, as its a 4x4 chip with a 1x1 chip.
#5 - 1MB - 9x1
#6 - 4mb - and with the 1996 date code, it seems like they stop caring about the parity and just shipped it with a larger parity chip
#7 - #9 - 4mb same as #1
#10 thru #12 - 1mb

Agree LuckyBob ! some may not work in some older boards due to it's chip type/density. All the 9 chip should be okay... just rambling ;p

Problem is you have mixed sets. Slow 386 would work with these mixed sets using 3 chips and perfect for 286 and 386sx (these are 2 simms per bank). Fast cached 386 and 486 will not be happy as they needs nearly matched sets (two and two), or ideally set of 4 matched set to make a bank.

Cheers,

pentiumspeed wrote on 2021-05-17, 00:04:

Problem is you have mixed sets. Slow 386 would work with these mixed sets using 3 chips and perfect for 286 and 386sx (these are 2 simms per bank). Fast cached 386 and 486 will not be happy as they needs nearly matched sets (two and two), or ideally set of 4 matched set to make a bank.

Cheers,

This really isnt much of a thing. At least in the 30-pin simm world. My OCD will agree with you 110%. Everything needs to match perfectly, down to the date codes! Thankfully I take a pill for it.

Early 2/386 may not like the 4mb 3-chip simms. Without going into the weeds, they use the same chips you might find on 16mb 30p simms. Not all motherboards like this. As long as you are using simms with the same physical setup, it wont matter. keep 9-chip simms in bed with the other 9's. etc. Usually the -60, -70, -80 denote speed. You can mix them, HOWEVER you cant tell the motherboard to go faster than the slowest chip. (lower numbers are FASTER). So if you have 3x 60's and 1x 70, you should set the jumpers/bios to 70ns.

yes, you SHOULD match brand and speed. But its not nearly as important as it is on later machines. You should ALWAYS match physical configuration. This includes different banks as well.

luckybob wrote on 2021-05-17, 00:18:

pentiumspeed wrote on 2021-05-17, 00:04:

Problem is you have mixed sets. Slow 386 would work with these mixed sets using 3 chips and perfect for 286 and 386sx (these are 2 simms per bank). Fast cached 386 and 486 will not be happy as they needs nearly matched sets (two and two), or ideally set of 4 matched set to make a bank.

Cheers,

This really isnt much of a thing. At least in the 30-pin simm world. My OCD will agree with you 110%. Everything needs to match perfectly, down to the date codes! Thankfully I take a pill for it.

You to ah. Mine are pink and red. 486/386DXs NEED 4 x30pin simms per bank because they are true 32-bit cpus.

That is all folk. PEEP PEEP!!...............

🤣 I am still a bit OCD with SIMMS and keep the gold plated 30pin sets separate from the tin plated even if have exact same chips on them 😀

luckybob wrote on 2021-05-16, 22:48:

in general, if you see 4000 or 4400, its going to be 4mb. 1000 will be 1mb. and so forth.

#1 thru #4 - 4MB - this one is a bit tricky, as its a 4x4 chip with a 1x1 chip.

Ah, I must have misunderstood something here. I assumed the 4400 meant it was 4 megabit in total, arranged in 4 bit groupings. Two of those chips would give 8 megabits, so would be a 1 megabyte module (plus a 1 megabit parity). A GM71C4400 datasheet describes it as "dynamic RAM organized 1,048,576 x 4 bit". So two of those in parallel with each other would give an 8 bit data bus with 2^20 addresses, which would be 1 megabyte.

Have I gone wrong somewhere?

Dionb, I found a picture of some Eagle RAM (simm 6), so it looks like that is the right part number for the chip. I also found an Amibay listing for a 72 pin, 16 chip module using the same chips, listed as 8MB. So I figure the 2 chips on this module makes it also 1MB.

luckybob wrote on 2021-05-16, 22:48:

the vast majority of memory chips can be easily identified from the numbers. After decades of experience, its second nature. T […]
Show full quote

the vast majority of memory chips can be easily identified from the numbers. After decades of experience, its second nature. This got a lot more complex with Dimms, and isnt as reliable.

in general, if you see 4000 or 4400, its going to be 4mb. 1000 will be 1mb. and so forth.

#1 thru #4 - 4MB - this one is a bit tricky, as its a 4x4 chip with a 1x1 chip.

[...]

#7 - #9 - 4mb same as #1

Eh? He writes 2x 4400 and 1x 1000, so that sounds like 2x 4Mbit = 8Mbit, with 1Mbit parity. Looks like 1MB to me.

snufkin wrote on 2021-05-17, 10:16:

luckybob wrote on 2021-05-16, 22:48:

in general, if you see 4000 or 4400, its going to be 4mb. 1000 will be 1mb. and so forth.

#1 thru #4 - 4MB - this one is a bit tricky, as its a 4x4 chip with a 1x1 chip.

Ah, I must have misunderstood something here. I assumed the 4400 meant it was 4 megabit in total, arranged in 4 bit groupings. Two of those chips would give 8 megabits, so would be a 1 megabyte module (plus a 1 megabit parity).

No, you didn't misunderstand anything, luckybob is wrong here. 4400 means exactly what you say 4 MBit total in a x4 organization, so 1M x 4.

snufkin wrote on 2021-05-17, 10:16:

Ah, I must have misunderstood something here. I assumed the 4400 meant it was 4 megabit in total, arranged in 4 bit groupings. […]
Show full quote

luckybob wrote on 2021-05-16, 22:48:

in general, if you see 4000 or 4400, its going to be 4mb. 1000 will be 1mb. and so forth.

#1 thru #4 - 4MB - this one is a bit tricky, as its a 4x4 chip with a 1x1 chip.

Ah, I must have misunderstood something here. I assumed the 4400 meant it was 4 megabit in total, arranged in 4 bit groupings. Two of those chips would give 8 megabits, so would be a 1 megabyte module (plus a 1 megabit parity). A GM71C4400 datasheet describes it as "dynamic RAM organized 1,048,576 x 4 bit". So two of those in parallel with each other would give an 8 bit data bus with 2^20 addresses, which would be 1 megabyte.

Have I gone wrong somewhere?

Dionb, I found a picture of some Eagle RAM (simm 6), so it looks like that is the right part number for the chip. I also found an Amibay listing for a 72 pin, 16 chip module using the same chips, listed as 8MB. So I figure the 2 chips on this module makes it also 1MB.

Is it just me or are the logos printed at different angles on different chips here?

Anyway, looks like 2x 4Mb + 1x 1Mb unfortunately, and the 72p SIMMs with same chips explain how some searches turned up "4MB", as if you stick 8 of these on 72p SIMM you get a nice 4MB module.

That leaves the SIMM with 2x Toshiba TC5117400CSJ-60 pretty lonely as only clear 4MB SIMM. 16Mb (4Mx4) per chip, 32Mb = 4MB per SIMM. And a 4Mb parity chip for good measure.

Well, yea obviously.

the fastest way to get the correct answer is to give someone the wrong one, so people will then come out of the woodwork to correct you. Its called Godwin's law.

dionb wrote on 2021-05-16, 22:39:

BLockOUT wrote on 2021-05-16, 22:07:

simm6: (3chips) : (3) eagle 9226 EM4103NJ-7

For what it's worth, I think you have two 4MB SIMMs there, but I can't be sure about eagle 9226 EM4103NJ-7 because that's a(n obscure) SIMM part number, not a chip code for the chips on it.

Be very wary about these SIMMs. I have a set of 4 MB SIMMs like that, consisting of 3 chips of 4M x 4 each, so there are actually 12 bits to choose from. There are 0 Ohm SMD "jumpers" to choose 3 of the 4 bits of each chip, so the module uses 9 of the 12 bits present in the chips.

The chips themselves have a type number that can nowhere been found on the internet. The modules work unreliably, depending on access patterns. Most annoyingly they work fine in memtest86+, likely because they get constantly accessed throughout the whole chip, but they fail with parity errors when I boot Linux. Certain bits seem to be weak and can flip in a rowhammer-like fashion if neighbouring rows are accessed.

Most likely, the chips are relabelled factory rejects, and the module makers tried to be tricky by excluding the weakest bit of the broken chips.

In addition to the modules having bad bits, they are not compliant to the 30-pin SIMM standard. The parity bit needs to be accessible using a separate CAS signal (so it must not be joined into a chip that also carries data bits), and the parity bit has separate read and write lines. Most x1 DRAMs provide them (as D and Q), whereas x4 DRAMs don't, but combine data in and data out into four bidirectional DQ pins). So you can not store parity bits in x4 chips. In practice, many 386/486 class boards short CAS (for data) and CASP (for parity) as well as PD (parity to module) and PQ (parity from module). On those boards, this kind of non-compliant module would work - if the memory chips were good.

luckybob wrote on 2021-05-17, 17:19:

Its called Godwin's law.

No, actually, that's not..... Wait a minute.
Oooh. Sneaky.

mkarcher wrote on 2021-05-17, 17:27:

Be very wary about these SIMMs. I have a set of 4 MB SIMMs like that, consisting of 3 chips of 4M x 4 each, so there are actually 12 bits to choose from. There are 0 Ohm SMD "jumpers" to choose 3 of the 4 bits of each chip, so the module uses 9 of the 12 bits present in the chips.

And then that's actually sneaky. Couldn't understand why they would have used the same size memory for data and parity, didn't seem possible for the benefits of lower unique parts to offset the increase cost per part. Didn't know about or think about them possibly binning parts like that.

No, actually, that's not..... Wait a minute.
Oooh. Sneaky.

*guy tapping on his forehead meme*

exactly. I'd wager the parity chip might be fine. But from the date code, it seems 30-pin stuff was on its way out, and it was cheaper to just use same chip three times, than source the 1x1 chips and use another spool on the pick and place.

but yea, I'm probably wrong about that size, as others have pointed out. I really want one of those simm testers that Adrian featured on his basement channel.

mkarcher wrote on 2021-05-17, 17:27:

Be very wary about these SIMMs. I have a set of 4 MB SIMMs like that, consisting of 3 chips of 4M x 4 each, so there are actually 12 bits to choose from. There are 0 Ohm SMD "jumpers" to choose 3 of the 4 bits of each chip, so the module uses 9 of the 12 bits present in the chips.

Perhaps it's modules like these that make people say "this mobo doesn't support 3-chip sticks, use 9-chip ones"? Because frankly I see no reason why it wouldn't work. A proper 2*4+1 stick should work just as well as 9*1.

Unless the mobo is so poorly designed that the parity chip must be of the exact same type and rating as the data ones, or else the faster/slower output of the parity bit will trigger an error during read. I've yet to see something like that though.

Deunan, for all I comprehend about the issues with the memory modules...

It tends to be a refresh issue.

luckybob wrote on 2021-05-17, 17:19:

the fastest way to get the correct answer is to give someone the wrong one, so people will then come out of the woodwork to correct you. Its called Godwin's law.

https://i.imgur.com/6TTc8hc.gif