Hi, I wanted to congratulate you on the excellent work, I am currently working on the restoration (therefore also repair) of some old PCs, most of these use 30 PIN RAM, usually with 1MB SIMMs you can install a maximum of 8MB (total) I've never had the ability to use larger SIMMs. I had recently seen 4MB 9 chip SIMMs on sale, but sadly they were sold out before I could place a bid.
I don't know if you are interested, but in the early 90's, to reuse 30PIN RAM, there was a 30PIN-4x30PIN adapter, it existed in two versions with right and left banks (or top and bottom?), unfortunately it took at least four to install 16MB of total RAM (1MBX4X4).
I put a photo to show you what it is.
Going back to the modules you made, could they work well on PCs with mixed banks?, That is 30PIN and 72PIN, I have some 486s that have both but on some they work without any problem, while others do not.

Such a thing presents a physical difficulty, you're not going to place 4 of them side by side on a typical motherboard to fill out one bank... Rest wouldn't be too much of a problem though.

----

That Acer stick pinout is definitely not the same, since it won't support parity.

Since there's 8MB quoted there must be an 11th address bit too. 10 lines indeed will only give 2MB with 16bit sticks. 4MB requires asymmetric chips with 11 address lines.

There are some pins that aren't connected that can hold the last address bit.

For the T5200 in particular I think the best course is to make a complete memory board, with as many chips needed to max stuff out, no sockets or anything else extra. I imagine the main connector is pretty straightforward though it would require a whole machine to deal with. I'd like to do such a thing to my T3200 one day.

PC@LIVE wrote on 2021-06-30, 19:26:

Going back to the modules you made, could they work well on PCs with mixed banks?, That is 30PIN and 72PIN, I have some 486s that have both but on some they work without any problem, while others do not.

On many of those boards with mixed banks, you have to adjust some jumpers (or resistor network) to indicate whehter the 30-pin or 72-pin socket(s) are bank 0. If a single bank (4 modules) of non-defective 1MB or 4MB 30-pin modules doesn't work in a normal 486-class mainboard, you are either using non-parity modules in a mainboard requiring parity, or you need to adjust some jumper.

mkarcher wrote on 2021-06-30, 20:39:

PC@LIVE wrote on 2021-06-30, 19:26:

Going back to the modules you made, could they work well on PCs with mixed banks?, That is 30PIN and 72PIN, I have some 486s that have both but on some they work without any problem, while others do not.

On many of those boards with mixed banks, you have to adjust some jumpers (or resistor network) to indicate whehter the 30-pin or 72-pin socket(s) are bank 0. If a single bank (4 modules) of non-defective 1MB or 4MB 30-pin modules doesn't work in a normal 486-class mainboard, you are either using non-parity modules in a mainboard requiring parity, or you need to adjust some jumper.

Thank you very much, I did not know these things, I will try to verify the existence of any jumpers, from memory a card is a BEK VLB V429S, in this I can not put more than 16MB, currently I have 2 SIMMs at 72PIN, and I can not put any more, for example it should be possible to reach 20MB with four SIMMs of 30 PINs (1mB each).

PC@LIVE wrote on 2021-06-30, 20:54:

Thank you very much, I did not know these things, I will try to verify the existence of any jumpers, from memory a card is a BEK VLB V429S, in this I can not put more than 16MB, currently I have 2 SIMMs at 72PIN, and I can not put any more, for example it should be possible to reach 20MB with four SIMMs of 30 PINs (1mB each).

That might be impossible. The 8MB SIMMs you use are dual-rank SIMMs (often produced and called "double-sided"). They occupy two /RAS lines of the chipset. Two of those SIMMs need four /RAS lines (think of them as "bank select" lines). It is quite common for 486 chipsets to only have four /RAS lines, and offering jumpers on how to distribute the /RAS lines to the 72-pin and 30-pin sockets.

The root cause of your problem is that you only have modules to equip 4MB per /RAS line (either a set of 4 1MB 30-pin SIMMs, or one side of a 8MB PS/2 SIMM). To get more than 16MB, you likely need modules with 16MB per RAS line (like 16MB PS/2 SIMMs or a set of 4 * 4MB as 30-pin modules).

In case of that Toshiba, 8MB is result of 16bit bus, equivalent of half of a 16MB 72pin SIMM. from what it looks like.

Tiido wrote on 2021-06-30, 21:39:

In case of that Toshiba, 8MB is result of 16bit bus, equivalent of half of a 16MB 72pin SIMM. from what it looks like.

from the info, provided by 16ShadesOfOrange & IanB, it seems that for the toshiba t5200 the 8MB is actually 2 40pin 4MB (16bit+2paritybit) modules.

Which is both halves of an 72pin 8MB parity module.

mkarcher wrote on 2021-06-30, 21:29:

PC@LIVE wrote on 2021-06-30, 20:54:

Thank you very much, I did not know these things, I will try to verify the existence of any jumpers, from memory a card is a BEK VLB V429S, in this I can not put more than 16MB, currently I have 2 SIMMs at 72PIN, and I can not put any more, for example it should be possible to reach 20MB with four SIMMs of 30 PINs (1mB each).

That might be impossible. The 8MB SIMMs you use are dual-rank SIMMs (often produced and called "double-sided"). They occupy two /RAS lines of the chipset. Two of those SIMMs need four /RAS lines (think of them as "bank select" lines). It is quite common for 486 chipsets to only have four /RAS lines, and offering jumpers on how to distribute the /RAS lines to the 72-pin and 30-pin sockets.

The root cause of your problem is that you only have modules to equip 4MB per /RAS line (either a set of 4 1MB 30-pin SIMMs, or one side of a 8MB PS/2 SIMM). To get more than 16MB, you likely need modules with 16MB per RAS line (like 16MB PS/2 SIMMs or a set of 4 * 4MB as 30-pin modules).

I haven't had time to look at what RAM I have on that BEK-V429S PC yet, I remember trying various modules but the maximum I got was 16MB.
Theoretically, however, there should be no problem putting more RAM with different capacities, but for some reason this is actually not possible.
It is possible that it depends as you suggest on the type of modules, even if visually they are always 72PIN or 30PIN, I believe there are such differences as to make them readable in one PC and illegible in another PC.
In fact I don't have 4MB 30PIN RAM, so I don't know if installing them would get a total of 32MB of RAM, or if it is possible to put more.
I have a 16MB 72PIN module aside, the other identical is broken, but that module is strangely read only in the PC where it was installed (iVX.P133), on the 486s they don't detect it.

PC@LIVE wrote on 2021-07-01, 20:36:

I have a 16MB 72PIN module aside, the other identical is broken, but that module is strangely read only in the PC where it was installed (iVX.P133), on the 486s they don't detect it.

might be that the difference is just one being FPM and the other being EDO.

weedeewee wrote on 2021-07-01, 22:26:

PC@LIVE wrote on 2021-07-01, 20:36:

I have a 16MB 72PIN module aside, the other identical is broken, but that module is strangely read only in the PC where it was installed (iVX.P133), on the 486s they don't detect it.

might be that the difference is just one being FPM and the other being EDO.

That is very likely indeed. Most Pentium boards support EDO RAM, and they are slightly faster with EDO than FPM. The additional cost of EDO RAM compared to FPM went down during the HX days, and was negligible during the VX days, so EDO RAM was a nice selling point at nearly no additional cost for the manufacturer.

On the other hand, most 486 mainboards fail to work with EDO RAM, even if the memory controller is supposed to support it. For example I have yet to see an SiS 496/497 board that properly works with EDO RAM.

PC@LIVE wrote on 2021-07-01, 20:36:

I haven't had time to look at what RAM I have on that BEK-V429S PC yet, I remember trying various modules but the maximum I got was 16MB.
Theoretically, however, there should be no problem putting more RAM with different capacities, but for some reason this is actually not possible.

I checked the mainboard jumper description for your board at TH99 / UH19. You have a single jumper that needs to indicate whether the 30-pin SIMM slots are occupied or not, that's JP5.

The list of supported configurations seems to be incomplete, though, as it has no dual-rank PS/2 modules listed, so your configuration of 2 8MB modules (technically: dual-rank, 1MB x 36 per rank) is not in that table. On the other hand, I checked the chipset datasheet. Your mainboard uses the Opti 895 chipset, which (as I suspected) only supports four ranks. With two dual-rank modules, the chipset is maxed out. To get above 16MB, you need 16MB or 32MB FPM PS/2 SIMMs, or you need a set of of 4 4MB 30-pin SIMMs. 30-pin SIMMs do not exist as EDO SIMMs, so no need to worry about the type.

Obviously, as your board allows you to plug in memory with 7 ranks (1 rank on the 30-pin array, 2 ranks on each of the PS/2 slots), some of the ranks have to be mutually exclusive. I guess your memory is plugged into banks 1 and 2, so you obviously find all four ranks from the chipset on those two banks. I furthermore guess that the secondary rank of bank 2 is also available as primary rank on bank 3, so you could replace the double-sided 8MB module in bank 2 with two single sided 4MB modules, one in bank 2 and one in bank 3, without the chipset noticing the difference. I furthermore guess that the jumper JP5 toggles either that rank, or the secondary rank of bank 1 to appear on the 30-pin sockets instead of the 72-pin socket. The second rank of bank 3 is most likely not connected at all, so an 8MB module in that bank only has 4MB accessible.

mkarcher wrote on 2021-07-02, 08:15:

I checked the mainboard jumper description for your board at TH99 / UH19. You have a single jumper that needs to indicate whethe […]
Show full quote

PC@LIVE wrote on 2021-07-01, 20:36:

I haven't had time to look at what RAM I have on that BEK-V429S PC yet, I remember trying various modules but the maximum I got was 16MB.
Theoretically, however, there should be no problem putting more RAM with different capacities, but for some reason this is actually not possible.

I checked the mainboard jumper description for your board at TH99 / UH19. You have a single jumper that needs to indicate whether the 30-pin SIMM slots are occupied or not, that's JP5.

The list of supported configurations seems to be incomplete, though, as it has no dual-rank PS/2 modules listed, so your configuration of 2 8MB modules (technically: dual-rank, 1MB x 36 per rank) is not in that table. On the other hand, I checked the chipset datasheet. Your mainboard uses the Opti 895 chipset, which (as I suspected) only supports four ranks. With two dual-rank modules, the chipset is maxed out. To get above 16MB, you need 16MB or 32MB FPM PS/2 SIMMs, or you need a set of of 4 4MB 30-pin SIMMs. 30-pin SIMMs do not exist as EDO SIMMs, so no need to worry about the type.

Obviously, as your board allows you to plug in memory with 7 ranks (1 rank on the 30-pin array, 2 ranks on each of the PS/2 slots), some of the ranks have to be mutually exclusive. I guess your memory is plugged into banks 1 and 2, so you obviously find all four ranks from the chipset on those two banks. I furthermore guess that the secondary rank of bank 2 is also available as primary rank on bank 3, so you could replace the double-sided 8MB module in bank 2 with two single sided 4MB modules, one in bank 2 and one in bank 3, without the chipset noticing the difference. I furthermore guess that the jumper JP5 toggles either that rank, or the secondary rank of bank 1 to appear on the 30-pin sockets instead of the 72-pin socket. The second rank of bank 3 is most likely not connected at all, so an 8MB module in that bank only has 4MB accessible.

Thank you very much for the suggestions, I think they will be very useful, because after repairing a 386DX-40 I have to try to fix some 486, apart from one others are probably repairable.
Regarding the P133 and the memories that do not work on the 486, it is undoubtedly a problem of the type of memory, at sight they all look the same but this is not the case.
Returning to the PC with BEK V429S, the chipset is Opti, the memories I photographed and they are double-sided, I installed them on the central banks, so 2 and 3 while the one and the four (one is the 30PIN one) are empty .
From the photos I can read in a memory a written on a sticker, RAM2MX32S70501, this has 16 chips (8 per side) signed TI-70 tms44400dj, the other has 4 large chips (2 per side), but I cannot read the abbreviations on the chips.

PC@LIVE wrote on 2021-07-02, 20:54:

Returning to the PC with BEK V429S, the chipset is Opti, the memories I photographed and they are double-sided, I installed them on the central banks, so 2 and 3 while the one and the four (one is the 30PIN one) are empty .

OK, so this is just as I expected. It's good that it works this way, even if your configuration is not listed in TH99 / UH19.

PC@LIVE wrote on 2021-07-02, 20:54:

From the photos I can read in a memory a written on a sticker, RAM2MX32S70501, this has 16 chips (8 per side) signed TI-70 tms44400dj, the other has 4 large chips (2 per side), but I cannot read the abbreviations on the chips.

The TMS44400dj is a 4MBit chip in a x4 organization. This means it has 1M of addresses and stores 4 bit at each of those addresses. Take two of those chips, and you get 1 MByte. You can build 1MB 30-pin SIMMs without parity exactly that way. Each side of the 16-chip SIMM is basically identical to four of those 1MB 30-pin SIMMs, just in an easier to handle form factor. This is a pattern that always holds: A single "side" of a PS/2 SIMM behaves very similar to a bank of four 30-pin SIMMs.

The other SIMM obviously has 16MBit chips (the next generation - at that time, each generation of memory chips had 4 times the size of the previous generation) in a x16 configuration, so two of them are enough to make the 32 bits your need for each side. 16 MBit in a x16 configuration means the 16 MBit are distributed over 1M addresses, and each of these addresses stores 16 bit. So a single of these chips behaves like a 16-bit bank made out of two 1MB 30-pin SIMMs.

Notice how I always write "behaves like 1MB 30-pin SIMMs". Your chipset is maxed out at 16MB, as long as you only use modules that "behave like (a quadruple of / an octuple of) 1MB 30-pin SIMMs". To get to the full 64MByte the chipset supports, you need modules that in total behave like 16 30-pin 4MB SIMMs. Your options for modules that behave like 30-pin 4MB SIMMs are:

4 4MB 30-pin SIMMs
a 16MB PS/2 SIMM that behaves the same as 4 4MB 30-pin SIMMs
a 32MB PS/2 SIMM that behaves the same as 8 4MB 30-pin SIMMs

4 30-pin SIMMs + 3 16MB PS/2 modules is a configuration explicitly listed in your mainboard documentation. 2 32MB PS/2 modules in the same slots you currently have the 8MB modules in is also very likely to work. And of course, you don't need to go all the way to 64 MB, but you can stop at 32MB (e.g. from one 32MB or 2 16MB PS/2 modules).

mkarcher wrote on 2021-07-02, 22:14:

OK, so this is just as I expected. It's good that it works this way, even if your configuration is not listed in TH99 / UH19. […]
Show full quote

PC@LIVE wrote on 2021-07-02, 20:54:

Returning to the PC with BEK V429S, the chipset is Opti, the memories I photographed and they are double-sided, I installed them on the central banks, so 2 and 3 while the one and the four (one is the 30PIN one) are empty .

OK, so this is just as I expected. It's good that it works this way, even if your configuration is not listed in TH99 / UH19.

PC@LIVE wrote on 2021-07-02, 20:54:

From the photos I can read in a memory a written on a sticker, RAM2MX32S70501, this has 16 chips (8 per side) signed TI-70 tms44400dj, the other has 4 large chips (2 per side), but I cannot read the abbreviations on the chips.

The TMS44400dj is a 4MBit chip in a x4 organization. This means it has 1M of addresses and stores 4 bit at each of those addresses. Take two of those chips, and you get 1 MByte. You can build 1MB 30-pin SIMMs without parity exactly that way. Each side of the 16-chip SIMM is basically identical to four of those 1MB 30-pin SIMMs, just in an easier to handle form factor. This is a pattern that always holds: A single "side" of a PS/2 SIMM behaves very similar to a bank of four 30-pin SIMMs.

The other SIMM obviously has 16MBit chips (the next generation - at that time, each generation of memory chips had 4 times the size of the previous generation) in a x16 configuration, so two of them are enough to make the 32 bits your need for each side. 16 MBit in a x16 configuration means the 16 MBit are distributed over 1M addresses, and each of these addresses stores 16 bit. So a single of these chips behaves like a 16-bit bank made out of two 1MB 30-pin SIMMs.

Notice how I always write "behaves like 1MB 30-pin SIMMs". Your chipset is maxed out at 16MB, as long as you only use modules that "behave like (a quadruple of / an octuple of) 1MB 30-pin SIMMs". To get to the full 64MByte the chipset supports, you need modules that in total behave like 16 30-pin 4MB SIMMs. Your options for modules that behave like 30-pin 4MB SIMMs are:

4 4MB 30-pin SIMMs
a 16MB PS/2 SIMM that behaves the same as 4 4MB 30-pin SIMMs
a 32MB PS/2 SIMM that behaves the same as 8 4MB 30-pin SIMMs

4 30-pin SIMMs + 3 16MB PS/2 modules is a configuration explicitly listed in your mainboard documentation. 2 32MB PS/2 modules in the same slots you currently have the 8MB modules in is also very likely to work. And of course, you don't need to go all the way to 64 MB, but you can stop at 32MB (e.g. from one 32MB or 2 16MB PS/2 modules).

Ok thanks a lot for the detailed explanations.
I must say that now I understand how I could do to increase the RAM beyond 16MB, while before I was trying and in the end I was unable to read further.
I remember trying various RAM of lower and higher capacity, those of lower capacity went but I could not go beyond 16MB, I remember trying 72 pin RAM of higher capacity, two pieces of 16MB each (in total therefore 32MB), but not I remember if one read it and the other did not, for convenience I put the two 8MB, so as to use the two 16MB in another PC, currently those modules I think ended up in the P.PRO-200.
So I'll try to look for 8MB RAM similar to those with 16 chips, or 16MB with 16 chips, I remember having 4 identical modules for a total of 32MB (4x8MB), maybe I have others similar but I should see how they are, anyway thanks to your suggestions I might be able to bring the total RAM to 32MB, if I found 64MB (2X32MB) even if they would be fine I would not use them in that PC, rather I would put them in the P.PRO-200 which currently has 64MB (4X16MB) or in the P. 133 (idem).
Going back to 30-PIN RAM, so the 4MB ones use the same chips as the 16MB 72-PIN ones?
I posted the photo of that SIMM30-4X30PIN, but from what I understand besides being bulky, it would require the use of 30 PIN RAM all the same?
However I really appreciate the work of creation and production of "Tiido", it could be very useful probably for a 486 Overdrive DX4-100 VLB, in that case I would have to decide whether to put 16MB (4X4MB) or 32MB (8X4MB), at the limit I could easily get there to 20MB (4x4MB + 4X1MB).

PC@LIVE wrote on 2021-07-03, 20:57:

So I'll try to look for 8MB RAM similar to those with 16 chips, or 16MB with 16 chipsI remember having 4 identical modules for a total of 32MB (4x8MB), maybe I have others similar but I should see how they are,

Don't bother with 8MB PS/2 modules. No matter how many chips they have, they always have 2 ranks of 32 bits with 1M locations. The only difference between different 8MB modules is how many of the 32 bits per location is stored inside one chip. With chips/modules offering only 1M addresses, you are maxed out at 16MB.

PC@LIVE wrote on 2021-07-03, 20:57:

Going back to 30-PIN RAM, so the 4MB ones use the same chips as the 16MB 72-PIN ones?

Essentially, yes. 4MB 30-pin SIMMs offer 4M addresses and 8 bit per address. 16MB 72-PIN ones offer one rank of 4M addresses with 32 bits. You can build a 16MB PS/2 SIMM using 2 chips of 4M x 16 chips, but you can't build 30-pin SIMMs, because cutting the 16-bit chips in half won't make two 8-bit chips.

PC@LIVE wrote on 2021-07-03, 20:57:

I posted the photo of that SIMM30-4X30PIN, but from what I understand besides being bulky, it would require the use of 30 PIN RAM all the same?

I don't see any jumpers on that 4x30 pin adapter. I don't see how you can build an adapter like that without jumpers that supports multiple capacities. From the look of it, I guess it is so old that it is used to convert 4 30-pin SIMMs of 256KB into one virtual SIMM of 1MB, not to convert 4 30-pin SIMMs of 1MB into one 30-pin SIMM of 4MB. As the general idea is the same, you could cut some traces and add bodges to convert it to a different capacity, but I don't think it is worth the trouble: You would need four 30-pin SIMMs of the same type for the board to recognize them. In theory, you can mix "real" 4MB SIMMs and virtual 4MB SIMMs made using an adapter like that. You likely can't fit four of these adapters next to each other into the board, just for mechanical reasons, so you would need some real 4MB SIMMs anyway. And if you get 4MB SIMMs, it's best to acquire a pack-of-4 in the first place. Furthermore, the adapter needs some time to decide which of the 4 SIMMs needs to be addressed, so the access time of the virtual 4MB SIMM would be worse than the access time of the individual 1MB SIMMs. And last, but nor least, the adapter might translate the accesses in a way that it break fast page mode, so it would only work in mainboards not using page mode (mostly 286 mainboards).

mkarcher wrote on 2021-07-04, 07:33:

Don't bother with 8MB PS/2 modules. No matter how many chips they have, they always have 2 ranks of 32 bits with 1M locations. T […]
Show full quote

PC@LIVE wrote on 2021-07-03, 20:57:

So I'll try to look for 8MB RAM similar to those with 16 chips, or 16MB with 16 chipsI remember having 4 identical modules for a total of 32MB (4x8MB), maybe I have others similar but I should see how they are,

Don't bother with 8MB PS/2 modules. No matter how many chips they have, they always have 2 ranks of 32 bits with 1M locations. The only difference between different 8MB modules is how many of the 32 bits per location is stored inside one chip. With chips/modules offering only 1M addresses, you are maxed out at 16MB.

PC@LIVE wrote on 2021-07-03, 20:57:

Going back to 30-PIN RAM, so the 4MB ones use the same chips as the 16MB 72-PIN ones?

Essentially, yes. 4MB 30-pin SIMMs offer 4M addresses and 8 bit per address. 16MB 72-PIN ones offer one rank of 4M addresses with 32 bits. You can build a 16MB PS/2 SIMM using 2 chips of 4M x 16 chips, but you can't build 30-pin SIMMs, because cutting the 16-bit chips in half won't make two 8-bit chips.

PC@LIVE wrote on 2021-07-03, 20:57:

I posted the photo of that SIMM30-4X30PIN, but from what I understand besides being bulky, it would require the use of 30 PIN RAM all the same?

I don't see any jumpers on that 4x30 pin adapter. I don't see how you can build an adapter like that without jumpers that supports multiple capacities. From the look of it, I guess it is so old that it is used to convert 4 30-pin SIMMs of 256KB into one virtual SIMM of 1MB, not to convert 4 30-pin SIMMs of 1MB into one 30-pin SIMM of 4MB. As the general idea is the same, you could cut some traces and add bodges to convert it to a different capacity, but I don't think it is worth the trouble: You would need four 30-pin SIMMs of the same type for the board to recognize them. In theory, you can mix "real" 4MB SIMMs and virtual 4MB SIMMs made using an adapter like that. You likely can't fit four of these adapters next to each other into the board, just for mechanical reasons, so you would need some real 4MB SIMMs anyway. And if you get 4MB SIMMs, it's best to acquire a pack-of-4 in the first place. Furthermore, the adapter needs some time to decide which of the 4 SIMMs needs to be addressed, so the access time of the virtual 4MB SIMM would be worse than the access time of the individual 1MB SIMMs. And last, but nor least, the adapter might translate the accesses in a way that it break fast page mode, so it would only work in mainboards not using page mode (mostly 286 mainboards).

ok thanks for the detailed explanations, they are very useful, I thought it was easier, but I understand it's not what I thought.
In fact the best solution is to get the 4MB 30PIN RAM, rather than anything else, obviously it depends on the PC, a 286 with 4MB total is already enough, a 386 with 8MB is fine, going beyond 8MB I guess will speed up the system, I don't know than what I think is visible.

Awesome effort. I'm sure it is greatly appreciated by the community as a whole...😉

Tildo,

Could you send me a PM so I can place an order for some of these please?

Amibay registrations are closed, and I am too "junior" on this board to send a PM.

Thanks in advance!

damianmcmillan wrote on 2021-10-28, 05:40:

Tildo, […]
Show full quote

Tildo,

Could you send me a PM so I can place an order for some of these please?

Amibay registrations are closed, and I am too "junior" on this board to send a PM.

Thanks in advance!

you can find the contact email at the bottom of this page: http://tmeeco.eu/ABOUT.HTM

Yeah, the e-mail seen in the image at the bottom is a sure way of reaching me.