Seems you´re not alone, here are reports that the sticks won´t work in 386 boards without the 3rd chip (parity) populated, _unless_ your mainboard provides a jumper for disabling parity.
https://dosreloaded.de/forum/thread/5088-self … ories/?pageNo=2

majestyk wrote on 2024-04-05, 09:14:

Seems you´re not alone, here are reports that the sticks won´t work in 386 boards without the 3rd chip (parity) populated, _unless_ your mainboard provides a jumper for disabling parity.
https://dosreloaded.de/forum/thread/5088-self … ories/?pageNo=2

Thank you very much for the link, majestyk! I'll read it through some online translator as I don't understand German language at all.

I thought that with disabled parity error check in BIOS there will be no problem to use these DIY FP-RAM sticks and I started to think that size and rated voltage for the ceramic SMD capacitors are the problem. Sadly my 386 boards don't have hardware option for disabling of the parity error checks 🙁

Does anyone have any idea what memory chips I can use for parity chips on these 4 MB fast page mode RAM modules?

Ok, so there's nothing wrong with the smaller SMD ceramic caps.

Now I see that these 4M x1 FP RAM chips are pretty expensive even as scrapped components from trashed electronic devices. Is there any chance to find 4 or 8 of them on some old 72-pin memory stick?

If I only knew about this 386 incompatibility problem, I'd never bothered to do this kind of self-made 4 MB FP-RAM sticks.

Shouldn´t it be possible to use 4M x 4 chips for parity as well and just leave data lines 1 - 3 unused, maybe with a 10K pullup resistor?
Seems like only data line 0 (pin2) is connected on your pcb and 1 - 3 (pins 3, 22, 23) are not.
You would then have 4M x 4 /8= 2MB, 2M / 4 = 512KB for parity.

majestyk wrote on 2024-04-05, 10:28:

Shouldn´t it be possible to use 4M x 4 chips for parity as well and just leave data lines 1 - 3 unused, maybe with a 10K pullup resistor?
Seems like only data line 0 (pin2) is connected on your pcb and 1 - 3 (pins 3, 22, 23) are not.
You would then have 4M x 4 /8= 2MB, 2M / 4 = 512KB for parity.

I have no idea how it will be possible to make 4M x 1 chip out of 4M x 4 one. I can provide a higher resolution photos of unpopulated PCB if this will help.

I still have 8 spare unused brand new 4M x 4 Samsung KM44C4100CK-6 FP-RAM chips and I'm open to any suggestions for saving this project, otherwise the DIY RAM modules are going in the trash bin. So far I haven't had more disappointing failure related to any computer projects 🙁 A lesson learned - never ever again blindly trust some colorful promising retrocomputer YT-content.

The pads 2 and 3 for the parity chips are intended and designed to be connected/used, I'm not sure about 22 and 23:

P.S. I've downloaded gerber-files for the project and since it's free project (still not fully open source - only gerber-files are distributed freely) I think it will not be a problem to provide any of them here in order to resolve this unexpected issue.

Would it not all be solved simply by soldering your leftover chips on the free space (IC3) on the PCB? I mean, use the same type of chips and solder them on in exactly the same way. Sure you will then use 6MB's of Chips for every Net 4MB stick, but anyway... That is what parity means in the hard drive world after all, like in the smalles possible Raid 5 configuration, you need 3 disks. Two for data and one for parity. So your net capacity is in a 3-way config, the result of 2.

hilram wrote on 2024-04-05, 11:17:

Would it not all be solved simply by soldering your leftover chips on the free space (IC3) on the PCB? I mean, use the same type of chips and solder them on in exactly the same way. Sure you will then use 6MB's of Chips for every Net 4MB stick, but anyway...

I think I can't use 4M x 4 chips in place of 4M x 1 parity chips directly. The full capacity using right two 4M x 4 chips (I have only these) and one 4M x 1 chip (these are expensive and hard to find) is 4.5 MB for stick, not 6 MB.

Maybe you can try to resolder the pins of all the chips, I use the same memory chip, but the PCB version is Rev.24, it does not have the EDO/FPM switch, it works fine on both the 486DLC and 386SLC motherboards, in the BIOS Memory parity is turned off.

zwrr wrote on 2024-04-05, 11:58:

Maybe you can try to resolder the pins of all the chips, I use the same memory chip, but the PCB version is Rev.24, it does not have the EDO/FPM switch, it works fine on both the 486DLC and 386SLC motherboards, in the BIOS Memory parity is turned off.

Already done that, checked again for shorts, bad solder joints and there was no positive result. Maybe this was the cleanest soldering job done by me so far.

It was strange to me that these sticks partially worked (in some occasions they managed to pass the POST RAM checking with parity check disabled in BIOS) without the capacitors soldered, maybe due to some intereference in the missing parity part (I don't have any other explanation for this strange behavior). Now it seems that my two boards with on-board soldered QFP 386 CPUs just don't like non-parity RAM sticks at all 🙁 Mind that 386SLC is more of a 486 kind processor and I'm trying these sticks on pure 386 motherboards.

4M x 4 and 4M x 1 seem to have different pinouts Here´s one for a 4M x 1:

As you can see, where the 4M x 4 has data 0 the 4 x 1 has WE#, at the 2 additional datalines of the 4M x 4 chip (pins 3 and 22->24) the 4M x 1 chip has CAS#, RAS# and at pin 23->25 it has data0. (counting differs, because it´s 24-pin vs.20 pin)

If the layout is designed for a 4M x 1 parity chip - an the 4 x 5 pads footprint says so, the pads are used, of course, but differently.

Adapting these connections with bridges on all 4 SIMMS is nonsense, so either find 4M x 1 chips or find a mainboard that takes the SIMMS without parity.
Maybe there´s another solution??

majestyk wrote on 2024-04-05, 10:28:

Shouldn´t it be possible to use 4M x 4 chips for parity as well and just leave data lines 1 - 3 unused, maybe with a 10K pullup resistor?
Seems like only data line 0 (pin2) is connected on your pcb and 1 - 3 (pins 3, 22, 23) are not.
You would then have 4M x 4 /8= 2MB, 2M / 4 = 512KB for parity.

Checked this suggestion. Well, theoretically it is possible, but in reality the 4M x 4 and 4M x 1 pin's signals does not match even if 20 out of 24 pins phisically fits on same pads. I marked all the via connections on PCB for the parity chip eleven data lines with red circles:

The pads from the upper picture don't match electrically with the 4M x 4 chip's pins (corresponding 4M x 1 pin's signals are written in red):

It is impossible some 4M x 4 chip to be used as 4M x 1 parity chip on this PCBs with any modifications. So, case closed, these useless FP-RAM sticks go to trash.

P.S. Let's never again make such stupid mistakes with unconfirmed projects. My "inner voice" told me to just buy four 4 MB FP-RAM parity factory modules on fleabay instead to waste a lot of time and some money in this project, but the stupid curiosity to make rare things by myself...

P.S. 2: Sorry, majestyk, just saw your last post which gives the same info.

Yeah, there's no options for such a modifications. One 4M x 1 FP-RAM chip costs more than the one of these modules equipped with two 4M x 4 chips not including the shipping costs. And when I "draw the line" all this will exceed the cost of four pieces of 4 MB parity factory modules bought used. There's absolutely no point in messing with these non-working RAM modules anymore.

Once again, thanks for your participation and information on the topic!

analog_programmer wrote on 2024-04-05, 13:53:

Yeah, there's no options for such a modifications. One 4M x 1 FP-RAM chip costs more than the one of these modules equipped with two 4M x 4 chips not including the shipping costs.

I was surprised by that claim. 4M x 1 chips were very common in those days, and should be abundant even today. I checked https://www.ardent-tool.com/memory/Chip_Conversion.html for chip numbers of 4M x 1 chips, and immediately found an offer at https://shop.tvsat.com.pl/en_GB/p/10pcs-M5M44 … ITSUBISHI/16417 offering 10 chips for 7,88€, shipping below 5€ to European countries and below 10€ to the US. That's below 2$/2€ per chip. This is not an endorsement for this specific offer, but just meant as an example that the current market price for 4M x 1 chips is not that absurdly high that you should scrap the otherwise perfectly good 4M modules.

mkarcher wrote on 2024-04-05, 21:34:

I was surprised by that claim. 4M x 1 chips were very common in those days, and should be abundant even today. I checked https://www.ardent-tool.com/memory/Chip_Conversion.html for chip numbers of 4M x 1 chips, and immediately found an offer at https://shop.tvsat.com.pl/en_GB/p/10pcs-M5M44 … ITSUBISHI/16417 offering 10 chips for 7,88€, shipping below 5€ to European countries and below 10€ to the US. That's below 2$/2€ per chip. This is not an endorsement for this specific offer, but just meant as an example that the current market price for 4M x 1 chips is not that absurdly high that you should scrap the otherwise perfectly good 4M modules.

Thanks, but I can't order these. I can order from another place for about EUR 6 each 4M x 1 chip without shipping included. As I already wrote, one of these modules in the state as on the picture in my first post (without parity chip) costs me about EUR 5.50. I think still 5.5 is less than 6. Yes, the 4 MB 30-pin FP-RAM modules are rare, but here nobody cares about ancient computers, so if I find any on the flea-market, they will cost me no more than EUR 10 for piece. My English is not very good, but exactly which one of these words you don't understand?

this place is an old school shop, I think they are around since at least the eighties. they also sell on allegro and ebay
https://allegro.pl/oferta/10szt-m5m44100cj-6-12961808412
https://www.ebay.com/itm/393079894607

I bought from them countless times and the service was perfect each time.

rasz_pl wrote on 2024-04-06, 07:47:

this place is an old school shop, I think they are around since at least the eighties. they also sell on allegro and ebay
https://allegro.pl/oferta/10szt-m5m44100cj-6-12961808412
https://www.ebay.com/itm/393079894607

I'm sorry, it was my mistake. In your previous post I saw in the link "shipping to UK" and I thought it's a UK-based shop, but actually it in Poland. The shipping cost is OK if there's no VAT or custom's taxes to other eu regions.

Thank you for the links! I'll think about whether it's worth ordering 10 pcs of 4M x1 FP-RAM chips. Maybe if I find suitable memory chips for RAM upgrade of some old VGA cards it would be justified to order several types of chips with one delivery.

I'll check this allegro.pl through some online translator too.

My ill-fated non-parity RAM DIY modules are sill lying around.

majestyk wrote on 2024-04-06, 07:55:

I bought from them countless times and the service was perfect each time.

From which shop you've ordered before?

P.S. It will be of use if there was a topic in the forum with info for such online electronic parts shops (excluding well known digikey, farnell, mouser, etc.) with international shipping in the eu countries.

analog_programmer wrote on 2024-04-06, 08:13:

From which shop you've ordered before?

This referred to "tvsat" in Poland.

all three links are to the same tvsat shop from Poland 😀