..and the rarest of the rare is the 40 pin proprietory simm as used in the T5200 ..
Ian_B found a way tound this by utilising a 72 pin EDO hand wired..
Had there been an alternative ..then maybe he and others might have been interested.
A worthwhile project though..
keep up the good work buddy.
rgds
VS

4MB 30-pin Simms are readily available for fairly cheap on eBay.

Is it even worth the time and money to make some yourself?

I have had thoughts of making some 16MByte 30pin SIMMs and 64MB+ 72pin ones but finding suitable DRAMs has turned out to be difficult. Making a PCB and puttings chips on them doesn't take much effort, I could probably make 1 module every 2..3 minutes or so.

cyclone3d wrote:

4MB 30-pin Simms are readily available for fairly cheap on eBay.

Just found only two offers below 10 €/ a piece on ebay.de, thats not cheap in my opinion. Needing 8 of these for a usual 486er makes the situation even worse. Most offers are 4 MB/set.

Tiido wrote:

finding suitable DRAMs has turned out to be difficult

As I wrote there are plenty of PS/2 SIMMs availiable to harvest the chips from. Of coarse only FPM modules can be used for 30pin conversion.

Needing 8 of these for a usual 486er makes the situation even worse.

486 boards with 30 pin memory are not optimal anyway.

eisapc wrote:

As I wrote there are plenty of PS/2 SIMMs availiable to harvest the chips from. Of coarse only FPM modules can be used for 30pin conversion.

Ok, taking a closer look, 4MB is easily doable, but for 16MB there are no options in that route that I see and those are what I'm most interested in. After I'm done with my current project I'll make a handful of 4MB modules to max out my 386 and convert few 72pin 32MB EDO sticks into FPM ones that my 486 can use.

With help from CPLD and monster sized sticks it would be possible to get 16MB going with multiple 2MByte sized chips... I don't need more ideas right now hahahaha

Yes, I am thinking of doing 16MB 30pin SIMMs as well...though I want to do it with modern components so that I can run them at 50MHz with as few wait states as possible.

For PLDless operation there aren't any choices that I have found but once you decide on the PLD route you can do anything as long as oversided modules work. There's also the 5V intolerance on all the faster DRAMs (i.e under 50ns). Digikey has some insanely expensive 5V 2Mbyte 4bit 70ns EDO chips, other places I looked at don't have any 5V offerings and none of the 3.3V chips I see have 3.3V tolerant IO, level translation will allow some leeway though but at reduced performance due to 5...10ns latency the add. I haven't had luck finding larger DRAMs from China either, neither good price or any sort of quantity.
If speed is the only factor then one could make a 4MByte 5V SRAM based module for a bit under 25€ a stick, with guaranteed performance of 80MHz without wait states. 8 chips are needed + a CPLD making for a pretty large module. For a 386 or a 486 you're gonna need 100€ worth of sticks to get going at all in case of 30pin SIMMs... is that 100€ of 16Mbytes of insanely fast memory worth the money ?

386_junkie already has a project on this:

Re: Best 386 Motherboard?

I think he's just removing ICs from old SIMMs and soldering on NOS 40ns parts, probably 1Mbit density. What we really need is a new PCB using modern ICs.

kixs wrote:

386_junkie already has a project on this:

Re: Best 386 Motherboard?

Thanks kixs.

To confirm what most people are saying here... with the number, varieties, and different densities of SIMM's out there... there is really no reason to create your own SIMM's unless there is a specific application for which there is not a product or solution that already exists.... hence the requirement for custom parts.

I already have a collection of 1MB, 4MB, and 16MB... 30 pin SIMM's. The lowest latency available in each of these is 60ns... which is a limiting factor and is the very reason I have been experimenting and testing the various ways of making my own SIMM's.

Why am I doing this? On some of my 386 systems, running at 40MHz; 1/40 = 25ns clock cycles .... and 50MHz; 1/50 = 20ns clock cycles, the lowest latency DRAM requires more clock cycles than required with the CPU running 3x faster: -

60ns DRAM... 3 x 25ns clk cycles = 75ns or 3 x 20ns clk cycles = 60ns!

The solution to reduce latency was to create 40ns 30-pin SIMM's, which I have done, thus: -

40ns DRAM... 2 x 25ns clk cycles = 50ns or 2 x 20ns clk cycles = 40ns! 😁

The result... reducing DRAM as a bottleneck, creating a 386 that has lower system latency and faster RAM access.

Anonymous Coward wrote:

I think he's just removing ICs from old SIMMs and soldering on NOS 40ns parts, probably 1Mbit density. What we really need is a new PCB using modern ICs.

This is the quick and dirty method yes.

Up until now, whilst sacrificing, learning, testing... and wasting, 1MB is fine.

This method will also work with 4MB SIMM's (as pictured below)... I have enough IC's and will start to convert these also when I am ready.

The Serpent Rider wrote:

Needing 8 of these for a usual 486er makes the situation even worse.

486 boards with 30 pin memory are not optimal anyway.

The bus is not the limiting factor... it is the latency of the IC chips that determine system performance.

60ns 30 pin SIMM's are more optimal than 70ns 72 pin SIMM's! If you find 72 pin SIMM's with a lower latency than 60ns then yes, they will be more optimal than what is typically available on a 30 pin PCB.

cyclone3d wrote:

4MB 30-pin Simms are readily available for fairly cheap on eBay.

Is it even worth the time and money to make some yourself?

Yes there are plenty of them around however at the prices that most go for put people off.

nforce4max wrote:

cyclone3d wrote:

4MB 30-pin Simms are readily available for fairly cheap on eBay.

Is it even worth the time and money to make some yourself?

Yes there are plenty of them around however at the prices that most go for put people off.

Guess I am just lucky in the US. I can get brand new 4x 4MB 60ns simms for $22 shipped.
https://www.ebay.com/itm/16MB-4x-4MB-30pin-SI … PC/162289926468

What is the shipping coming up as for where you are?

The Serpent Rider wrote:

486 boards with 30 pin memory are not optimal anyway.

Will be O.K. for me as there are only few EISA boards around with PS/2 slots, most come with 30 pin, probably the same for VL bus boards.
You are right if talking about PCI based 486er.

386_junkie wrote:

60ns 30 pin SIMM's are more optimal than 70ns 72 pin SIMM's! If you find 72 pin SIMM's with a lower latency than 60ns then yes, they will be more optimal than what is typically available on a 30 pin PCB.

60 ns FPM chips seem to be pretty rare. Most in my stock is 60 ns EDO or 70 ns FPM.

Anonymous Coward wrote:

Yes, I am thinking of doing 16MB 30pin SIMMs as well...though I want to do it with modern components so that I can run them at 50MHz with as few wait states as possible.

16 MB should be possible using 2 Mbit x8 chips from 8 MB SIMMs. As mentioned before modern components are usually designed for 3.3 V or even less supply voltage and not capable of the 5V.

cyclone3d wrote:

What is the shipping coming up as for where you are?

They ask 14.50 $ shipping to Germany.

I've heard of custom SIMMs before, albeit mainly for the 64-pin variety that the Macintosh IIfx uses instead of the more standard 30-pin or 72-pin of its contemporaries. It's apparently the only way you're ever going to get the full 128 MB in one by maxing out all the DIMM slots.

The trick lies in finding the correct RAM ICs, hopefully in some kind of new-old-stock quantity and not pulls salvaged from existing SIMMs. If you can source those, then designing the PCB to put them on is relatively straightforward.

386_junkie wrote:

Thanks kixs. […]
Show full quote

kixs wrote:

386_junkie already has a project on this:

Re: Best 386 Motherboard?

Thanks kixs.

To confirm what most people are saying here... with the number, varieties, and different densities of SIMM's out there... there is really no reason to create your own SIMM's unless there is a specific application for which there is not a product or solution that already exists.... hence the requirement for custom parts.

I already have a collection of 1MB, 4MB, and 16MB... 30 pin SIMM's. The lowest latency available in each of these is 60ns... which is a limiting factor and is the very reason I have been experimenting and testing the various ways of making my own SIMM's.

Why am I doing this? On some of my 386 systems, running at 40MHz; 1/40 = 25ns clock cycles .... and 50MHz; 1/50 = 20ns clock cycles, the lowest latency DRAM requires more clock cycles than required with the CPU running 3x faster: -

60ns DRAM... 3 x 25ns clk cycles = 75ns or 3 x 20ns clk cycles = 60ns!

The solution to reduce latency was to create 40ns 30-pin SIMM's, which I have done, thus: -

40ns DRAM... 2 x 25ns clk cycles = 50ns or 2 x 20ns clk cycles = 40ns! 😁

The result... reducing DRAM as a bottleneck, creating a 386 that has lower system latency and faster RAM access.

You've built the better mousetrack (err 386 machine)

eisapc wrote:

386_junkie wrote:

60ns 30 pin SIMM's are more optimal than 70ns 72 pin SIMM's! If you find 72 pin SIMM's with a lower latency than 60ns then yes, they will be more optimal than what is typically available on a 30 pin PCB.

60 ns FPM chips seem to be pretty rare. Most in my stock is 60 ns EDO or 70 ns FPM.

Not as rare as you may think...

https://www.ebay.com/itm/5x-HYUNDAI-HY531000A … te/182736425076

https://www.ebay.com/itm/V53C408HK60-MOSEL-20 … AM/111184910259

24/26 pin SOJ's @ 40ns? ... now they are rare!

NamelessPlayer wrote:

The trick lies in finding the correct RAM ICs, hopefully in some kind of new-old-stock quantity and not pulls salvaged from existing SIMMs. If you can source those, then designing the PCB to put them on is relatively straightforward.

This is true... but finding NOS is getting more difficult, almost impossible. A lot of dealers I've found and spoken to sell refurbished parts... I've bought small quantities and they did work ok.

IC's can be of various lengths... graphics cards IC's tend to use 4-bit wide DRAM... @ 256Kb, hence their code... 4256!

Other IC's I've come across are between 8-bits and 16-bits wide. For my project, I can only use 8-bits wide IC's as they are being used with 30-pin SIMM's.

oldpcguy wrote:

You've built the better mousetrack (err 386 machine)

Mousetrack? ... Great! Windows 95 would form part of the testing.

The SIMM's can be used on any 30-pin system... and are not limited to a 386.