Hi,

Good to see someone else looking into making their own SIMM's... I look forward to seeing your project as it progresses.

In response to your post...

rocksolid_1997 wrote:

its a simple board, with very little that can go wrong.

This depends on your soldering ability and if you kill an IC/chip with the heat... i've been soldering for years and still I can make mistakes.

rocksolid_1997 wrote:

For this project, I would love to be able to completely loose the parity, but im not sure if its needed. Does anyone have any information about whether or not the parity chip is needed by some systems ? Was it needed on most systems or actually none ?

Generally all PC systems use parity, it is other systems like Apple which does not. In any case, why do you want to loose the parity?

I had a Contaq MS-3124 386DX board that won't boot with non-parity sticks unless I clear CMOS and load BIOS default.

386_junkie wrote:

This depends on your soldering ability and if you kill an IC/chip with the heat... i've been soldering for years and still I can make mistakes.

They are fine, this isn't my first soldering job 😀

386_junkie wrote:

Generally all PC systems use parity, it is other systems like Apple which does not. In any case, why do you want to loose the parity?

Right, so parity seems to be "a thing". I wanted to loose it, solely to keep the number of ICs to a minimum. As an example, 16mb in 4mx4bit or 8mb in 2mx4b is still very much available as salvaged from e.i. 72 pin ram, but they often lack the same 1bit type parity-IC's that are found in 30 pin SIMM (this is not verified, its just a hunch).

I am still rambling to find the proper components, it would be preferred for something that is available as new - and in plenty supply. At the moment I am leaning towards something like the KM44C4100 (or compatible) with the KM41C1000 (or compatible) as parity - there is "some" stock out there, and its used commonly on (at least those i have) 4/8/16mb 72 sticks.. Alternatively find something more modern surface-mount based, but im having a hard time finding anything thats not 16 bits - given, 16bits can be used if cheap enough, and just throw away the last 8.. but.. yearh,.. 😀

Any thoughts or inputs appreciated 😀

lazibayer wrote:

I had a Contaq MS-3124 386DX board that won't boot with non-parity sticks unless I clear CMOS and load BIOS default.

But it worked with both ?

Ah, the days of the 8 chip, 9 chip, 2 chip and 3 chip (the 3 chip having two 4 bit chips and a 1 bit for parity, but with questionable compatibility)

My recollection is that it varied, some used it, some didn't, and on some it was selectable.

The other way might be, with fast enough logic, a fake parity circuit

rocksolid_1997 wrote:

lazibayer wrote:

I had a Contaq MS-3124 386DX board that won't boot with non-parity sticks unless I clear CMOS and load BIOS default.

But it worked with both ?

It worked with parity sticks without any hassle.
After loading BIOS default it works with non-parity sticks, but any mod to the settings in the BIOS will cause it freeze at memory check at boot - it won't pass 64KB of RAM.

Matth79 wrote:

The other way might be, with fast enough logic, a fake parity circuit

Isnt the problem that the parity is made from, and build by, the RAM-controller, not the RAM itself ? By calculating it, there would be parity errors on all systems with "another parity model" so to speak.

rocksolid_1997 wrote:

Matth79 wrote:

The other way might be, with fast enough logic, a fake parity circuit

Isnt the problem that the parity is made from, and build by, the RAM-controller, not the RAM itself ? By calculating it, there would be parity errors on all systems with "another parity model" so to speak.

Thge parity bit is written by the RAM controller, as the XOR of the other bits, and is read back expecting the same, so if the RAM module had a circuit to generate it on read, it could spoof parity.

http://www.pcguide.com/ref/ram/errFalse-c.html - not an original idea, and as described, a problem where the system uses clever logic across the multiple parity bits to provide ECC

Matth79 wrote:

Ah, the days of the 8 chip, 9 chip, 2 chip and 3 chip (the 3 chip having two 4 bit chips and a 1 bit for parity, but with questionable compatibility)

My recollection is that it varied, some used it, some didn't, and on some it was selectable.

IIRC, parity modules were more expensive than non-parity modules, and on some boards you could mix the two types.

Matth79 wrote:

Thge parity bit is written by the RAM controller, as the XOR of the other bits, and is read back expecting the same, so if the RAM module had a circuit to generate it on read, it could spoof parity.

Was it always calculated the same way ? I mean, was it in some specification ? (I can't seem to find any specs on how this was done). Fake Parity works, but there might as well just has been a board or two where it wouldn't work, unless ofc. theres an actual definition of it, in a spec some where.

But the hack would be simple as point out with the link, Parity in => nothing, parity out => the XOR, /CASP => nothing. I will consider it 😀

Matth79 wrote:

The other way might be, with fast enough logic, a fake parity circuit

Oh, I didn't know about it. So that's why my 386 mobo doesn't boot with these: 😵

The first 4 use 4+4+1 configuration. I have sticks like these and so far they work fine in all mobos I've tried. Though some mobos prefer 9 identical chips - that way the output should appear on all data lines at the same time so less chance the parity detection circuit will randomly trip.
Last 2 sticks are 8 chip only, clearly missing the partity one. Those might work on some mobos if in a set of similar sticks but trying to mix them with partity-enabled sticks is a bad idea.