Sorry for not keeping the exact schedule, here are some more results:
The problem doesn't seem to be affected by the video card so far.
I have tried the cirrus logic, mentioned in the opening post, and got the same results with a tseng et4000-1mb.
Also, tried every viable cdrom driver from hirens site, no success (mostly hard freeze).
The specific problem seems to be with the ram modules, as the cd driver fails to load when the 4meg motules occupy bank0 and bank1.
For the last hour, I've been leaving bank2 and bank3 empty.

Here are the advanced chipset (probably) relevant settings:
(changing these have no effect regarding the problem)
(also, these were the default values)
Hard Disk Type 47 RAM Area: 0:300 (alternative option: DOS 1KB)
Fast Gate A20 Option : Disabled
Video ROM Shadow C000, 16K: Enabled
Video ROM Shadow C400, 16K: Enabled
Adaptor ROM Shadow C800, 16K; CC00, 16K; .... ; DC00, 16K; E000, 64K: All disabled
System ROM Shadow F000, 64K: Enabled
384KB Memory Relocation: Disabled
Fast I/O Speed Option: Disabled

It's an American Megatrends 1991 bios, the limegreen-magenta-light brown uglyness.

Further findings:
When the 4x1meg modules are inserted, the bios counts 4096.
When 2x 4megs installed, bios counts 7424KB.
With 4x 4megs installed, bios counts 15872KB.

Not sure if it's relevant, the 4meg modules seem to warm up a little. Not too much, roughly body temperature. The 1meg modules stay cold.
Will try if the 4megs warm up in a 486 board or not.

Now trying with a different IDE controller, since that's the quickest of the remaining options.
Also, quick question:
I have a few ISA ram extension cards (XMS cards). Do those bottleneck the system?
Also attaching photo of the board.

Different IDE controller card, same results.
Kind of sleepy, will try tomorrow what happens with the 34pin sony with it's own controller, then again all the drivers with 386emu.

Ooo one of those city gate chipsets. I've been curious about those for a while

Thanks for the pictures !!
That chipset (Citygate D70-275 + ALD 93C001-B) was also used on a 386SX micro-AT motherboard called DAT302. It even has same Samsung KS82C6818A.
From the orig poster its bios string is 31-0100-009999-00101111-121291-TD70-O and was sold from eastern Europe
Have not dug deeper into the chips or the bios string yet.. but just did a basic search...

matskatsaba wrote on 2021-05-10, 21:14:

Not sure if it's relevant, the 4meg modules seem to warm up a little. Not too much, roughly body temperature. The 1meg modules stay cold.

That's expected. Your 1M modules use CMOS chips, as indicated by the letter "C" after "KM44", whereas your 4M modules use NMOS chips (no C after the MN41 prefix). CMOS chips typically have extremely low standby power consumption, whereas NMOS chips roughly use the same power (and produce the same heat) if accessed or not.

Have you tried this atapicd.sys https://www.hiren.info/download/dos-files/atapicd.sys driver ?
No idea if it's 386 only.
Came from here https://www.hiren.info/downloads/dos-files

There's more drivers to be found there for you to try if you feel like it.

Yeah, as per advised in previous posts, I did. None of them seemed to work at all - probably not compatible with the 286 memory management.

Today I was swamped at work (and probably tomorrow too), hopefully tomorrow evening I will have a few hours free time to try the following:
-Do a throughout ram test with the above suggested ram test programs and also the ramdrive copy (a good few passes of it)
-try a sony 4x 34pin drive with its own controller card
-regardless of previous working or not, will retry all the IDE drivers with 40pin drives using 386emu
-trying a SCSI drive

Will post results as soon as I'm done with some or all of those tests, regardless if success or fail (since I really hate it with passion when people ask for help, and disappear without trace after a few answers, so we would never which solution worked, just in case anyone else runs into the same problem.

Thanks for your help guys, I really apreciate all of you wasting your precious free time on a semi-retarded problem like this (since no sane person would want 16MB RAM in a 286, or a CD-ROM for that matter).

Also, any idea about the ram size disrepancy?
4x1meg : 4096KB -as it should
2x4meg: 7424KB -768KB less (was expecting 8192KB)
4x4meg: 15872KB -512KB less (was expecting 16384KB, or 15616KB, for the sake of consistency)
What am I missing?

Could be the bios setting is excluding some memory for other items in use, remember 286 can only address so much and the 24bit addressing lines is exactly 16,384K so any cards and others that needs the space between 640K and 1MB boundary will be taken away. Remember cards and other items that needs some address range will "steal" the memory space, leaving you with less memory capacity. Normal. Like the 4GB is seen as 3.5GB for example.

Good practice to reserve some ranges for the cards they needs via the bios setting if it has one.

Secondly, 286 can only do 64K blocks and that can also cause this.

Cheers,

matskatsaba wrote on 2021-05-12, 22:36:

Thanks for your help guys, I really apreciate all of you wasting your precious free time on a semi-retarded problem like this (s […]
Show full quote

Thanks for your help guys, I really apreciate all of you wasting your precious free time on a semi-retarded problem like this (since no sane person would want 16MB RAM in a 286, or a CD-ROM for that matter).

Also, any idea about the ram size disrepancy?
4x1meg : 4096KB -as it should
2x4meg: 7424KB -768KB less (was expecting 8192KB)
4x4meg: 15872KB -512KB less (was expecting 16384KB, or 15616KB, for the sake of consistency)
What am I missing?

I'd like me a 286 that has support for 16MB onboard. though I only got one that supports 4MB max.

I would've expected it to be 384KB less for all, since that's the amount of space between 640K and 1M where the roms normally reside.
Is there a 'memory hole remap' option in the bios that you might be able to turn off/on?

Just the usual stuff (Video ROM Shadow, adaptor ROM Shadow, System ROM Shadow, and 384KB Memory Relocation).
The thing I fail to understand why the difference is inconsistent.
You see, I would expect the 4096KB to be less, or if not, then the 8192KB to be present since the ram is not using up the whole memory map.
But then if I theorize that the above 4MB part suffers losses, then it should be consistent for all configurations above 4MB, so either the result should be exactly the same, or at least, more taken up when adding more RAM. But I'm "losing" 768KB with 8 megabytes, and that would suggest I would lose 1280KB with 4 modules, or even worse, if the 768kb memory loss would happen only on the parts above 4MB, so that would suggest a total of 3x768 KB to be lost.
So I would expect a constant amount of memory loss regardless of how much ram above 4mb there is, or progressively more if the memory is used up by some behind-the-scenes memory management by the chipset.
Basically, I would expect constant or more amount of memory to be lost, not less.
My first idea off the bat was a faulty module, but it was the same no matter which two modules I used for the 8MB configuration.
I'm trying to source more 4mb simms to be able to test this with different modules, since it might be by design, knowing there are only ram chips on the 4meg modules, unlike on the 1meg ones.

On the weekend, I will also test a few new things (ISA RAM expansion cards, a 4meg and a 2,5meg one) and check how much those report.
I have been messing with those, and some of them do actually suffer from the less amount counted by the bios, but the memory is there, the testExt does see those, however with the 4meg modules the memory seems to be actually lost, the extra 768 and 512KB not even being mapped.
Well, that was a good month ago, so I will have to retest the whole thing again for actual test results, because vague memories are not really useful in a scientific-ish experiment.

On another note:
Since I have literally kilograms of 1meg and 256KB modules, I really wonder if someone bothered to design a PCB that goes in a simm slot and can take in either 1meg simm modules or a bare board I can solder the chips on, as the price of the 4meg modules are going higher and higher for no real reason (2sided pcb, cheap AF memory chips - or even free if having a shoebox full with ram modules), so pcbway-ing seems to be an actually viable solution.

pentiumspeed wrote on 2021-05-12, 23:41:

Secondly, 286 can only do 64K blocks and that can also cause this.
Cheers,

The 286 MMU has a max segment size of 64KB..
But if memory serves, it can also handle smaller values just fine. 😀

matskatsaba wrote on 2021-05-13, 09:40:

Just the usual stuff (Video ROM Shadow, adaptor ROM Shadow, System ROM Shadow, and 384KB Memory Relocation).

the 384KB memory relocation is the one. different label, same beast. memory hole...

The thing I fail to understand why the difference is inconsistent. You see, I would expect the 4096KB to be less, or if not, the […]
Show full quote

The thing I fail to understand why the difference is inconsistent.
You see, I would expect the 4096KB to be less, or if not, then the 8192KB to be present since the ram is not using up the whole memory map.
...
I'm trying to source more 4mb simms to be able to test this with different modules, since it might be by design, knowing there are only ram chips on the 4meg modules, unlike on the 1meg ones.

Agreed, though I'm thinking this is more chipset/bios related than simm modules, since I don't believe that any on boot bad memory remapping was ever done in that day and age.

On another note:
Since I have literally kilograms of 1meg and 256KB modules, I really wonder if someone bothered to design a PCB that goes in a simm slot and can take in either 1meg simm modules or a bare board I can solder the chips on, as the price of the 4meg modules are going higher and higher for no real reason (2sided pcb, cheap AF memory chips - or even free if having a shoebox full with ram modules), so pcbway-ing seems to be an actually viable solution.

You'll need extra logic chips to make higher capacity modules with lower capacity chips like that.
There are some modules which allowed you to use 4 lower capacity ones. They even came in a left sided and right sided version.

Do you recall running any tests with the 384KB relocation off and on ?

Tried once without success. Will redive into this project this weekend and try it again, though. For most of the time, 384k reloc was disabled.

matskatsaba wrote on 2021-05-09, 09:28:

is a board print that I guess the board type: TD60K

good afternoon! do you have the ability to read the bios firmware? better programmer. I have the same board and with bios from td60c it does not start.
image - https://disk.yandex.ru/d/zxoBBbrYGTQsXw/IMG_2 … 1218_010102.jpg

matskatsaba wrote on 2021-05-13, 09:40:

On another note:
Since I have literally kilograms of 1meg and 256KB modules, I really wonder if someone bothered to design a PCB that goes in a simm slot and can take in either 1meg simm modules or a bare board I can solder the chips on, as the price of the 4meg modules are going higher and higher for no real reason (2sided pcb, cheap AF memory chips - or even free if having a shoebox full with ram modules), so pcbway-ing seems to be an actually viable solution.

They were around. Variously called SIMM stackers SIMM trees etc. There was a magazine that was distributed in the UK in mid 90s, best I can remember it was between 93 and 97, published a PCB design for one. I had it, but think I don't any longer. Possibly Practical Electronics, could also have been Electronics: The Maplin Magazine, or Everyday Electronics. ... and some chance it was some other... I think I might have bought it just for that article, so could have been one I wouldn't normally look at. I think it might have been later in the period, as I think I was thinking that if it appeared a couple of years earlier one could have "made bank" knocking them out for the local market. I have a vague image of a white background cover with a blueish illustration, I figure I'd know it if I find a gallery of covers... I'll look around a bit online.

Edit: Aha, something popped, this wasn't the issue I remember, but it's a "project collection" issue, so maybe one I saw was earlier, or maybe it was a repeat/reprint I saw later, anyway, page 62 https://worldradiohistory.com/UK/Elektor/90s/ … -1994-07-08.pdf

Is ALD the same as Citygate?

*edit*

I just found this:

Mystery “Super-286” motherboard from 1994

It seems to be the same board with no sticker on the chipset. (the actual chipset is SARC/Toshiba).
So what this implies is that "Citygate" is just a PC chips branding.

led178 wrote on 2021-12-18, 22:52:

matskatsaba wrote on 2021-05-09, 09:28:

is a board print that I guess the board type: TD60K

good afternoon! do you have the ability to read the bios firmware? better programmer. I have the same board and with bios from td60c it does not start.
image - https://disk.yandex.ru/d/zxoBBbrYGTQsXw/IMG_2 … 1218_010102.jpg

Yeah, I actually do. Sorry for the long delay, had some rl issues, but more or less back on track now.
If you still need the eprom dump, do let me know and I'll do something about it. (would have gotten it already, but the pc is stored away in a mountain of boxes prepared for moving but the renovation of my house is going waaaay slower than planned, part of the reason it took me months to do anything hobby related.

matskatsaba wrote on 2022-01-07, 10:52:

If you still need the eprom dump

Good day matskatsaba!
Glad to hear you still have the board! Yes, I'm still looking for firmware, it's not urgent, the board has been around for many years and over time it will only get better 😀
I like rescuing hardware from a landfill and I really don't want to return it there.

matskatsaba wrote on 2021-05-13, 09:40:

On another note:
Since I have literally kilograms of 1meg and 256KB modules, I really wonder if someone bothered to design a PCB that goes in a simm slot and can take in either 1meg simm modules or a bare board I can solder the chips on, as the price of the 4meg modules are going higher and higher for no real reason (2sided pcb, cheap AF memory chips - or even free if having a shoebox full with ram modules), so pcbway-ing seems to be an actually viable solution.

BitWrangler wrote on 2021-12-19, 01:45:

They were around. Variously called SIMM stackers SIMM trees etc. There was a magazine that was distributed in the UK in mid 90s, best I can remember it was between 93 and 97, published a PCB design for one. I had it, but think I don't any longer. Possibly Practical Electronics, could also have been Electronics: The Maplin Magazine, or Everyday Electronics. ... and some chance it was some other... I think I might have bought it just for that article, so could have been one I wouldn't normally look at. I think it might have been later in the period, as I think I was thinking that if it appeared a couple of years earlier one could have "made bank" knocking them out for the local market. I have a vague image of a white background cover with a blueish illustration, I figure I'd know it if I find a gallery of covers... I'll look around a bit online.

Edit: Aha, something popped, this wasn't the issue I remember, but it's a "project collection" issue, so maybe one I saw was earlier, or maybe it was a repeat/reprint I saw later, anyway, page 62 https://worldradiohistory.com/UK/Elektor/90s/ … -1994-07-08.pdf

Something else turned up while I was browsing for something else...
https://worldradiohistory.com/Archive-Radio-E … /EN-1995-09.pdf
printed page 35, page 33 of the PDF, SIMM PCBs to put DRAM on, 3 varieties.