The BIOS (and KBC) stickers and the BIOS ID-code (1223) say Biostar.
This means the board was probably at least marketed and possibly made by Biostar (assuming the 3 chips are original...).

Thanks! Yeah. It looks like Biostar. Didn't know that they also used VLSI chipsets.

I actually could make it work with 4 MB of RAM. But unstable. Don't know why yet.

This chipset definitely supports 4MB.
Are all the contacts of the SIMM slots clean? Are all the solderings of the PLCC 84 chips beyond all doubt?

I´m having a similar issue at the moment:
Parity Errors On A 286 Unitron U3911-V2 Mainboard

majestyk wrote on 2023-08-30, 19:38:

Are all the contacts of the SIMM slots clean? Are all the solderings of the PLCC 84 chips beyond all doubt?

SIMM contacts are fine, in the slot and on the SIMMS itself. But I cleaned them up just in case. Did not help. And also checked this memory on 386 DX40. Works great there.
The weirdest thing is that the memory is actually passing the test on boot, and even passes the memory test on Himem.sys, and even in Checkit3. And then when I'm trying to use for instance as a ram disk it fails with different errors.
Completely the same thing on 386 dx40: same memory, and the same XT-IDE drive - works perfectly.
I tried different CPU, different video card, jumper settings. Turned off turbo mode. Nothing helped.
p.s. Thanks for the URL to your thread. Seems we have a pretty similar problem.

majestyk wrote on 2023-08-30, 19:38:

I´m having a similar issue at the moment:

I actually was wrong. We have very different problems. In my case, this seems to be a software problem. In your case, you are experiencing a real physical memory fault. Some of the chip on your board doesn't work as it should. And it should be easy to debug because BIOS tells you, where exactly the problematic address. In my case HIMEM.SYS from MSDOS 6.22 seems to be doesn't support properly memory above 1Mb on a 286 CPUs. I found this article https://dosdays.co.uk/topics/ht12_xms_ems.php in which the author suggests EMM286 and HT12MM.SYS instead of HIMEM.SYS, but I can't make it work yet.

maksg wrote on 2023-08-30, 23:49:

SIMM contacts are fine, in the slot and on the SIMMS itself. But I cleaned them up just in case. Did not help. And also checked […]
Show full quote

majestyk wrote on 2023-08-30, 19:38:

Are all the contacts of the SIMM slots clean? Are all the solderings of the PLCC 84 chips beyond all doubt?

SIMM contacts are fine, in the slot and on the SIMMS itself. But I cleaned them up just in case. Did not help. And also checked this memory on 386 DX40. Works great there.
The weirdest thing is that the memory is actually passing the test on boot, and even passes the memory test on Himem.sys, and even in Checkit3. And then when I'm trying to use for instance as a ram disk it fails with different errors.
Completely the same thing on 386 dx40: same memory, and the same XT-IDE drive - works perfectly.
I tried different CPU, different video card, jumper settings. Turned off turbo mode. Nothing helped.
p.s. Thanks for the URL to your thread. Seems we have a pretty similar problem.

Your problem is that you are using SIMMs with only 3 chips. This chipset requires SIMMs with 9 chips. The issue (and symptoms) you are experiencing are due to the 3 chip modules having slightly different memory refresh timing requirements that the VLSI 100/200 series chipsets do not meet.

It's not software related, the reason you only see problems with using ramdisks and the like is that the data in memory is only "going bad" after it sits for a period, due to inadequate refresh cycles.

It behaves oddly and unpredictably. Last night I booted that SBC into a plain DOS and everything worked just fine. This morning I can't even boot it without errors. You are probably right - this memory is not physically compatible with this SBC.

maksg wrote on 2023-08-31, 15:55:

It behaves oddly and unpredictably. Last night I booted that SBC into a plain DOS and everything worked just fine. This morning I can't even boot it without errors. You are probably right - this memory is not physically compatible with this SBC.

I'm definitely right. I have a board with that exact same chipset. Find some 1MB modules with 9 physical ICs on them, trust me.

maksg wrote on 2023-08-31, 14:20:

majestyk wrote on 2023-08-30, 19:38:

I´m having a similar issue at the moment:

I actually was wrong. We have very different problems. In my case, this seems to be a software problem. In your case, you are experiencing a real physical memory fault. Some of the chip on your board doesn't work as it should. And it should be easy to debug because BIOS tells you, where exactly the problematic address. In my case HIMEM.SYS from MSDOS 6.22 seems to be doesn't support properly memory above 1Mb on a 286 CPUs. I found this article https://dosdays.co.uk/topics/ht12_xms_ems.php in which the author suggests EMM286 and HT12MM.SYS instead of HIMEM.SYS, but I can't make it work yet.

In my case it´s definitely not the chips (I just updated my topic this morning).
I tried 5 different sets of RAM, tested them all for hours with Memtest, swapped the chips a hundred times, but the errors persist.

Update: with an oscilloscope and VL82C101 datasheet I could find which switch is responsible for the RAM wait-state.

SW1.5-on/off = RAM wait state. on=low(0), off=high(1)

Also noticed quite a weird thing on the SBC...

Traces with an "open end" are often grounded and have some shielding function to prevent signals from interfering with each other.
Otherwise I could think of a sloppy rearrangement between different revisions of this SBS, but that wouldn´t be Taiwan-like...

I was looking for any evidence that the memory wait state really works and affects something. And found after being hooked to pin 30(RAS) and pin 66(ENDRAS) on the VL82C101. So it's definitely affecting but without positive results in general. Still having parity issues and waiting for 9 chips SIMMS which I already bought.

BIOS

maksg wrote on 2023-09-03, 20:05:

Also noticed quite a weird thing on the SBC...

majestyk wrote on 2023-09-04, 06:45:

Traces with an "open end" are often grounded and have some shielding function to prevent signals from interfering with each other.
Otherwise I could think of a sloppy rearrangement between different revisions of this SBS, but that wouldn´t be Taiwan-like...

https://resources.altium.com/p/guard-traces-hit-or-myth

I'm not a complete noob in electronics. It’s not that I’ve never heard of impedance and tracks that serve as capacitance on the board. 😉 But in my opinion, this is clearly not the case. This track comes out of the capacitor connected with the 9.6 Mhz oscillator which is connected to the WD IDE controller. I doubt that such a solution was necessary there at that time for 286. But who knows?

maxtherabbit wrote on 2023-08-31, 16:05:

I'm definitely right. I have a board with that exact same chipset. Find some 1MB modules with 9 physical ICs on them, trust me.

Thanks for the advice! It helped. Problem solved. It works just perfectly with 9 chips SIMMs. 4 Mb now on the board.