FWIW I have found a later version of the extended memory test program, functions the same as the one above but has some extra checks to make sure the CPU is not in V86 mode when testing. I'm attaching it here because the only place where it can be downloaded demands registration to download.

(edit) found a free download of the same at https://archive.org/details/msdos_TESX34_shareware

Thanks for that! For me, the former version is running well. But as already mentioned, I dont think its the optimal program for that purpose. Much time is wasted for the testing, a simple Scan would be sufficient and should speed up the procedure.

Also figured out most Jumpers of the no-name board, see above. As for the other one, they are not "in order" and randomly distributed.

However, its really fun to test some combinations and figure out the remaining settings by doing some bit-wise calculations. I test around 10 settings, try to use all differrent switches and the results are sufficient to calculate the remaining ones.

After having some experience the following hints:
- make sure all chips are OK. If you have 1 defective one in between, it scrambles the whole board and you get totally weird results.
- I found it useful to populate 1 bank completely (shows as 0000 in testext.exe) and the other ones only with one chip (will show as e.g. FFFE depending on its position). With this method you can see and distinguish the location of the different banks.
- I think most boards are designed for 256 k chips and not 64 k

Predator99 wrote:

- I think most boards are designed for 256 k chips and not 64 k

I have a board (Kouwei KW-518) that supports 64k chips but only in the first bank, and only if they're used to fill conventional memory from 512k to 640k.

Hm here's something interesting. I have had this board for a while. I have been unable to find any information about it at all. It's marked "M A Systems" which doesn't bring anything in any searches. There are 2 PCBs stacked; each has 10 banks of 9 chips, the chips are all 256kx1 DRAM. So theoretically it has 5MB DRAM.

The current state of the switches (which I haven't changed since I got the card)

And the result of testing tonight. I only had time for 1 run, and I chose to start with the default switch setting as pictured above.

Weird pattern of errors in the middle of the memory range detected. Also only 2.5MB show up instead of the 5MB expected. Unfortunately all memory chips are soldered not socketed so I can't remove/replace easily. Also the 2 PCBs are connected by nylon standoffs which would need to be cut to separate them as they don't come apart (they do have screw tops but the bottoms are melted together). There is a connector with 2 rows of approx. 30 pins each between the 2 boards - about as long as a 8-bit ISA socket which is IIRC 2x31 contacts. Looks like a standard pin+socket connector.

Nice board and already a good result to start with!

There is no random pattern. 4000 hex = 0100000000000000. There is exactly 1 chip defective and you have a good idea where it is located.

Maybe there is a jumper to activate the daughter-board? You can already start to play with the jumpers and fix that bit-error later.

I would propose to switch the 10 DIPs 1-10 one after one and note the results. Set one siwtch back after modifying the next. You have 10 runs and get a good idea what the DIPs are for.

Good point. I have to think in binary more.
There are no jumpers on the board, but maybe one of the DIP switches controls that.
I think I will have to take the assembly apart at one point, but before that I will do some more test runs. Which I may have to repeat later just with the base board w/o the daughterboard. Especially since I don't know which direction the banks are numbered. The only visible bank mark is "bank 4" which shows under the only column of chips not covered by the daughterboard. That would be the 10th column counting from the left, or the first from the right. DOesn't make much sense: assuming 1 bank=2 columns, if counting from the left the 10th column would be in bank 5 not 4.

(edit) since the repetitive error shows up over a range of 512k and the chips are 256k, that would make 2 defective chips, I think.

stamasd wrote:

(edit) since the repetitive error shows up over a range of 512k and the chips are 256k, that would make 2 defective chips, I think.

No 😉

One "block (half row)" in testext.exe correspond to 512 kbyte.
16 bit (e.g. FFFF) are 64 kbyte.
8 x 16 bit = 1 "block" = 512 kbyte

Therefore 16 bit in testext.exe correspond to one 256kbit "chip".

However, organisation of the card seems to be special. 1 chip (out of 16 chips) of a bank correspond to one specific bit of the 16 bits in testext.exe. Therefore its repeated 8 times.

--> "4000" is 1 defective chip to make it short 😉

Gonna have to take a moment to wrap my head around that.

stamasd wrote:

Gonna have to take a moment to wrap my head around that.

I show you some quick pictures as you can not test with your soldered-chips board 😉

Board is set to starting adress 2.0 M.

Bank 1, 2 and 6 are fully populated, showing at 2.0, 2.5 and 4.5 MB respectively. 512 kbyte each.

The spare chip I set on bank 5 seems to be defective, ignore that...

But the one on bank 3 appears as "FDFF" = 11111101 11111111. The "0" indicates where this single IC is set. Maybe its similar on your board.

I'd be interested in finding cards that work with 8-bit ISAs. The lowtech cards can provide Expanded Memory or Upper Memory Blocks -- but I want both!

The AST Rampage is one of them. 8-bit ISA, can provide up to 2M UMB/EMS.

Made a little progress today on the card. Wish I had more time/energy, but I only get 2h max after I come from work before I have to go to bed or I'll fall asleep on the couch.

Flipped all switches to "on", testing shows this:

All switches to "off", surprise! At next boot, 2M extended RAM found by the BIOS.

Testing shows this:

So 3.5M XMS seen, of which 2M can be used due to the defective chip.

Next tests were with: all switches off except for sw1=on; all off except for sw2=on; all off except for sw3=on; all off except for sw4=on. I'm not going to post images of those because they were identical to the test above with all off. That's all I had time for today.

stamasd wrote:

The AST Rampage is one of them. 8-bit ISA, can provide up to 2M UMB/EMS.

How long is the card though....?

infiniteclouds wrote:

stamasd wrote:

The AST Rampage is one of them. 8-bit ISA, can provide up to 2M UMB/EMS.

How long is the card though....?

It's a full size ISA card.

If it's a full-size 16-bit ISA that is 8-bit compatible than that won't do for a Tandy 🙁

No it's a full size 8bit ISA card. There are later 16bit versions but the original Rampage is 8bit.

stamasd wrote:

Made a little progress today on the card. Wish I had more time/energy, but I only get 2h max after I come from work before I hav […]
Show full quote

Made a little progress today on the card. Wish I had more time/energy, but I only get 2h max after I come from work before I have to go to bed or I'll fall asleep on the couch.

Flipped all switches to "on", testing shows this:

All switches to "off", surprise! At next boot, 2M extended RAM found by the BIOS.

Testing shows this:

So 3.5M XMS seen, of which 2M can be used due to the defective chip.

Next tests were with: all switches off except for sw1=on; all off except for sw2=on; all off except for sw3=on; all off except for sw4=on. I'm not going to post images of those because they were identical to the test above with all off. That's all I had time for today.

Same for me, not much spare time at the moment but at least this hobby gives some relaxation and feeling of success...

You should start playing lotto when I see this. 1st try and you hit the correct combination 😎

Therfore you have 2.0 MB onboard? 384 kb selected as shadow-RAM and not available? With this, RAM-card starting adress should be 2.0MB. But this does not match the screenshots you posted before? Because on these there is nothing at 1.0 MB? Or was this a different Mainboard?

Never seen this BIOS, can you post here please?
80286 BIOS image collection

Predator99 wrote:

Same for me, not much spare time at the moment but at least this hobby gives some relaxation and feeling of success... […]
Show full quote

stamasd wrote:

Made a little progress today on the card. Wish I had more time/energy, but I only get 2h max after I come from work before I hav […]
Show full quote

Made a little progress today on the card. Wish I had more time/energy, but I only get 2h max after I come from work before I have to go to bed or I'll fall asleep on the couch.

Flipped all switches to "on", testing shows this:

All switches to "off", surprise! At next boot, 2M extended RAM found by the BIOS.

Testing shows this:

So 3.5M XMS seen, of which 2M can be used due to the defective chip.

Next tests were with: all switches off except for sw1=on; all off except for sw2=on; all off except for sw3=on; all off except for sw4=on. I'm not going to post images of those because they were identical to the test above with all off. That's all I had time for today.

Same for me, not much spare time at the moment but at least this hobby gives some relaxation and feeling of success...

You should start playing lotto when I see this. 1st try and you hit the correct combination 😎

Therfore you have 2.0 MB onboard? 384 kb selected as shadow-RAM and not available? With this, RAM-card starting adress should be 2.0MB. But this does not match the screenshots you posted before? Because on these there is nothing at 1.0 MB? Or was this a different Mainboard?

Never seen this BIOS, can you post here please?
80286 BIOS image collection

The motherboard I use has 1MB onboard. It's a SCAT chipset motherboard that I discussed recently Biostar MB-1212C (SIPP version) extended memory problem. The original BIOS of the motherboard is attached to that thread, but this is not what shows in the picture above. I found a compatible BIOS at http://chukaev.ru54.com/bios_en.htm file 2ctc001.zip (original computer was GW286CT) which I like more and this is what is in the picture.

The on-board 1M is configured as 640k conventional and 384k upper memory, which is unused while I'm testing the memory card (booting from a plain DOS floppy with no drivers, just the memory tester). Without the memory card, there is nothing above 1MB. So everything you see above 1MB comes from the card. None of the memory on the motherboard contributes to the extended memory.

That is why I said above that with the current switch settings, I can see 3.5MB from the card. The original switch settings only showed me 2.5MB from it. I don't know where the remaining memory from the card is (2.5MB missing with the original settings, 1.5MB missing with the latest tests out of a 5MB total). Hopefully further testing will elucidate that.

Ah OK, understood.

Most likely there is another DIP setting that selects the amount of available memory. And another one for the starting adress.

stamasd wrote:

No it's a full size 8bit ISA card. There are later 16bit versions but the original Rampage is 8bit.

Nice! How many inches is a full size 8-bit?