I continued with this board.
Installed the nwX287 instead of the empty Dallas, works.
Then I switch to 64MB RAM (2 times 32) and the Ark 1000 VL. Still with the DX2-66 and 256kB cache with 20ns.
But the Bios is really triggering me. Look at the chipset options:

That is all? No DRAM timing parameters? No L2 write back / write through switch?

By the way the picture shows the values after choosing "Optimal" values.

I am not 100% convinced that everything is stable, because the Windows 98SE installation got stuck. After a hard reset, the installation finished. I have to check now in detail.
The Windows 98 SE installation was done with a minimal hardware. So additional to the above listed: VLB I/O card, CF flash adapter with 16GB flash card, 3Com NIC with XTIDE Bios. IDE DVD drive.
So the "minimal" configuration is already a full configuration, only a sound card is missing. The NIC is needed, because I need the XTIDE ROM for the 16GB card.

EDIT: Of course also a 1.44MB floppy was part of the config, and a serial mouse (so I had to enable first a com port on the controller, normally I don't use any serial or parallel devices).

But this Bios is really weird.
I have to ask @pshipkov, if he can upload the Bios of his board, because he mentioned more options.

bios 2800 is much more prickly than 1800.
dont let go just yet.
use the "optimal" settings and try to work it out from there.
disable external cache.

pshipkov wrote on 2024-03-31, 19:50:

bios 2800 is much more prickly than 1800. dont let go just yet. use the "optimal" settings and try to work it out from there. di […]
Show full quote

bios 2800 is much more prickly than 1800.
dont let go just yet.
use the "optimal" settings and try to work it out from there.
disable external cache.

I gave up on the 2800 Bios. It is not about being more prickly (whatever that means, I have no clue haha).
It does NOT post, so it is stuck. No chance to enter the Bios and change something.

But I tried the one Bios from theretroweb version 2200 (Sprint OEM?). With that one the board starts and there are more options, too.
But I am not sure, if the Bios is really better than the original .... But there are more options for sure.

For now I continue with the 2200 Bios.
I tried to upgrade the cache, as expected by me, it was a pain in the ....
Finally I was successful, before I had one "bad cache do not enable the cache".

Next thing is Am5x86:
I tried those jumper settings

Maybe I got too greedy, because I tried to enable a setting with multiplier 2 (which is 4 for Am5x86) and WB.
So the board started very slow and got stuck .....
I put back the Intel DX2-66, similar behaviour.
Do you still know which settings you used?

But what is most frightening is the 3V, 5V switch. Hopefully I did not put 5 volts to the Am5x86 😀
These 3 jumpers open means 3volts, is it correct?

I can make more high res pictures of the jumpers.

EDIT:

I added JP1 and JP2 with green jumpers.
JP4 was also not set, I added it with a white jumper. JP4 is near the upper vesa local bus slot.
Some changes in the main jumper section.
With main jumper section I mean RN1 and RN2 (each 4 jumpers when closed, they are closed) JP41.
JP10, JP8, JP19, JP6, JP42, JP12, JP11.

Clock still at 33MHz (not changed, can be done later).

When you say 2800 does not work - what you mean exactly ?
No lights at all, or mem counter completes but strem hangs afterwards ?

pshipkov wrote on 2024-04-01, 08:00:

When you say 2800 does not work - what you mean exactly ?
No lights at all, or mem counter completes but strem hangs afterwards ?

With 16MB or 32MB module(s) it counts very quickly to 640k. More or less the first screen I see, is with memory 640k. Then it does not react anymore.
With a 4MB module: No lights.

EDIT: Probably I do one last test with the 2800 Bios: I re-install the empty Dallas. That should clear the settings. And then I try to run it with the empty Dallas chip, but I don't have high expectations, I guess it will hang at 640K, too.

I did the test: In order to come into a know good state again, I restored all the jumpers to the inital state, re-inserted the DX2-66 and removed all cards except graphic card. And removed the battery from the nx -thing.
All values are 100% lost.

I now have no lights with the 2800, no matter which RAM modules I use. => completely dead.
With the original Bios it still posts with 64MB of RAM (the maximum I used before, I could use 2 times 32 + 2 times 16, but don't have 4 free 32MB modules unfortunately, but maxing out is not the primary task yet).
So now I really give up with the 2800 Bios....
I now think I should continue with the original Bios, because I am not sure if the 2200 Bios is for the exact same hardware. On the other hand it is very unlikely that I will keep the board, when I have crappy performance, but one step after another .....

Since the Bios 2800 does not work on my board, my board must be somehow different to the one of @pshipkov. Bad.

Very strange. Visually the two motherboards look identical, but so be it.
Btw, i see some diffs in the jumpers. The photo in the 3+3 thread shows them for am5x86 cpu., not sure if the config in your pic is used with that cpu, or thats before you adjusted them. The issue can be something trivial like that.

pshipkov wrote on 2024-04-02, 06:17:

Very strange. Visually the two motherboards look identical, but so be it.
Btw, i see some diffs in the jumpers. The photo in the 3+3 thread shows them for am5x86 cpu., not sure if the config in your pic is used with that cpu, or thats before you adjusted them. The issue can be something trivial like that.

Thx for clarifying.
In your picture those 3 jumpers (near the voltage regulator) are open, that is most important, I don't want to fry a 3 volts cpu.
I tested the 3 jumpers open, the 5 volts DX2-66 still works. So I think the board does a voltage detection, closed seems to be 5 volts hard-coded, open: voltage detection.
With the same setting an Am5x86 works, too and is detected as DX4-100. So far so good.

But I wonder how you derived the jumper setting for Am5x86, because it is different to the one listed on the retroweb. Was L1 working in WB mode with your setting?

Also I checked the backside of the mainboard:

I am sure, I have an older version than you.

But the only visual difference on the top side is the socket for the Bios, my board has space for an 1MBit chip, yours not I think.

EDIT: I now have the original jumper setting (like in my first picture), except the 3 jumpers for the voltage.

What eeprom did you wrote the 2800 bios to ?
I would go with dip-32 chip like winbond w29ee011 instead the one you have on the photos.

pshipkov wrote on 2024-04-04, 07:00:

What eeprom did you wrote the 2800 bios to ?
I would go with dip-32 chip like winbond w29ee011 instead the one you have on the photos.

Ahhhhh. I did not see that there is a 1Megabit ROM in your board. Haha. Yes, of course I used a 512kbit Winbond W27E512, I don't have 1Mb EEPROMs at the moment.
That is a very good reason why it crashed in my board, I will buy some larger ones.
But on the other hand your dump was exactly 65536kiloBytes which fits perfectly in a W27E512, can you make please a full dump of your Bios?

pshipkov wrote on 2024-04-04, 07:00:

What eeprom did you wrote the 2800 bios to ?
I would go with dip-32 chip like winbond w29ee011 instead the one you have on the photos.

I ordered W29EE011 chips, should arrive in 2 weeks.
Can you please make a full dump of your 2800 Bios? Not urgent of course ...

Done.

pshipkov wrote on 2024-04-08, 02:47:

Done.

Thank you very much.

pshipkov wrote on 2024-04-08, 02:47:

Done.

Finally I got some bloody W29EE011-15 chips.
Of course I had to try the 2800 Bios immediately.
What shall I say? It works much better with a 1Mb chip. Thank you!

I am still using a Am5x86 with the jumpering of a normal intel DX/DX2/DX4. So for sure it will be L1 WT and multiplier is 3, so now 100MHz.
In the meantime I also installed 128MB RAM, could not resist.
Windows 98SE still runs, but only tells about 64MB RAM, in Bios the whole 128MB are counted. That has nothing to do with the Bios, was the same already with the 1800 version.

Next step will be now some more serious testing, I will apply the jumper settings from your picture and check if I get L1 WB and multiplier 4.

Cool.
This looks better indeed.

That 64Mb RAM only in Windows looks like stability/incompatibility problem.
I don't think i tested this board with more than 64Mb system memory, so don't know.

Also, do you get SpeedSys showing on your system ?
That can show you if L1 is in WB or WT mode.

pshipkov wrote on 2024-04-23, 20:23:

Cool. This looks better indeed. […]
Show full quote

Cool.
This looks better indeed.

That 64Mb RAM only in Windows looks like stability/incompatibility problem.
I don't think i tested this board with more than 64Mb system memory, so don't know.

Also, do you get SpeedSys showing on your system ?
That can show you if L1 is in WB or WT mode.

Short update:
I played a bit today with the board, it was not too successful. I followed the jumper settings from your picture.
So the 4 times multiplier works, the Bios always shows 100MHz (no matter if it is really 100MHz, 133MHz or 160MHz)
Board is now running at 160MHz.

First of all, everything was unstable, it improved a lot when I removed RAM, so now only 64MB (2 times 32 instead of 4 times 32MB).
Speedsys shows extremely low values for memory and cache write. It is weird. Even if all timing parameters are on lowest wait states = fastest settings @ 40MHz. Super weird.
But speedsys clearly shows that the 1MB L2 cache is working.

But 100% the L1 is working in WT mode 🙁
So I guess the jumpers in your picture are not for L1 WB mode.
In your picture JP1 is open, JP2 is closed, JP4 is open (JP1 and JP2 are next to the keyboard controller, JP4 is between two ISA slots "above the highest VLB slot" so to say)
According to the 2 manuals on retroweb, it should be JP1 closed, JP2 closed, JP4 closed.
But now there is a BUT ......
JP4 is a 3pin block. So the manual must be wrong / for another revision. I see you have also the older revision of this board (without voltage regulator) and there is JP4 only 2 pin (so open or closed). So the manuals are for the older revision as it seems. I have a long try and error session ahead.

====
EDIT:
I apologize, obviously I need better glasses:
On my board JP4 is only 2pin, so my board is similar to your older revision board.
But your newer board with the voltage regulator has the 3 pin JP4.
====

I will install more tools soon. At the moment, I have more or less a blank CF card with DOS7.1 boot, speedsys and sysinfo.
While playing around the Windows 98SE installation got corrupted, so I formated the CF card again and started over.

I was not able to get SpeedSys working on the motherboard here.
High chance L1 cache is in WT mode, which is not necessarily a bad thing. Many systems like it better that way.

My notes don't talk about manual discrepancy, but there is a chance i simply didn't log it. Cannot remember details anymore.

How bad is the performance you are seeing ?
Looking at the numbers i shared, they are average or better when compared to the rest of the other systems.

pshipkov wrote on 2024-05-01, 00:24:

I was not able to get SpeedSys working on the motherboard here. High chance L1 cache is in WT mode, which is not necessarily a b […]
Show full quote

I was not able to get SpeedSys working on the motherboard here.
High chance L1 cache is in WT mode, which is not necessarily a bad thing. Many systems like it better that way.

My notes don't talk about manual discrepancy, but there is a chance i simply didn't log it. Cannot remember details anymore.

How bad is the performance you are seeing ?
Looking at the numbers i shared, they are average or better when compared to the rest of the other systems.

I continued.
First I did another Vogons search about Symphony Wagner, which turned out to be a very good idea.
I found a modified Bios by user Oerg866 for the Abit AB-PW4
https://theretroweb.com/motherboards/s/abit-ab-pw4#bios
OK, as a first thing I installed the Bios (used the 1Mb variant, since I have now enough Winbond W29EE011 chips 🤣) and the computer still booted. The Bios also correctly identifies the CPU as a Am5x86 and also detects the 160MHz, so somehow nicer than the Ami BIOS.

But what about L1 WB?
It was easier than expected. I just closed JP1 and JP4 (JP4 seems connect to the CPU WB pin) => aaaaaannnnnndddd L1 WB is working
I am really excited.
I also set all parameters to the best possible values, except memory faster (best is fastest).
Later I noticed that the system was crashing when trying to boot from floppy.

I played with the jumpers. I followed
https://theretroweb.com/motherboard/manual/vs … 69536830598.pdf
and used as a reference P24D (WB).
And it seems to be stable now with the same fast settings, it can handle correctly the floppy disk. YES!
For little money I can provide everyone who is interested my jumper settings. Kidding of course.

Next step will be the installation of Windows 98SE, my favourite stability test and afterwards I will run some benchmarks.
I guess the ASUS SV2GX4 will still be better in Doom and Wolf3D, the speedsys graphics throughput shows only 35000 (it is 38000 with the Asus).

The L2 throughput is acceptable, but the main memory is a bit sub-par