Ok .. summary :-
Refitted a DX4-100 (100NV8T)
result incomplete boot from floppy or hard drive unless cpu cache is disabled
disabled CPU cache boots either HD or floppy

This is not the same as previous..so refitted the second DX4-100

AMD DX4-100 (SV8B)
CPU cache enabled
Boots from HD
Will not boot from floppy - hangs the system

CPU Cache disabled
Boots from HD and Floppy

This is the same as before I tried the DX2-66

So it would seem there is a definate difference between the 2 x 486DX4-100 cpus
Neither allow to boot from floppy with cpu cache enabled
Both will boot either with cpu cache disbled (100NV8T)
DX4-100 (SV8B) will boot from either with cpu cache disabled
but only from HD with cpu cache enabled.
are these different chips ?
Something is obviously critical with this mobo
also attached a photo showing jumpers currently as requested.
Basically the same as the DX2-66 except for JP39/40 for core voltage change.
Surely not the 2 cpus at fault...this motherboard or chipset just doesn't like either unless cpu cache disabled...though one is more tollerant.

Sorry Snufkin..
didn't get to read you response until I'd finished posting my summary

Vipersan wrote on 2021-09-30, 16:15:

Ok .. summary :- Refitted a DX4-100 (100NV8T) result incomplete boot from floppy or hard drive unless cpu cache is disabled disa […]
Show full quote

Ok .. summary :-
Refitted a DX4-100 (100NV8T)
result incomplete boot from floppy or hard drive unless cpu cache is disabled
disabled CPU cache boots either HD or floppy

This is not the same as previous..so refitted the second DX4-100

AMD DX4-100 (SV8B)
CPU cache enabled
Boots from HD
Will not boot from floppy - hangs the system

CPU Cache disabled
Boots from HD and Floppy

This is the same as before I tried the DX2-66

So it would seem there is a definate difference between the 2 x 486DX4-100 cpus
Neither allow to boot from floppy with cpu cache enabled
Both will boot either with cpu cache disbled (100NV8T)
DX4-100 (SV8B) will boot from either with cpu cache disabled
but only from HD with cpu cache enabled.
are these different chips ?
Something is obviously critical with this mobo
also attached a photo showing jumpers currently as requested.
Basically the same as the DX2-66 except for JP39/40 for core voltage change.
Surely not the 2 cpus at fault...this motherboard or chipset just doesn't like either unless cpu cache disabled...though one is more tollerant.

Manual gives different jumper settings for the (N)V8T and (N)V8B (pages 14 & 15), try fiddling with those maybe?

[hmm, other motherboard have one jumper different between NV8B and SV8B. Getting to the point where you'll need to find out what those jumpers actually connect to so you can figure out what the settings should be for CPUs not listed in the manual]
[Oh, and I think the 8T is write-through internal cache, 8B is write-back]

Thanks Snufkin..
8B and 8T ...that would make sense.
Perhaps it is just a case of the manual being issued before certain CPU variations became available...thus not every permutation is covered.
It certainly doesn't make life easy though does it...

I'd give the settings on page 15 a whirl. It lists it as an AMD Enhanced 486 Plus, which elsewhere gives as an SV8B.

[also, for the NV8T that's now not booting from either HD or floppy, try under clocking it by removing the jumpers JP21&22, that should set the clock to 75MHz]

..Dont think I missed anything Snufkin...
Please check as my eyes are not good these days..
But currently appears to be booting from HD and floppy with the jumper config you suggested....and both external and internal cache enabled.

Sounds good. The only one I can spot different from the manual is JP34, which should have 1-2 jumpered (1-2 for 75 or 100MHz, 2-3 for 66 or 80MHz).

So you've got an Intel DX2 with write-through L1 cache and an AMD DX4 with write-back L1 cache, that both boot fine from floppy or hard disk.
You've also got an AMD DX4 with write-through cache that boots ok from the hard drive, but stops half way through booting from a floppy if the internal cache is enabled. only boots ok from HD or floppy if the internal cache is disabled.
[edit to fix my lack of reading comprehension...]

Is that right? Could be a faulty L1 cache on the AMD NV8T?

There was a rumor that there were some downbinned dx5s marked as NV8T despite having 16kB writeback.

snufkin wrote on 2021-09-30, 20:06:

Sounds good. The only one I can spot different from the manual is JP34, which should have 1-2 jumpered (1-2 for 75 or 100MHz, 2 […]
Show full quote

Sounds good. The only one I can spot different from the manual is JP34, which should have 1-2 jumpered (1-2 for 75 or 100MHz, 2-3 for 66 or 80MHz).

So you've got an Intel DX2 with write-through L1 cache and an AMD DX4 with write-back L1 cache, that both boot fine from floppy or hard disk.
You've also got an AMD DX4 with write-through cache that boots ok from the hard drive, but stops half way through booting from a floppy if the internal cache is enabled.

Is that right? Could be a faulty L1 cache on the AMD NV8T?

Not quite Snufkin..
from my post above..

"DX4-100 (100NV8T)
result incomplete boot from floppy or hard drive unless cpu cache is disabled
disabled CPU cache boots either HD or floppy."

As you say ...
Could well be faulty L1 cache.

TBH I'm just happy we have ALL made such progress..
I appreciate everyones contributions and help...and learned a lot in the process.
I may even return with fresh enthusiasm to the once working GA-486VF for some more digging.
As I said ..it might well be the 85C471 gone bad....as I think the cpu initialisation/reset is generated there...and the reset line at the CPU is permanently high at power up.
I would have expected Low High Low.....and not Low High >>>>
Must find and download the PDFs for the chipset if they are out there.
I do have 3 of each of the QFP chips on order from a guy called Adam in Poland.
(brand new old stock)apparently
So theoretically there is nothing on the mobo that is not replaceable...though those QFPs wont be easy.
I think the damage may have been done by that faulty from new Post card...which grounded data line 3 on the ISA bus.
Ive tried basic stuff like leaving it unpowered with the CMOS clear jumper in place for several days....just in case.
but it didn't help.

BitWrangler wrote on 2021-09-30, 20:32:

There was a rumor that there were some downbinned dx5s marked as NV8T despite having 16kB writeback.

Nothing would surprise me B W..
In my trade (TV engineer) I came across a lot of marked down 'seconds' when it came to 'silicon'
rgds
VS

Vipersan wrote on 2021-09-30, 21:40:

Not quite Snufkin.. from my post above.. […]
Show full quote

Not quite Snufkin..
from my post above..

"DX4-100 (100NV8T)
result incomplete boot from floppy or hard drive unless cpu cache is disabled
disabled CPU cache boots either HD or floppy."

One of these days I will finally learn to read. I've edited my post to not mislead people.

Don't know if it's possible to edit the thread title to mention the Soyo, given that's what it's ended up being about?

I may even return with fresh enthusiasm to the once working GA-486VF for some more digging.
As I said ..it might well be the 85C471 gone bad....as I think the cpu initialisation/reset is generated there...and the reset line at the CPU is permanently high at power up.

Have you checked if the power good signal from the PSU is at +5. I think the motherboard might stay in reset until it sees a good signal. Might be easiest to start tracing forward from there to figure out how the reset system works.

Checked the Power Good line from the psu 5.2v
Traced this to a diode D4 top edge of the mobo in front of the first 16 bit ISA slot.
measured again ..about 5v ...then of course the signal dives either underneath or inside the board layers....so attached a wire ...meter in continuity/beep mode...connected the wire to one side of the meter and dragged the other probe across every pin of every chip till I got a response..
..and yes as expected I'm back at the 85C471...pin 113 to be exact.
So I expect there is a gate in the chip that is connected to the internal reset circuit.
looking more and more like the 85C471 has died.
cest la vie.
anyhoo ..enough for tonight ..
my eyes are tired.
Scroll down to section 2.4 in the pdf

Hmmm there is reference to the power good line being gated with A20 line..
If this is an input ..pin 56 ...maybe it is generated elsewhere ?
maybe the hunt goes on...

Hmm...
This extract from the pdf might indicate the keyboard controller could possibly prevent reset and boot ?
"
2.4 Fast A20GATE And CPU Reset/Port 92h
In the original PC/AT design, the A20GATE and CPU Reset (RC) are controlled by the
8042 keyboard controller to switch the CPU between the real and protected address
modes and to warm boot the system. The operation of 8042 is slow. If address mode
switching happens frequently, execution speed will be affected. The SiS85C471 provides an 8042 emulation to generate the A20GATE and CPU reset
signals in hardware. This feature is software transparent. The SiS85C471 also provides a control port, I/O port hex 0092, to generate the
A20GATE and the CPU reset. The following shows the definitions of bit 1 and bit 0 of
the port 092h that are used for the generation of those two signals.
 Bit 1: This bit is internally OR'ed with the FAST_A20_GATE. Writing an "1" to this bit
will force A20M to go high active. Upon reset, this bit goes low.
 Bit 0: This bit can be set to "0" by a CPU reset or a IO write operation. A transition of this
bit from "0" to "1" will have the INIT signal go active. The minimum delay from the
transition of this to the INIT active is 2 to 6"

am I clutching at straws ?

Though that means that setting "Fast gate A20" in setup should hand that duty to the chipset rather than KBC, so if it is that problem with the KBC and that setting works, it confirms it, if it doesn't do anything different then you're probably looking for a different problem.

Had a good think about this today..
Apart from the supply lines and the reset ...why else would the cpu not run..
Obviously it would need a clock signal which I believe is 14mhz
Well the board has a 14mhz crystal ...and an MX8315 dedicated clock generator...
Just in case the cpu was loading it ..this was removed and a small component leg inserted into the ZIF at grid ref 4D
which I would have assumed would have a free running clock signal.
Nothing on the scope which I haven't given back yet.
I have the mobo on the bench atm hooked up to a bench psu to supply 5v
Looking at the pinout of the MX8315 ...Pin 8 is CK - Pin 4 is XCK - and Pin 5 is PCK
all of which are clock ouputs.
PCK = peripheral clock ...and should output 24 Mhz
XCK = Crystal Clock ou = 14.318 Mhz
and CK = CPU clock output..
..the only pin showing any signal is PCK ..
XCK and CK have nothing coming out..
CK is the signal which I assume makes its way to the CPU ZIF.
though I suspect not directly ...jumpers dividers logic or buffers etc
Point is there is no CK signal.
Now unless this chip is gated or controlled in some way there should be a free running clock signal.
Yes there are control pins on the chip presumably how clock speed is modified...but surely I should expect an output even without changes...I mean there IS something output on PCK.
Might be worth trying to find a replacement clock chip.
But before I go searching I'd like to run this by the knowledgeable folk in this thread.
rgds
VS

Huh, odd, the datasheet seems to say that XCK is a direct buffered copy of the crystal input. So if you're getting an output on PCK (which'll be generated from the 14.318 crystal input), then there should definitely be something on XCK. What frequency do you measure on PCK?

[Looking at the 85c471 datasheet there are at least 2 clock inputs: pin 1 should be the 14.318MHz input and pin 98 should be the CPU clock. The datasheet says the 14.318 is used for an internal timer, which might affect when it releases the reset]

[2nd edit... That said, given the damage to the board previously, it's probably worth rechecking the traces and connections you had to repair near the keyboard connector and controller]

Ok ..this is interesting ...
before condemning the chip ...which _could_ still be faulty..
I checked the supply pins ..9 and 14 digital power and analog power respectively..
Both are less than 1 volt..
tracing one of the pins (14) it surfaces at what looks like an inductor ...by it's designation L1
0.9v on one side and 5v on the other..
Pretty sure this is high resistance or O/C....as tested with the meter.
Is this and inductor though ? ...no part number visible....and what caused it to fail ..
Old age ...or loading.
..and what to replace it with as I have no idea what value it was ?
..anyone want to take a stab at it ?
I checked both sides of L1 with the meter to ground ...and nothing close to a dead short.

Vipersan wrote on 2021-10-01, 20:33:

Ok ..this is interesting ... before condemning the chip ...which _could_ still be faulty.. I checked the supply pins ..9 and 14 […]
Show full quote

Ok ..this is interesting ...
before condemning the chip ...which _could_ still be faulty..
I checked the supply pins ..9 and 14 digital power and analog power respectively..
Both are less than 1 volt..
tracing one of the pins (14) it surfaces at what looks like an inductor ...by it's designation L1
0.9v on one side and 5v on the other..
Pretty sure this is high resistance or O/C....as tested with the meter.
Is this and inductor though ? ...no part number visible....and what caused it to fail ..
Old age ...or loading.
I checked both sides of L1 with the meter to ground ...and nothing close to a dead short.

L would be the usual indicator for an inductor, so should read pretty close to 0 ohms. It's probably there to filter out the high frequency switching noise from the main digital 5V supply from the sensitive analog side of the PLL. Both the DVdd and AVdd should be +5 +/-0.5V. What are you using to power the board at the moment? Check the resistance from pin 9 to the PSU 5V.

Though I'm confused why you're getting an output on PCK if there's no power going to the chip.

snufkin wrote on 2021-10-01, 20:47:

Vipersan wrote on 2021-10-01, 20:33:

Ok ..this is interesting ... before condemning the chip ...which _could_ still be faulty.. I checked the supply pins ..9 and 14 […]
Show full quote

Ok ..this is interesting ...
before condemning the chip ...which _could_ still be faulty..
I checked the supply pins ..9 and 14 digital power and analog power respectively..
Both are less than 1 volt..
tracing one of the pins (14) it surfaces at what looks like an inductor ...by it's designation L1
0.9v on one side and 5v on the other..
Pretty sure this is high resistance or O/C....as tested with the meter.
Is this and inductor though ? ...no part number visible....and what caused it to fail ..
Old age ...or loading.
I checked both sides of L1 with the meter to ground ...and nothing close to a dead short.

L would be the usual indicator for an inductor, so should read pretty close to 0 ohms. It's probably there to filter out the high frequency switching noise from the main digital 5V supply from the sensitive analog side of the PLL. Both the DVdd and AVdd should be +5 +/-0.5V. What are you using to power the board at the moment? Check the resistance from pin 9 to the PSU 5V.

Though I'm confused why you're getting an output on PCK if there's no power going to the chip.

Thats the thing Snufkin..there is 0.9v getting to the chip...
Checked the oscillator again with the scope..
a lot lower than I think it should be and the frequency fluctuates a bit.
So I guess it is sort of running..though well below par.
currently using my bench power supply which has over current protection ..and a good constant voltage

I'm pretty sure replacing L1 will fix this ...assuming the chip is ok and didn't cause L1 to fail..
Now I just have to take a gamble on the inductor..
100uh - ish ?

Ok, if the frequency is fluctuating then it's definitely not working right. Could be doing some sort of stop/start thing. I had that once with a high resistance on a power supply. Voltage would be fine under no load, then when things started up the voltage would drop, things turn off, voltage recovers, things turn on, and so on.

L1 will probably be a lot less than that, I'd guess somewhere around 1 to 10uH (partly based on it's size). If it's doing what I think it's there for, then it's to remove the high frequency components of a digital switching edge (so multiples of the fundamental). For a quick test, I wouldn't be surprised if it worked just shorting it out. A lot of filters are there to make the board pass the various EMC regulations, so it may work without them, but be a bit less reliable and more likely to interfere with other equipment.