NF4UK8AA - Is this Lazarus, just sleeping?

[socketa]

Ok, so i lifted the three legs of the PWM controller, and there was no voltage on it's OUT pins that are going to the three drivers, even when shorting the VID pins to ground.
Then i put the overclock chip back in, and there is the same voltage on each of the PWM controller's OUT pins (50-250mA - it changes each time that i power it up)
Shorting the VID pins to ground makes no difference.

[socketa]

Removed the overclock chip, so it's VIDOUT pins are definitely disconnected now
But lower mosfets still don't turn on when shorting VIN pins to ground

[прямо]

I'm not sure if # symbol means the signal is active low, but you can probably verify it by toggling it between active low and active high and see what happens to the chip.

And yes, it does seem the chip receive VCORE setting directly from the CPU and then passes it onto the PWM controller chip. In this middle man position, the chip can also overrides what the CPU has requested.

I think you need to temporarily disconnect the chip VIDOUT pins from the PWM controller to see if the controller will then respond to your manual VID settings.

[socketa]

Yes, i disconnected the PSU and moved the CMOS reset jumper to reset position and then back again
I also flashed the BIOS because it's checksum had changed
So even though i ground the VIN pins that go to the controller which tell it what voltage to regulate at, there seems to be something disabling the chip from outputting signals to the drivers,
Like you say, maybe the overclock chip is telling the controller chip to not output, irregardless of whether i try to force VCORE by grounding the VID pins
I found a schematic of a board with that overclock chip
https://elektrotanya.com/gigabyte_ga.../download.html
and see that the VIN pins of the overclock chip go to the CPU (and on the other side of the chip are VIDOUT pins which go to the controller's VID pins)

On the overclock chip there is a watchdog timeout reset # pin which is low (does the # indicate that it's active low?),
and the CPU change detect pin still correctly shows when a CPU is present or absent

And on the VIDIN pins of the overclock chip, from VIDIN0 to VIDIN5, which i think are connected to the CPU VID pins, there is 101001, which could possibly mean that part of the CPU is asking for VCORE to regulate at 00101 (1.425V), or 01001 (1.375V),
1.375V is between what CPUworld says should be the CPU voltage (1.35 - 1.40V)

I'm guessing that the CPU needs VCORE before the BIOS will POST

I found a video of how to remove the cover from a AM4 socket, so i might try that if i have to
https://www.youtube.com/watch?v=JTWgNGoylo8

[прямо]

Have you tried CMOS reset? it seem like the jumperless overclock IC gets a configuration from somewhere (from a previously saved BIOS setting most likely) and it's overriding what the CPU is actually telling the PWM controller.

[socketa]

Maybe due to this?
Top right and bottom right

Bent the CPU pins a bit, when taking it in and out, so wiggled it into the socket
Bad move!
If the CPU doesn't fall in by itself, do not apply any pressure whatsoever
Tried to open a socket 754 socket once, but destroyed the top of of in the process - they break really easily, and you'd be very lucky to remove it without breaking the plastic,
and this looks like the same deal
The damaged hole is MEMDATA[105], but that's to do with the RAM, so it's possibly not the reason why there is no VCORE or any POST
Checked the holes for VID0 to VID5, and they all look good

[socketa]

The VID pins of the VRM controller chip are direclty connected to the VIDOUT pins of the jumperless overclock chip
It seems like the job of the overclock chip is to override the voltage that the controller chip is putting out.
Shorting pins VID 3&4 of controller chip to ground should give 00111 (1.35V)
But this new PWM controller chip is also not sending out any PWM signal voltage to the drivers.
The three driver chips each have 12V power at VIN, and voltage on drains, and the controller chip has PGood voltage,
and the jumperless overclock chip correctly detects when a CPU is installed
I get the feeling that this is something simple
Maybe the original controller chip was going faulty, but i suspect that when i accidentally blew up the mosfet twice, that that may have created some extra damage, because before that i could turn off the board with the power switch, but can't now, and was also getting some codes on the PCI slot analyzer card, but now there is nothing.

[socketa]

Well, it looks like that VRM controller chip was possibly OK
because i replaced it, and there's still no voltage on it's PWM out pins, even when grounding the VID pins (which are all high)
so no Vcore
and the board won't shut down using the power button

[socketa]

Yep, have ordered some
Will update when the chip has being replaced

[прямо]

I suppose the controller output is pretty much toast because of the short to ground accident.

I guess the next logical move is to replace the chip.

[socketa]

I was sure that the high and low side mosfets were shorted to ground
Now, after checking again, it's only the high side mosfets drain and source shorted together, and not to ground
I found the faulty mosfet (was the same position that the previous one went short), and replaced it

Now, with the CPU removed, the board starts, but this time no VCORE, and the VID begins at 11111 (off), and then, over the course of about ten seconds, VID0 falls to low to make 11110, but still nothing on VCORE
When i push the power button with the CPU installed, VID is 11111 (off), but still no VCORE
Unlike last time, grounding VID4 (or any of the other pins), while powering up, with, or with or without the CPU installed, fails to bring up any VCORE voltage.

Another odd thing is that all of the pull up resistors are now measuring about the same in-circuit (3.4 - 3.5kohm); whereas before, across VID3's resistor, there was 1.3Kohm

So now it seems to be behaving as you expected
or maybe that's what happened if the controller clapped out (no working controller to pull the VID lines down)

Regardless of whether i ground any VID pins or not, there's now no voltage on any of the PWM OUT pins.

[прямо]

Oh damn! The controller chip is probably dead now.

So aside from the CPU, something else is able to set VID. Perhaps that jumperless overclock IC is the culprit. I was pretty sure that without a CPU installed, VID would be all pulled to low, telling the PWM chip to turn itself OFF, but instead it is set to always put out 1.0V

which is really weird, because if 1.0V VCORE is like the default start up setting, but the CPU obviously need more than that to start, then it's no brainer that the mainboard doesn't work and can't POST.

[socketa]

Did that (sorry, it was VID4, not VID3, that i pulled low to get 1.375V)
There's 12V on the three drivers supply pins
There's 8.57V - 8.64V on the low side gates,
but there's no voltage on the high side gates,
and of course, there's no VCORE

The controller is putting out a signal 10110 (1V)
and when grounding VID4, there is still no VCORE, since there's no voltage on the high side gates

There is no voltage on the PWM OUT pins

Then i accidentally shorted OUT1 to AGND, and it shut down, now the board wont start
It tries to start (nForce4 fan twitches), and shuts down nearly instantly because the high side mosfet's drain and source are shorted to each other

[прямо]

Try pulling VID3 low again, but without a CPU this time, and see if you will actually get 1.37V or not, because I don't think the phase controller chip is faulty, after all it was able to put out the correct voltage for 0.975V, 1.0V and your earliest attempt at 1.2V

[socketa]

Yeah, maybe the 3000 CPU was just about dead after the upper vrm shorted,
and it had enough life in it to to POST, but then fully died

This board also has an IT8206R "jumper free overclock controller" that's located next to the VRM controller, and VID 0-4 also go to it

There's five 4.7k pull up resistors
with the PSU and CPU removed, they all measure 3.4kohms in-circuit,
except for VID3's resistor, which is 1.3K

With the CPU removed, there is 1.0v VCORE, which corresponds to what's on VID 0-4 10110
With the CPU installed, there is 0.975 VCORE which corresponds to what's on VID 0-4 10111

OK, so i felt confident to start it up while shorting VID3 to ground
which put 1.370V (probably 1.375V) on the VCORE, and code 45 on the pci card,
but still couldn't swtich it off with the power button

And it obviously didn't like it, becuase now the VRM is not being switched (no voltage on the mosfet gates), so nothing on VCORE
i checked the CPU - fortunately, it's VCORE isn't shorted to it's ground like the other one.

Tried resetting the bios, but no change

Was reading that resistors don't usually fail short, so the pull up resistors are probably OK
Am thinking that that out-of-line reading across VID3 pull up resistor is possibly an indication of a faulty controller chip

[прямо]

It doesn't make sense to me that when it's being forced to pull low, it will end up overheating the mosfets and killing the CPU, because these pins are supposed to change states real-time, in respond to whatever voltage the CPU requested, such as when a user is under/over-volting the CPU on the fly.

Most likely what killed the CPU was the high side mosfets gate were opened and 12V passed through.

Other than pull up resistors, I don't see any additional circuitry. Pins are direct connection from the CPU to the phase controller and shared with super IO for monitoring.

Maybe try turning on the mainboard without a CPU and see if the VID pins change state accordingly (all pulled low = OFF)

The alternative of injecting VCORE from your bench supply poses a much higher risk factor because it involves exposing the components to an excessive heat from your hot air blower gun, since you will have to remove the mosfets and then force the phase controller to stop working (either by pulling VID pins to low or remove the chip)

[socketa]

Thanks for that
Re-flashing made no difference
Am assuming that i should be looking at the K-8 output voltage chart that's on that datasheet, since the CPU is AMD socket 939
On VID pins 0 to 4 there is 10110 (3V = high),
and that equates to 1.00V

Meter shows about 1.0V
Shorted a few of those pins to ground and had it up to 1.2V and the PCI card moved on to a different code and stopped there
But now i noticed that the mosfets were getting hot, Vcore 800mA, and now the CPU is shorted
Maybe i briefly put it up to 1.55V, and killed the 3000 CPU since it's maximum specified voltage is 1.35/1.4V
If that's true, i wasn't aware that an extra 0.15V could kill a CPU
I've now put a 3200 CPU into it, but of course there's not enough voltage to fully power the CPU 1.35/1.4V
It's still 10110 (1.00V)

00110 would be 1.4V

VID4 is 2.77V when the other two highs are 3.0V
It's a bit suspicious
Maybe there is a faulty component in the VID4 circuit that is stopping it from being pulled low,
because 00110 is only one digit different from 10110

Hmmm
If the VID signals are generated from the CPU, then it seems that the CPU is asking for 1.4V, but it's not getting it, so it's not POSTing.
At the moment i'm not keen enough to ground VID4 in case doing so kills this CPU also

[прямо]

According to the controller datasheet, VCORE is set by VID0...VID4 and on the schematic, these pins are directly connected to the CPU socket.

Perhaps you can compare their states (high or low) with this table,



and instead of injecting 1.5V directly into VCORE, I'd force these pins to match what the table says for 1.5V

[socketa]

I flashed the bios, but still get code 44 on the PCI slot card
the only noticeable difference is that the board now powers down when the power button is pressed,
1.02V VCORE sounds too low to me

...A few hours later , now a hard short has developed across VRM mosfets
...Found, and replaced, the shorted mosfet, and it shorted again as soon as power was applied
...Replaced that one, and now get code D0, and now can't shut down with the power button - VCORE still 1.02V
Am wondering that since i replaced the mosfet with a different brand, maybe i should replace all of the other paralleled mosfets, so as to make them all exactly the same?
Maybe that's what causing the low voltage?

... replaced the other mosfets so that they all match, but that made no difference
...maybe the bios got corrupted when the VRM shorted
...will try re-flashing it

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

Without the CPU there is 46 ohms from controller feedback pin to Vcc.
With the CPU there is a hard short from controller feedback to Vcc
Couldn't find a schematic for this board, but i found one for a similar board
Attached, is the schematic (NF3UK8MA - same controller, drivers, and MOSFETS)
Then, i calculated (from the schematic) the equivalent resistance of the series/parallel resistors that are between feedback and Vcc, and got 46 ohms
So, unfortunately, that appears to be OK