NF4UK8AA - Is this Lazarus, just sleeping?

[momaka]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

If the board worked before you messed with the VRM... or at least POSTed every once in a while... then perhaps you did something to the VRM area that's causing the low V_core now.
My suggestion is to first check the voltage feedback network going to the VRM controller. Get a datasheet for the VRM controller and see which/where is the Feedback pin. Then run back from there and see if the resistors that connect this pin to V_core and ground have proper values. The VRM controller is also likely receiving I2C data either from CPU or BIOS for the voltage control. So bad BIOS / CMOS data could lead to wrong CPU voltage. Same with the feedback resistors on the VRM controller.

Also, while I do usually suggest to remove and reinstall CPU(s), I also suggest not to do it too many times. CPU sockets aren't meant for too many connect/disconnect cycles. IIRC, it was in a tech note for Intel LGA socket, and said that the number of guaranteed connect/disconnect cycles was only something like 6 or 7 (i.e. only 6-7 CPU changes.) AMD sockets might be able to take more... but who knows.

So while a good idea to try different CPUs, also a good idea not to overdo it. If I have 2 known good CPUs and one doesn't work, I will try the other one. If that doesn't work too, then I don't bother swapping CPUs anymore.

It's also very rare and hard for a mobo to kill a CPU, especially the older CPUs. I've had a particular LGA775 CPU for example, survive 5 direct high-side shorted MOSFETs.

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

I tried a different CPU, an Athon 3000, and this one went to POST code 44 (again) with Vcore at 1.01V
Thought that i must have destroyed the 3200 CPU that i was using, after reinserting quite a few times previously,
so i mucked around trying to bend the pins, with a meter probe, that looked a bit off (wound up bending one right over in the process), until i thought that using a credit card would be a much better/safer way to align them....
Put it back in, but still just horizontal dashes on the analyser card
Tried a 3500 CPU but still horizontal dashes again
Tried reseating these CPU's many times - same result
Thought that it was a bit odd that two CPU's would be buggered, or that the socket would be so fussy (all of the contacts look good when moving the CPU lever under the microscope), so tried all three CPU's in another board, and that board POSTed no problem

Could it be that there is power issue that is preventing the board from starting up, as there is enough voltage at 1.01V to just start the POST process with a 3000 CPU, but not the 3200 or the 3500 CPUs?
Vcore on the other board is 1.4V
So i suspect that that is what's halting the boot process, because, with the 3000 CPU, the board is obviously POSTing, but it's getting stuck at a certain point in the process because of low Vcore.
It really looks like 1.01V is not enough voltage to to allow the POST to fully complete.
The code 44 (Phoenix) says "Initialize BIOS interrupts.(Beep)=2-1-2-1. Verify video configuration." so i'm thinking that whatever that is, something changes at that point, and then there's not power to continue with the boot sequence. (just thinking out loud)

As previously mentioned, for some reason the board was working no problems, after i removed and put back MOSFETS; then as soon as i put that video card in, and restarted, something changed:
Maybe the video card did a bad thing to the motherboard, or maybe that was just coincidence...
or maybe there is something dicky going on with the VRM on the board (it's not the PSU - i tried another good one just in case it was) because a couple of times there was no response when i pushed the power button, and another time i could power off the board simply by tapping the corner of the board.
Anyway, i think that the issue at hand is the fact that Vcore at 1.01V is just enough to perform some of the POST with an Athlon 3000, but not enough to complete it.
Any thoughts on how to approach troubleshooting that?

[momaka]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

In the CPU is constantly resetting due to something being buggered with the chipset, that could be the reason why you see the low voltage on V_core.

Given how it's been slowly going downhill, it might really be time to flow the nVidia chipset... unless you have a programmer and can do what Dan81 did and flash a different ROM on there (after backing up the original, of course.)

BTW, have you tried removing and installing the CPU a few times? I've had issues before where a system slowly became more and more crash-prone until it stopped booting. Turns out the CPU pins had the slightest bit of oxidation where I touched them on the side, and that caused the board not to boot. Some cleaning + a reinstall of the CPU in the socket, and it was back up and running.

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

Yeah, i used a hot air station.
Thanks for providing a better technique/alternative
That "bulged driver chip" now appears to be a good chip with a coating of flux on top of it (i realized that after measuring all of the pins to ground, and they all were the same)
So, i cleaned it up, plugged the two ATX connectors back in, pushed the power button, and now get horizontal dashes on the POST analyser card. (which is what happens if i forget to plug in the 4-pin CPU ATX connector)
So then i check the voltage of Vcore, and it's 1.01V, which is significantly less than what it was before. The drain pin is 11.94V.
I didn't do anything, except wipe the flux off of the top of the chip. Now i turn it back on, and Vcore has dropped to 1.01V

[Dan81]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

Got a NF4UK8AA as well, mine boiled down to the weirdest thing ever:

-won't boot with its own BIOS freshly flashed on a blank chip.
-will successfully POST with a DFI NF4 SLi INFINITY BIOS chip.

Anyone have any ideas what the actual ever living bejeezus is going on with this mobo? I've tried flashing the BIOS over at least two mobos (an ASUS P5PE-VM and the above-mentioned DFI) and the same result happens.

To recap what I tried:

- two BIOS chips - one is the original SST 49FL004B chip, the other a PMC/PMIC Pm49FL004T-33JCE - both won't POST on the Winfast NF4UK8AA, but will work if I put them on the DFI mobo
- various RAM sticks - even as far as using ADATA Vitesta DDR500 sticks (as they came with the Winfast mobo, along with a pair of 2x1GB Kingmax sticks) to no avail - none will work with the Winfast BIOS, but will work fine with the DFI BIOS.

[momaka]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

While looking over the area again, i noticed a bulged driver
Now, i don't know it it was like this when i got the board, or if it was a consequence of spending too long heating the MOSFETS when removing and replacing them (probably the latter)
So i'll order another one and another MOSFET and hopefully replace them without destroying any other components.

What soldering tools are you using to replace the MOSFETs and drivers? Hot air?

The VRM area is pretty hard to heat, so it's definitely possible the driver IC could have been heat-damaged from the hot air.

I myself prefer using just two soldering irons to remove such parts and save the hot air only for TSSOP or finer pitch ICs. In case of MOSFETs, I use my 75W station to heat the tab, and then a 30W iron alternating between the two legs. Once all heated, I slide the MOSFET away. Same with 8-pin ICs, like that driver. Takes only a few seconds to remove a MOSFET that way and there is minimal heat leakage to any other components. Before doing any of that of course, I first add fresh solder to the tab and legs of the MOSFET. And in the case of 8-pin SO/SOIC, I add solder to bridge all of the pins on each side.

Could a damaged driver be a reason why the board is faulting?

Unlikely, but I won't say no.
If the driver IC is bad in such a way that it's not shorted and also not doing anything else in the circuit (i.e. not driving the MOSFETs at all), then that phase in the VRM won't be contributing its share of providing power to the CPU, and so the other two phases will have to work harder. And since phases are usually staggered on multi-phase motherboards, one phase missing can cause the ripple current and noise to increase significantly. If that's the case, then that could well be why the motherboard is crashing. But more often than not, a bad driver IC typically causes one of the MOSFETs to get stuck, and then the whole VRM tanks and doesn't work and/or PSU goes into protection from overload due to "stuck" MOSFET.

Also when i got the board i noticed that the NB driver fan was noisy/erratic, and had being replaced, and that the third wire was disconnected (which i reconnected, and the BIOS health screen showed that it was monitoring it's RPMs, and also spins lot quieter now) - So the NB may have being running without a fan for some time.
The fan makes a big difference to the NB temperature - from fairly hot when it's not connected, to comfortable when it is.

Yeah, my Biostar NF4U AM2G was in a similar condition - NB fan was very noisy and possibly may have been stuck for a while before the seller / previous owner cleaned it to get it running. It's still very noisy (BB's are shot and make "sand grinding" noises every once in a while.)

I also had an ECS MCP61PM-AM recently die on me. But it's been on a deathbed since 2018, so not like it was unexpected. Would die, then I'd get mad at it and punch the chipset heatsink (literally! ), and back to working for a few more months. It has a GeForce 6100/6150 chipset. If you ever run into a board with one of those, AVOID IT by all means. I didn't believe it, but these things are definitely all doomed to fail. In the case of the above board, the LAN started going out back in 2014 (and probably why I found that PC in a dumpster.) Gotta take it out of the PC one day and do a reflow. Maybe that will buy it a few more years. CPU's a X2 6000+, which holds up surprisingly well even today.

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

Ok, thanks
Yeah, i did all of that
Blasting it is a last resort - thanks for the tips though, in case it comes to that.
While looking over the area again, i noticed a bulged driver
Now, i don't know it it was like this when i got the board, or if it was a consequence of spending too long heating the MOSFETs when removing and replacing them (i only had the hot air gun on the MOSFETs)
So i'll order another one and another MOSFET, and hopefully replace them without destroying any other components.
Could a damaged VRM driver be a reason why the board is faulting?
Also when i got the board i noticed that the NB driver fan was noisy/erratic, and had being replaced, and that the third wire was disconnected (which i reconnected, and the BIOS health screen showed that it was monitoring it's RPMs, and also spins lot quieter now) - So the NB may have being running without a fan for some time.
The fan makes a big difference to the NB temperature - from being fairly hot when it's not connected, to being luke warm when it is.

[momaka]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

It's an nVidia chipset board... and a socket 939 at that... so I probably need not say more. (But I always do. )

Yes, the nForce 4 is affected or at least partially-affected by the bumpgate issue. In fact, I have one Biostar board that's acting similar to yours - sometimes it boots, other times no. It still boots more often than it doesn't. But I'm sure it will get worse.

The stock heatsinks on socket 939 also had a tendency to warp the board very badly, so the issue could be bad BGA under the CPU socket too.

That said, I'm willing to bet my money on a bad/dying nForce chipset.
Before you go nuts with the heatgun, though... try booting the board with the CPU heatsink on the CPU, but with the retention spring loose. Any difference?

Also, is the thermal paste on the nForce chipset's heatsink good? Tried replacing it? If not, try that too.

And if it still misbehaves, give it heat from hell... er, I mean the heat gun.
You may not need to reflow it to full RoHS solder melting temperatures to get good results. 150-180C on the chip could be enough. But if not, do it again and crank it up more.

BTW, if a board is turning On and not shutting down instantly, then there's no need to mess with any of the MOSFETs. A shorted MOSFET will either make the PSU shut down right away or it would be overheating and causing serious discoloration/burn marks on the board. So if you're dealing with an intermittent issue, the problem will almost never be a bad MOSFET.

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

I used about a tablespoon of dishwashing liquid in about 7 litres of hot water with a soft toothbrush, (photo attached),
then gave it a rinse in another container of hot water.
Prior to removing the MOSFETS, there was one time where i could power down the board simply by gently tapping it with my finger (yes, the ATX connectors were fully plugged in) - tried it a few times, and could repeat it; but it didn't happen again after that - i did try flexing the board to see if that made any difference when starting, but no.

[DynaxSC]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

What agent did you use to wash the board?

[socketa]

Re: NF4UK8AA - Is this Lazarus, just sleeping?

Now i can't power down the board with the power button.
It's got worse without me doing anything.
Still get error 44
When i hold down the power button the reset LED on the analyser card lights up (after several seconds, like normal), but stays lit when i release the button.
The only way to power down the board is by turning off the PSU.
What could this indicate?

It's like a chip is recognizing that i held down the power button, but it's not passing it on
Maybe is a bad nVidia NF4 chip? (i seem to recall that they are part of nVidia bumpgate)
Odd how it went from sometimes reaching BIOS config and freezing, to working perfectly, to not even getting to the point where is detects missing RAM/RAM in wrong slot and not being able to be be powered off
It's like some stories that you hear when folk are on their death bed, where for a brief period they sit up and become very conscious, then a few minutes later they're dead (so to speak)

[socketa]

Good CPU, PSU, RAM, keyboard and graphics card
Phoenix BIOS
POST card tester error 44 - "Initialize BIOS interrupts. Verify video configuration."
The board would occasionally fully POST, but then freeze (with error code 44) after a short time when in BIOS config.
The VRM caps all test good (MCZ and MBZ)
There's normal-looking voltages on all of the big MOSFETS that are on the board.
The VRM MOSFETS are not shorted.
Anyway, i mistakenly thought that there was too low resistance on the high side MOSFETS (12Kohm) but after checking out a similar board schematic saw that that was normal.
But by that time i had removed the three high side MOSFETS, three electrolytics, and the six little SMD decoupling caps that were in parallel.
So i put the MOSFETs, electrolytics, and three of the decoupling caps (lost the other three), and powered it up.
To my surprise, it booted normally, no problem, and didn't freeze when in BIOS, i verified this several times, and all appeared to be working well, and all of the rail voltages that were listed in "health" section looked great, "yay!"...
That was, until i swapped the video card (which i had previously tried, and which didn't work) and the board went back to giving a 44 error (even with the good graphics card).
I washed the board, and sun dried it with the help of compressed air, but sill got error 44, (along with no keyboard led flash, or speaker beep).
i can hear the speaker click when i push the power switch, but the POST sequence probably doesn't get to the point where it will beep, like it did when i removed the RAM or put the RAM in the wrong slot.
The four rail voltage LEDs, and the clock LED, light up on the analyser card
Any suggestions that might help resurrect this board again?