Oh NO!
Radeon 9200 is AGP 8x -> 0.8v signaling
Asus P3B-F is AGP 2x -> 3.3v signaling

You most likely fried a voltage regulator or the chipset or both.
Sorry to say, that usually that kills the chipset.
You MIGHT get lucky and it's only a voltage regulator so it is worth checking.
.
W83977EF isn't just an ISA chip. It's a Super I/O chip and those are tied into just about everything.
Anything it's tied into could be trying to work at the wrong voltage.
.

Radeon 9200 is AGP 8x -> 0.8v signaling

No. 9200 has universal AGP.

As it's been said a 9200 can run 3.3v but that doesn't mean that there was nothing else wrong with the card.

The Serpent Rider wrote:

Radeon 9200 is AGP 8x -> 0.8v signaling

No. 9200 has universal AGP.

There has always been a lot of misinformation about AGP card-slot compatibility.
Since I don't have one here I can only go by what the spec references online say.
I looked at many and they all said 8x and none said it supports 1x/2x.
Just some examples.
https://en.wikipedia.org/wiki/Radeon_R200_series
https://www.cnet.com/products/ati-radeon-9200/specs/
https://en.wikipedia.org/wiki/Radeon_R200_series#Radeon_9200
http://www.gpureview.com/radeon-9200-card-69.html
https://videocardz.net/ati-radeon-9200-se/
http://www.gpureview.com/radeon-9200-card-69.html
https://www.amazon.com/ATI-Technologies-Radeo … s/dp/B00009KTMA
I did see one ad for a cheap Chinese card that said 4x/8x. That still only gets you to 1.5v, not 3.3v.

In reality 8x only AGP cards don't exist unless something new happened in the last ~10 years or so.
Any that claimed to be 8x only turned out to be 4x/8x and supported 0.8v or 1.5v.

Radeon 9200 refers to the chip on the card, not the whole card.
Mfr 1 may chose to make it as a 4x only card while mfr 2 makes it in 8x only.
I don't see why they would, but they could.
Same with 1x/2x. The chip may support it but the specific card doesn't.

I also don't see why all the ads and specs only say 8x when they are 4x/8x.
From a marketing perspective that's not too bright.
Same-same if they support 1x/2x. They would sell more if they actually said so.

Just to add confusion...
I went back and looked at several of the actual cards and most were keyed for 3.3v and 1.5v.
That's not definitive either because AGP cards that were keyed incorrectly weren't that uncommon.

I have no way to determine if many of the cards were keyed wrong or all the ads and specs online are wrong.

I do however have an Asus P3B-F right here and it's definitely keyed for 3.3v only.
That a card that is probably 4x/8x was plugged in and the board blew up suggests that (at least the card in question) doesn't support 3.3v.

[edit] And yes, as said, it could have simply been a bad video card.

Radeon 9200 was Radeon 9000 refresh with added 8x support, similar to GF4MX-8x. It will work with any AGP slot.

The Serpent Rider wrote:

Radeon 9200 was Radeon 9000 refresh with added 8x support, similar to GF4MX-8x. It will work with any AGP slot.

So they all work fine in MAC AGP slots? [teasing]

I suspect you are right (about the chip itself) but I haven't seen any information that backs you up.

I remember from back in the day there were lots of problems with card manufacturers implementing AGP incorrectly on their cards which fried 1x/2x AGP chipsets even when the chip supposedly supported it.
What I don't remember is which cards and chips were problems.

During that time the only ATI AGP I used came in pre-fab systems or integrated into server boards.
You definitely know more about ATI than I do.

Having literally fixed thousands, I'm pretty sure I know more about blown motherboards.
This board is blown.
Knowing how it happened only helps to figure where to look to see if it's repairable or not.
.

PCBONEZ wrote:

Having literally fixed thousands, I'm pretty sure I know more about blown motherboards.
This board is blown.

It doesn't take the knowledge of having fixed thousands of mainboards to know that the board is blown, because that information was there from the beginning.

Replacing the W83977EF and the VRMs is the minimum required to get the board running again. I wouldn't even care however and just replace the board. It's just an average Slot 1 ATX board without anything special about it.

derSammler wrote:

PCBONEZ wrote:

Having literally fixed thousands, I'm pretty sure I know more about blown motherboards.
This board is blown.

It doesn't take the knowledge of having fixed thousands of mainboards to know that the board is blown, because that information was there from the beginning.

DUH! What a brainiac comment.
The OP specifically asked....

Paar wrote:

... I assume that something got fried but the board itself looks undamaged. Do any of you have experiance with this? I would be sad if I damaged the board to the point I could throw it to a trash bin.

I'm simply trying to answer his/her question.

derSammler wrote:

Replacing the W83977EF and the VRMs is the minimum required to get the board running again. I wouldn't even care however and just replace the board. It's just an average Slot 1 ATX board without anything special about it.

That's nothing but an assumption without working the board so at this point it's total BS.

I have checked the voltage going into the W83977EF and it's exactly 5.0V as specified by manufacturer (Asus is good at this). Even with that the chip is hot like chilli pepper. I have checked the data sheet of the chip and it's frequency is dependant on external clock. There are two crystal oscillators on the board but they are rather far away from the chip. It could be worth to measure them but unfortunately I don't have proper equipment right now. What else could get the chip so hot if the voltage is fine?

derSammler wrote:

Replacing the W83977EF and the VRMs is the minimum required to get the board running again. I wouldn't even care however and just replace the board. It's just an average Slot 1 ATX board without anything special about it.

Asus P3B-F is one of the best Slot 1 motherboards. It's not that rare but in time there won't be many left. That's why I would rather try to fix it before tossing it out.

Paar wrote:

I have checked the voltage going into the W83977EF and it's exactly 5.0V as specified by manufacturer (Asus is good at this). Even with that the chip is hot like chilli pepper. I have checked the data sheet of the chip and it's frequency is dependant on external clock. There are two crystal oscillators on the board but they are rather far away from the chip. It could be worth to measure them but unfortunately I don't have proper equipment right now. What else could get the chip so hot if the voltage is fine?

Those are good checks. An oscillator at the wrong frequency could definitely heat it up.
Are the DC voltages to the oscillators okay?
What I was suggesting with the earlier comment about many outputs is that whatever is at the other end may be shorted and drawing excess current through it.

Check if any capacitors are abnormally hot. If so that's a place to look closer.

Are the main rail voltages okay?

With the main plug out check the resistance between the +voltage pins and ground.
Best to do that with the CMOS battery out.
I can compare your readings to my board and that might give a clue as to which rail to focus on.
Tell me the exact test points so I check the same ones.
We should also have all our jumpers and switches in the same positions.

Look very closely at the LDO's and MOSFETs for shinny spots or pits that suggest melting or arcing inside the package.
Same-same with the main chipset. All the ICs in fact.

Paar wrote:

Asus P3B-F is one of the best Slot 1 motherboards. It's not that rare but in time there won't be many left. That's why I would rather try to fix it before tossing it out.

I picked mine up just a few weeks ago for the same reasons. I was really happy to see all the caps are Japanese already.
Asus was not very good about using quality caps after the 90's.
.

I have checked LDOs and MOSFETSs for shiny spots but coudln't see anything. None of capacitors are hot, only the Super I/O chip.

Resistances (measured with multimeter set to various settings):

+3.3V (orange) - 14.4 (200k)
+5V (red) - 132 (200)
+12V (yellow) - 39.9 (2000k)
-12V (blue) -83.5 (2000k)
+5V VSB (purple) - value keeps rising, cannot measure that

Is that helpful?

Definitely a problem in +5v.
-Might- be a problem in 3.3v.

The rising one on +5vsb is due to the meter charging up a cap.
Mine does the same.

On my board +5v behaves the same as +5vsb and it goes much higher than the (200) range you used can even see. It's more than 200 in a fraction of a second and starts slowing down around 800.

That you have a steady low reading suggests a partial short somewhere.
On this kind of board there are lots of caps on +5v.
Because they are all in parallel through the board you have to remove them to check them.
There may be a +5v powered 3.3v (out) regulator to the AGP slot.
The CPU VRM is powered by +5v on this.

My day has ended so I'll stop by tomorrow. Good luck!
.

Thank you for the tips. I could try to do complete recap once I'm at it as I have no way to know which caps are on the +5V rail. Could replace all +5V regulators as well.

I thought I wrote it but I didn't.
Your 3.3v read 14.4, mine read 19.6.
I'm just noting that for now. Not sure if it's a problem or not.
It may be that we have a jumper of switch on different settings.

Paar wrote:

I have no way to know which caps are on the +5V rail.

Do a resistance check between the +5v pin in the PSU connector (rather the solder point for it as you will be on that side of the board) and the pins to the caps.
Any that read zero are on the +5v rail.

If you pull one out and the problem with the +5v reading from earlier goes away then you found the last shorted one on +5v.. Checking the +5v to ground after pulling each cap might save you some time.

Paar wrote:

I could try to do complete recap

As you are so determined I say go for it. The caps are like 20 years old and don't last forever.
I counted 20 caps and all but one look easy.
That 6.3mm near the CPU slot might be harder to get out but still not THAT hard.
Do a Cap-Map first. Don't want to get them out then forget which way was negative.
If the crystals are cheap you might consider just replacing those too.

Please don't spend more than the board or the experience is worth to you.
The outlook is not good on something like this. It just isn't.

Paar wrote:

Could replace all +5V regulators as well.

I wouldn't blanket replace those. Least not the ICs.
The back side is usually soldered all across and they are a bugger to change out.
Look up the numbers on them and find the datasheets.
Most can be checked with a DMM but they aren't all 'wired' the same.
There are tutorials online that tell how to check them.
I can't promise to be here often as I have a busy time ITRW coming up.
Checking the little SMD parts right near them isn't a bad idea.

You/we are expecting shorted caps at this point.
Shorted ones don't need much to find them once they are out.
Shorted is shorted. Capacitance and ESR don't even matter for that.
An analog resistance meter works better but you can sort-a do it with a DMM.
=Connect the leads, the reading should rise as the meter charges the cap.
=Reverse the leads. Should start negative/go to zero/then repeat what it did on the other side.
That is the charge you just put on it discharging and then it charges up again the other way.
If you get a low reading or not much response both ways the cap is shorted or partially shorted.

This is easier to do on an analog meter because you can watch the needle move as the cap charges/discharges. DMM's are harder for this because it's hard to tell if rapidly changing numbers on the display are going up or down. It can be done with a DMM. It's just not so easy.

Obviously you have a DMM but I dunno what other equipment you have available.
I also dunno your skill level but so far you seem to know what you're doing.

[I am not suggesting you buy one. Just telling things to consider if you want one.]
You do not HAVE to have an analog meter. It's just easier for this check with one.
You do NOT need an expensive one. A $10-$15 cheapie is fine for this kind of check.
One with a reasonably sized meter face is a good idea.
The very tiny ones are hard to read and almost useless if you ever need it for actual numbers rather than just to see a trend.
.

I have removed all caps on +5V rail and the resistance didn't change. After that while I was in the process proceeded with complete decapping. Still no change. After that I thought that the W83977EF chip gets hot and sits on +5V rail... So I have removed it with hot air station and finally got some progress. New resistances:

+3.3V - 1592 (2000 range)
+5V - 610 (2000 range)

All values are immediate as I don't have any caps soldered on the board. Still not the same values you have but it is a progress. All my dip switches are set to OFF, jumper JP20 is set to 1-2 and jumper JEN is set to 2-3. There is one more jumper between PCI slots and it is set to 3-4.

That's really good news.
Hopefully that's the only problem with it.

Yes, it won't work to compare caps to no-caps.
.

Fingers crossed... I have ordered new chip so we'll see how to board will behave when the chip is replaced. Thank you for your help, I would be doomed without you.

Did you fix it?