I got a card with this Chip broken and if I remember correctly it was a Dual Channel MOSFET. Sold the card. Dont know more Details Sorry.

Missing part is an STS8DNF3LL or comparable. Dual N-channel fet. 30V rating. A.

https://www.st.com/resource/en/datasheet/sts8dnf3ll.pdf

tehsiggi wrote on 2026-04-16, 18:50:

Missing part is an STS8DNF3LL or comparable. Dual N-channel fet. 30V rating. A.

https://www.st.com/resource/en/datasheet/sts8dnf3ll.pdf

Thank you very much!
Did you check this on your board? (I just want to be sure; my board is a bit simpler than the Sapphire in terms of components; it's a PalitDaytona).

@AncapDude, thank you too for trying!

adamsr wrote on 2026-04-16, 20:24:

Thank you very much! Did you check this on your board? (I just want to be sure; my board is a bit simpler than the Sapphire in t […]
Show full quote

tehsiggi wrote on 2026-04-16, 18:50:

Missing part is an STS8DNF3LL or comparable. Dual N-channel fet. 30V rating. A.

https://www.st.com/resource/en/datasheet/sts8dnf3ll.pdf

Thank you very much!
Did you check this on your board? (I just want to be sure; my board is a bit simpler than the Sapphire in terms of components; it's a PalitDaytona).

@AncapDude, thank you too for trying!

Does it have the same layout as the card on the picture? Then yes, it's the right part number. It's the same as on the 9700Pro for which the reference schematic is available. I also repaired a card with a lost fet a while ago.

It looks like a capacitor C94 is missing too.
I guess the Q8 might be as Q5 E6N03 - NTMD6N03R2G 30V 6A (Dual N-Channel SOIC-8)
But mentioned STS8DNF3LL has better parameters

tehsiggi wrote on 2026-04-16, 18:50:

Does it have the same layout as the card on the picture? Then yes, it's the right part number. It's the same as on the 9700Pro for which the reference schematic is available. I also repaired a card with a lost fet a while ago.

Yes, the board in the photo is what I got. I don't even know which 9800 it is (SE or Pro), but it's clearly 256-bit.
Before the first launch, I need to restore the missing parts, at a minimum. Thanks again!

Roman555 wrote on 2026-04-17, 06:57:

It looks like a capacitor C94 is missing too.
I guess the Q8 might be as Q5 E6N03 - NTMD6N03R2G 30V 6A (Dual N-Channel SOIC-8)
But mentioned STS8DNF3LL has better parameters

Yeah, good find, I was focusing while looking at small 0402/0603 and didn’t notice the elephant in the room.
Thank you!

C94 is 470µF at 6.3V - low ESR.
Yes, without replacing the FET you'll get no memory power and thus the card will not work.

I see more ripped out parts from those pictures: three larger SMD caps by the lower-left corner of the GPU package, the fan connector, C212 (22 uF Tantalum between the two memory chips on the right side)... and probably more if I looked at it in more detail. And if there is that much damage on the front of the card, I expect the back would be even worse... but keeping my fingers crossed for you that it isn't.

In regards to the missing MOSFET: too bad it's in a bit of a tight spot. Otherwise you could "kludge it" with two TO-252 MOSFETs standing on their sides(been there done that a few times 🤣 since I have accumulated a good supply of TO-252 MOSFETs.)

momaka wrote on 2026-04-17, 19:09:

I see more ripped out parts from those pictures: three larger SMD caps by the lower-left corner of the GPU package, the fan connector, C212 (22 uF Tantalum between the two memory chips on the right side)... and probably more if I looked at it in more detail. And if there is that much damage on the front of the card, I expect the back would be even worse... but keeping my fingers crossed for you that it isn't.

In regards to the missing MOSFET: too bad it's in a bit of a tight spot. Otherwise you could "kludge it" with two TO-252 MOSFETs standing on their sides(been there done that a few times 🤣 since I have accumulated a good supply of TO-252 MOSFETs.)

Not really. There's nothing else broken on the card, including the back side (although that's hard to believe, considering the chip and capacitor broke off... I can't imagine how that could have happened).
It's not Sapphire, it's Palit (a bit simpler ). Everything you see as "missing" is unsoldered from the factory.

tehsiggi wrote on 2026-04-17, 11:45:

C94 is 470µF at 6.3V - low ESR.
Yes, without replacing the FET you'll get no memory power and thus the card will not work.

Unfortunately, restoring the electrolytic capacitor and dual-channel MOSFET didn't help—the computer won't initialize with this board.

What I did:
1. Soldered in an SDM9926 MOSFET, removed from a donor FX5200 board of unknown condition. I tested it by opening and closing it off the board—everything seemed fine.
2. Soldered in a 1000µF solid-state through-hole capacitor in place of the SMD capacitor (temporarily).
3. Under a layer of old thermal paste/yellow glue, I discovered a missing ceramic capacitor. I don't know what value it needs, so I didn't solder anything. But I think it's highly unlikely that this could be the reason the card won't boot.
4. I checked the AGP connector for voltage drop - no anomalies (I have some experience with this).
5. I measured the voltages at several points:

Unfortunately, I don't know where to look next...
I started collecting AGP cards about six months ago (I currently have about 70 video cards), and even less recently I started getting interested in repairing them, as a significant portion of them are inoperable. I managed to fix some, partially repaired others, but most often I get stuck when I see an anomaly, because I don't know where to look next. I hope you can give me some advice.

GPU resistance is about 12-16Ohm
Memory resistance - 69Ohm
BIOS VCC - 3.2V
Quartz resonator: 0V on one contact and 0.29V on the other (this seems to be the problem)

P.S. Replacing the quartz resonator did not produce any results...

adamsr wrote on 2026-04-18, 11:51:

GPU resistance is about 12-16Ohm Memory resistance - 69Ohm BIOS VCC - 3.2V Quartz resonator: 0V on one contact and 0.29V on the […]
Show full quote

GPU resistance is about 12-16Ohm
Memory resistance - 69Ohm
BIOS VCC - 3.2V
Quartz resonator: 0V on one contact and 0.29V on the other (this seems to be the problem)

P.S. Replacing the quartz resonator did not produce any results...

All your voltages appear to be in spec, just re-verified with my 9800 Pros. Resistance is same, so no shorts.

Measuring the quartz can be misleading, as it's part of a low voltage oscillator circuit. It's AC in the MHz range.

Do you have a high res picture of the back of the card as well?

tehsiggi wrote on 2026-04-18, 13:42:

All your voltages appear to be in spec, just re-verified with my 9800 Pros. Resistance is same, so no shorts. […]
Show full quote

adamsr wrote on 2026-04-18, 11:51:

GPU resistance is about 12-16Ohm Memory resistance - 69Ohm BIOS VCC - 3.2V Quartz resonator: 0V on one contact and 0.29V on the […]
Show full quote

GPU resistance is about 12-16Ohm
Memory resistance - 69Ohm
BIOS VCC - 3.2V
Quartz resonator: 0V on one contact and 0.29V on the other (this seems to be the problem)

P.S. Replacing the quartz resonator did not produce any results...

All your voltages appear to be in spec, just re-verified with my 9800 Pros. Resistance is same, so no shorts.

Measuring the quartz can be misleading, as it's part of a low voltage oscillator circuit. It's AC in the MHz range.

Do you have a high res picture of the back of the card as well?

Sure:

I didn't find anything criminal on the back side, I also soldered a 0.16μF ceramic capacitor onto the GPU substrate.
I probably need to measure something else, but I'm stumped. The oscilloscope is still on its way, and I need to figure out how to use it 😁

In diagnostic purpose you can add a PCI graphic card in pair to the Radeon.
Set the PCI card as a primary videocard in BIOS settings and then you'd see if the Radeon appeared in a PCI devices tab of any sysinform utility.
Maybe you can read videobios. Can it be the card from Apple computers?

Roman555 wrote on 2026-04-19, 17:59:

In diagnostic purpose you can add a PCI graphic card in pair to the Radeon.
Set the PCI card as a primary videocard in BIOS settings and then you'd see if the Radeon appeared in a PCI devices tab of any sysinform utility.
Maybe you can read videobios. Can it be the card from Apple computers?

The problem is that the PC won't boot if this card is inserted as the second-highest priority after the integrated/PCI card. It's possible there's a BIOS issue, and it might be worth removing it and trying to read it with a programmator, or trying shorting pins 1 and 8 of the BIOS chip on the board during startup... I don't know.
It's unlikely that this is a Mac card; it wasn't common in our area (Post-Soviet region) — firstly, and secondly, the R9800Pro cards I've seen for Macs look exactly like Sapphire cards.

I had a look at the pictures again. Nothing that strikes my eyes. The missing cap on the GPU substrate is for VDDC, shouldn't be an issue. You might try to read the BIOS chip. I had cards as well, that wouldn't even allow the board to post as a secondary GPU, no real idea why.
But please don't just short pins on the BIOS IC, it's running off of the PSU rails and that has hefty power.

adamsr wrote on 2026-04-21, 11:28:

The problem is that the PC won't boot if this card is inserted as the second-highest priority after the integrated/PCI card. It's possible there's a BIOS issue, and it might be worth removing it and trying to read it with a programmator, or trying shorting pins 1 and 8 of the BIOS chip on the board during startup... I don't know.
It's unlikely that this is a Mac card; it wasn't common in our area (Post-Soviet region) — firstly, and secondly, the R9800Pro cards I've seen for Macs look exactly like Sapphire cards.

Yes, maybe Palit can't be a mac card...
But I don't think it is a Video BIOS issue if the PC doesn't boot at all. At least I would expect booting with it as a secondary adapter. Correct me if I'm wrong.
Does the PC stop with some POST codes or just stay with active RESET signal? RESET state can be discovered using POST (debug) card or to measure voltage of RST signal in PCI slot. Obviously POST codes you can observe only on a POST card.

Roman555 wrote on 2026-04-21, 16:09:

Yes, maybe Palit can't be a mac card... But I don't think it is a Video BIOS issue if the PC doesn't boot at all. At least I wo […]
Show full quote

adamsr wrote on 2026-04-21, 11:28:

The problem is that the PC won't boot if this card is inserted as the second-highest priority after the integrated/PCI card. It's possible there's a BIOS issue, and it might be worth removing it and trying to read it with a programmator, or trying shorting pins 1 and 8 of the BIOS chip on the board during startup... I don't know.
It's unlikely that this is a Mac card; it wasn't common in our area (Post-Soviet region) — firstly, and secondly, the R9800Pro cards I've seen for Macs look exactly like Sapphire cards.

Yes, maybe Palit can't be a mac card...
But I don't think it is a Video BIOS issue if the PC doesn't boot at all. At least I would expect booting with it as a secondary adapter. Correct me if I'm wrong.
Does the PC stop with some POST codes or just stay with active RESET signal? RESET state can be discovered using POST (debug) card or to measure voltage of RST signal in PCI slot. Obviously POST codes you can observe only on a POST card.

The BIOS was unsoldered and read without any problems, the chip was erased and the same dump was written again.
On the postcard, it sometimes stumbles on the d0 code, sometimes it gets to the a2 code. There's 3.36V on the RST# pin of the AGP.