I am no expert in this, but I would start by checking with a multimeter if the 5v or 12v pins on the card's molex connector are shorted to either ground or each other .

Unless the issue is only on my end, the photo provided is not very useful as it looks to be both low resolution and out of focus . Providing good quality high resolution photos of both sides of the card would definitely not hurt .

Thanks darry for your answer.
Yes, I've checked Molex connector and it's Ok.

Sorry for the picture quality, I've tried to correct it but my ten years old phone is not the best 😀

Dunno if the direct photo links will work:
https://i.ibb.co/9hRmsY1/Photo4518.jpg
https://i.ibb.co/XSqkL3m/Photo4514.jpg

Some soldering looking a bit odd by the caps near the VGA connector.

[Oh, and what were the resistances from +12V to GND and +5 to GND (and +12 to +5 just for completeness sake)?]

Thanks snufkin for your answer.

Yes, I have seen it, but the capacitor is ok.
What can cause a short on such a card ? Capacitors ? And how can I find and test it ?

What setup do you have?

Just a thought. What PSU are you using ? Maybe the card is OK but your PSU is not up to task.

The base of that capacitor in that group at the top left (I think it's labelled as C1155???) looks odd - either it's leaked, or it has previously leaked and has been replaced and the flux not cleaned up; it's 'wet' looking compared to the others. There's a horrible blob of solder at the base of it too.

Thanks for all your answers.

The setup is fine and the power supply too. There is something wrong with the card as as soon as the computer is powered up, it stops. With others cards, there is no problem.
Concerning capacitors, there was something strange on it when I received the card, maybe corrosion related, but it seems that they are intacts.
Concerning this particular cap, yes I think it has been changed but it looks better than the others.

Strange thing is that without the molex connected, the PC do not shut down, but there is nothing on screen (cards like the 9800 pro displays a power related warning).

It sadly could just be a dead card so many of the ati cards of this era had terrible cooling systems and cooked to death. Have you tried it in another computer and was it as tested and working?

Yes, all the 9700 and 9800 card I bought are at best with artefacts and all the others, dead 🙁
But this one had a good cooling system and I think the problem is elsewhere (maybe I'm wrong).

Tested on other system and same results 😒

I have a Dell OEM 9700 that works fine . It has the same RAM and (crappy) cooling as a 9700 Pro, AFAICR, but slower clock speed, so it probably survived thanks to that . I did shim removal surgery on it and it still works fine .

Might be an option for you, if you can find one not too expensive .

I'd agree that the lower clocks saved it's life

parhelia512 wrote on 2021-05-04, 20:19:

Yes, all the 9700 and 9800 card I bought are at best with artefacts and all the others, dead 🙁
But this one had a good cooling system and I think the problem is elsewhere (maybe I'm wrong).

Tested on other system and same results 😒

Some of your artifacting 9800s may only need re-balling of the BGA RAM chips . I have one such card that artifacted heavily unless light pressure was applied to one of the BGA chips while running (not sure where I stashed it, as I know have a place less than a mile away that can do re-balling and reflows and would like to have it fixed). I have heard that this failure mode was common on those cards .

Darry, same here, I need to find a way to reball one vram IC on my 9800.

Problem is I don't have a stencil for that type of IC. Is there a way to get one so I can plant fresh leaded solder balls. I know this as I repair phones is my main job, reballing tiny chips before installing.

Cheers,

I'm still assuming that the PSU turning off when the molex is connected would be due to an overload. So if there's not a short from 12->G, 5->G or 12->5 (can you get resistances for those?), then perhaps there's one to one of the AGP connector pins.
You might be able to identify which voltage is causing the overload by removing the 12V wire from the molex plug and seeing if the PSU still turns off. If it does then you know the problem is with the 5v; if it stays on then the problem is probably with the 12V.
Then you just (it'll be tedious) need to take the card out and buzz through to each of the AGP pins. Looking at the pinout, it looks like only A1 is 12V, so no other pins should have a short to the 12V on the molex. And the pinout says only B2&3 are 5V.

Hmm, it's possible that the 12V and 5V on the edge connector don't connect to directly to the inputs on the molex since the card has to be able to sense whether power is connected to the molex, so there might not be a direct short from 12V molex -> 12V connector.

If there is a card edge pin shorted to the molex/power, that'd probably explain why there's nothing displayed even when the molex isn't connected and the PSU stays on.

Did you check all 11 8-pin MOSFETS on the top side and the one on the backside?
I think some are in VRM circuits and some are just for switching between Molex<->AGP power supply.