Dead 2900XT, and ways to fix it?

[M-XXXX]

My 2900XT is dead (for few months now). One day, screen has just gone black (it shows picture in one of 20 powers up, but its full of squares (in bios) so its almost 100% RAM) When I took it out and after close inspection I've seen some elements that were rusty and some "lines" in PCB are opened. Ive carried card to the shop for RMA, but they declined it as physical damage. Then I've tried to clean that element but that didn't do any good too:mad: Then I was bit more aggressive, but that didn't brought any good. Now I've sent it to my friend, who changed rusty element (didn't helped), then I've sent it again (nice and shiny), but it was refused again.
I've measured voltages (vGPU and VDDQ are OK), but main RAM voltage is close to 0, and jumping. My idea is to connect some external 2.2V cable and hope for the best? Is this possible?

This is situation after more aggressive "cleaning":

http://shrani.si/t/N/D4/1rutkmKs/dsc01778.jpg (http://shrani.si/?N/D4/1rutkmKs/dsc01778.jpg)

[Martijn]

That card looks pretty bad. 'Agressive' cleaning is not the thing to do with ICs. I really have no idea what tool you used, but the one you used wasn't the right one for sure. If the contacts are a little rusty, leave it as is. It's only the top surface, not the material below it. Right now, you've created an uneven surface, making corrosion even easier.

From the looks of it, it seems the center pin isn't making proper contact. Also, it could be possible you completely sanded away the lane on the 4th pin, if there even was any.

Long story short: I think you ruined your 2900XT with the 'cleaning'.

[cozwin]

what the hell did you clean it with a screw driver

[Pete]

Bin fodder now! Next time be nicer

[MichelinGuy]

The IC's on my 2900XT are goldenbrown, but it still works so I don't think your card failed because of that.

[STEvil]

Try attaching external +2.0v-2.3v source to vmem... cant kill it any more than it is.

[M-XXXX]

Try attaching external +2.0v-2.3v source to vmem... cant kill it any more than it is.
This is my intention... But I would like to get some more advices (where should I connect to? To vmem measure point or somewhere else? Probably from molex would be OK? (afcourse about 2.2V)

From the looks of it, it seems the center pin isn't making proper contact. Also, it could be possible you completely sanded away the lane on the 4th pin, if there even was any.
Those "scratched" elements are replaced, but "wires" inside PCB are missing (like in picture)

The IC's on my 2900XT are goldenbrown, but it still works so I don't think your card failed because of that.
When card failed some of the "wires" in PCB burn out (in picture) I dont know what else could kill it...

[boqonx]

Is the C1018 still making any contact with the board? Do some continuity checks on the components to see if you have any last chance of reviving the card. Can we use something like WD-40 to cover up the rusty parts in order to prevent further corrosion?

[M-XXXX]

Is the C1018 still making any contact with the board?
Im not sure- card is still in the RMA (they even won't sent it back- I have to get it (I will get it tomorrow), but Icy (person who tried to repair it) says that probably is (but he is not sure). I can tell you more as soon as I get card back.

Can we use something like WD-40 to cover up the rusty parts in order to prevent further corrosion?
As I've told before, "rust" part was changed and it was l lot (sorry but I don't know English word for lotting (or is it correct?)), but I can spray it with WD40.

[persivore]

On some cards I've worked with in the past, the RAM voltage supplies are turned on after the card has turned on. Its possible that there is nothing wrong with the memory voltage supply, and there is a fault elsewhere on the card, preventing it from turning on the memory voltage supply after power up.

IIRC there should be a voltage of around 2.5v on the output (pin 5) of the RT9173 if the 2900XT is anything like older ATI PIC-e cards.

[STEvil]

The best connection point for the power source for your memory would probably be a capacitor (or two) and also to the output side of the mosfet which supplies the power on the card normally so that you can get a vref point.

[M-XXXX]

Yesterday I went to get he card:


On some cards I've worked with in the past, the RAM voltage supplies are turned on after the card has turned on.
Probably you mend off? (but probably this is not it, because when I measured VMEM voltage it was low, and it was jumping like crazy.

IIRC there should be a voltage of around 2.5v on the output (pin 5) of the RT9173 if the 2900XT is anything like older ATI PIC-e cards.
I would be glad to give you answer, but I have no idea where this would be (could you give me data sheet, because I dont find it:( )

If you need any more pictures just say...

http://shrani.si/t/1x/10A/3sXJIW58/2900.jpg (http://shrani.si/?1x/10A/3sXJIW58/2900.jpg) http://shrani.si/t/s/Y/osfPjve/img1124.jpg (http://shrani.si/?s/Y/osfPjve/img1124.jpg) http://shrani.si/t/34/yp/4gPMUVl6/img1126.jpg (http://shrani.si/?34/yp/4gPMUVl6/img1126.jpg)

[oohms]

http://www.alasir.com/reviews/sapphire_radeon_9600xt_agp/rt9173.html

[M-XXXX]

IIRC there should be a voltage of around 2.5v on the output (pin 5) of the RT9173 if the 2900XT
Weird- I've measured 1.6V (on working card and on the broken one...)

Edit: This is on the working card:
http://shrani.si/t/3i/6X/3QhEQaV6/img1145.jpg (http://shrani.si/?3i/6X/3QhEQaV6/img1145.jpg)

I'm rely born under unlucky star (my Maximus is dead too- probably dead mosfet;(

[M-XXXX]

Anyone:shrug:

[tiborrr]

I think the GPU gives the signal to power up the memory voltage feed. If the GPU is damaged and not functionating properly, it may fail to start powering up the memory.

Had similiar problems with damaged X850XT(overvoltage - short circuit killed the GPU), it gave me ~ 0-1.1V on the memory instead of 1.9V, GPU was gone... Even after adjusting the Vddr & Vddq higher with voltmod, it wasn't any better.

All in all, my bet is on the GPU.

[M-XXXX]

tiborrr did your card shown any picture after failure? My shown picture few times (JUST in bios), after going to windows and loading drivers, picture was gone in no time, so this still give me some hope:p:

Because card doesn't work anyway, im at least going to try something, if that doesn't help, then we I will do:

http://shrani.si/t/2E/Wi/1W6NWBYc/ati-repair-kit.jpg (http://shrani.si/?2E/Wi/1W6NWBYc/ati-repair-kit.jpg)

:up: :)

[begrip]

R.I.P that was a good card dude :P:

[ZenEffect]

missing pcb trace could be replaced by conductive ink pen?

not sure but it seems like it might work, granted the ink is not too resistive

[celemine1Gig]

missing pcb trace could be replaced by conductive ink pen?

not sure but it seems like it might work, granted the ink is not too resistive

Or even better, by soldering a wire into place. But I think the trace has already been fixed some time ago by M-XXXX and it didn't help.

[M-XXXX]

Or even better, by soldering a wire into place. But I think the trace has already been fixed some time ago by M-XXXX and it didn't help.
Those all all good ideas. That connection wasn't touched.