in theory U504 supply is a mirror image of U503 one on the left, but they arent really perfect copies and I cant find corresponding resistors on the picture

Just throwing out a slightly random guess whilst waiting to see if anyone can measure the actual resistors, but given the rotation of the chips then I've circled what I guess are the matching 4 resistors and empty pad.

Something else that might be worth checking if you want to is if any of the pads with missing components are connected to either ground or a power plane. If so then they might be pull-up/down configuration strap resistors, in which case something like a 10k resistor may work. But getting the actual values is obviously better than guessing.

[picture grabbed from https://www.techpowerup.com/review/inno3d-gef … platinum/3.html ]

rasz_pl wrote on 2022-10-29, 06:35:

in theory U504 supply is a mirror image of U503 one on the left, but they arent really perfect copies and I cant find corresponding resistors on the picture

Yea I figured the same thing out, but like you sad the layout is not an exact copy.

snufkin wrote on 2022-10-29, 10:33:

Just throwing out a slightly random guess whilst waiting to see if anyone can measure the actual resistors, but given the rotati […]
Show full quote

Just throwing out a slightly random guess whilst waiting to see if anyone can measure the actual resistors, but given the rotation of the chips then I've circled what I guess are the matching 4 resistors and empty pad.
Possible_NVI02_match.jpg

Something else that might be worth checking if you want to is if any of the pads with missing components are connected to either ground or a power plane. If so then they might be pull-up/down configuration strap resistors, in which case something like a 10k resistor may work. But getting the actual values is obviously better than guessing.

[picture grabbed from https://www.techpowerup.com/review/inno3d-gef … platinum/3.html ]

If I recall all of them where grounded but I could be wrong.
I have measured resistors around the card and some are 0 ohm other 28K and if I recall one set of pads had very low resistans like 150 ohm, but could be wrong.

I was also thinking maybe 10-20K resistors just to see what happens but that dont make to much sens to me on say very low resistans pair of pads anyways.
And I do find like sets of 3 here and there where the values vary a LOT like 120 ohm VS 28K so yea guessing there values seems like at least one would be way out of spec.

The card do run BUT when I got it a few years ago it did not run in PCI-E Gen1 boards due to missing SMD caps and traces at the PCI-E connector.
Fixed that and it seem to run "fine" in UNIGINE Haven Benchmarks and I did play som Fallout 3 on it.
But a while ago we had a game night with Halo CE and the card keept doing a black screen randomly, no artifacting.

So I tested it now and it did the same thing at random times in Haven so I figured I fix the last SMD stuff that where missing but well it still doing it.
I do have a GTX 295 double sided card that works (parts from 2 broken cards).

Anyways I figured I try to fix everything someone scraped of the board first because the "Fan shroud" is broken, it fell off when I got it so I reinforced the plastic hooks with epoxy/glasfiber and that lasted well 1-2 years but now it fell of again.
But dont want to wast my time doing a new Shroud unless I can get the card stable again.

But it might just be toast GPU's, 200-500 series card seems to have chips that are prone to go bad.

Anyways if I got it going I do have an AM2 SLI board so the plan was Quad SLI, just for fun like any hobby.

Im gonna go and keep working on the card, see what can be fixed.

The top 3 pads are ground, I also measured the resistans to the bottom pads relative to ground and it was 153ohms for left and middle and 19K on right
Also I did not find them to connect directly to the caps behind the chips so dont seem to involve any VCC for those chips.

.

I also poked those 4 resistors as suggested on the other chip (180 degree relative) and two of them came out 0 ohm, one 34 ohm and one 10K so a bit all over the place, and yes in circuit so might be different values on the actual resistors..

The card is disassembled, needs new pads becuse the the 3 pads for the chips in the middle, PCI-E bridge etc where rock hard, dont think its a thermal issue, the card had new past 2 years ago and runs at about 60-62C but could be a local heat issue on those chips because the past was rock hard now so.

Well, pads shorted directly to ground probably means they're for configuring the chip. What are the resistances to ground for the 4 pads at the 2 empty locations (just marked with a yellow arrow)? If they look the same then maybe they are a matching pair, and then it might be worth removing the 4 resistors from the left-hand side, making sure you keep track of which one came from where. Then you can measure the 8 pads on the left-hand side and compare them with the 6 pads with missing components. If you can match them up with each other then you can see what value resistor was used on the left-hand side.

On the other hand, if they are configuration resistors and the card generally works, then fitting them might not help since if the configuration is wrong then I wouldn't expect it to work at all. On the third hand, 153 ohm might suggest it's part of a transmission line, and if the termination resistor is missing then maybe sometimes the signal gets corrupted.

I'd probably give it a few days to see if anyone can measure the resistors before taking a soldering iron to the card.

snufkin wrote on 2022-10-29, 12:23:

Well, pads shorted directly to ground probably means they're for configuring the chip. What are the resistances to ground for t […]
Show full quote

Well, pads shorted directly to ground probably means they're for configuring the chip. What are the resistances to ground for the 4 pads at the 2 empty locations (just marked with a yellow arrow)? If they look the same then maybe they are a matching pair, and then it might be worth removing the 4 resistors from the left-hand side, making sure you keep track of which one came from where. Then you can measure the 8 pads on the left-hand side and compare them with the 6 pads with missing components. If you can match them up with each other then you can see what value resistor was used on the left-hand side.
Possible_NVI02_match2.jpg

On the other hand, if they are configuration resistors and the card generally works, then fitting them might not help since if the configuration is wrong then I wouldn't expect it to work at all. On the third hand, 153 ohm might suggest it's part of a transmission line, and if the termination resistor is missing then maybe sometimes the signal gets corrupted.

I'd probably give it a few days to see if anyone can measure the resistors before taking a soldering iron to the card.

The board is drying right now from a clean but I recall measuring the empty pads where the missing once are and none of them where ground pads if I recall, I also think the resistans was pretty high so OL basically. Could try Mohm.

I can check in an hour or so again.