Repair a memory chip faulty Sapphire RX 470 Nitro+ 8GB

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
So I have a Sapphire RX 470 Nitro+ 8GB video card, which I think is memory chip faulty.

sapphire_rx470_memory_chip_faulty.jpg


I tried to measure the resistance and found only one problem:
- There should be a ~20 ohm resistor in the memory controller, but there is nothing.
- However, the 4c10n mosfet in front of the memory chip has a ~20 ohm resistor, which should not be.

What do you think could be the fault of your video card? I suspect one of the memory chips is faulty. In this case, how do I know which one? What is the most common case, could the chip fail, or could a reballing be enough?

Or maybe it could just be the fault of the mosfet? Would it be enough to replace it?

Thanks in advance for all the ideas and help.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,919
The component you have marked as the "memory controller" is an inductor. It's just a dumb spool of wire inside that ferrite overmolding, and is not part of your problem.

Look at the component I've marked in blue, here:
deadrx470_marked.jpg

You'll see it has two terminals - one on the left and one on the right, when you look at the card in the same orientation as your photo. Take a resistance measurement from each terminal, by putting your multimeter's red probe on the terminal, and the black probe on the housing of the DVI connector.

What is the resistance you get on the left terminal?
What is the resistance you get on the right terminal?
 

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
The component you have marked as the "memory controller" is an inductor. It's just a dumb spool of wire inside that ferrite overmolding, and is not part of your problem.

Look at the component I've marked in blue, here:
View attachment 348886

You'll see it has two terminals - one on the left and one on the right, when you look at the card in the same orientation as your photo. Take a resistance measurement from each terminal, by putting your multimeter's red probe on the terminal, and the black probe on the housing of the DVI connector.

What is the resistance you get on the left terminal?
What is the resistance you get on the right terminal?

I measured. This is abnormal. 1.5 ohms on both sides.

This should be 20 ohms on the left.

So can this be the problem? Should I replace it?
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,919
1.5 ohms on both sides means you have a short to ground. Your next step is to find the short.

Do you have the ability to inject voltage into the circuit with a bench power supply?
 

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
1.5 ohms on both sides means you have a short to ground. Your next step is to find the short.

Do you have the ability to inject voltage into the circuit with a bench power supply?

Awesome, thank you for your help so far. I feel we are close to a solution.

Yes, I have a lab power supply.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,919
Solder the positive lead to one of the 12V input pins on the 8 pin connector. Solder the ground lead to one of the ground pins.

Set the voltage of the power supply to 1.5V or less. Do not exceed 1.5V. Limiting the amperage is less important, but start around 10 amps and see if that's enough. If not, go higher.
 

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
Hm, I was trying to introduce 1.5V and the current went up to 1A. I let it warm up for a few minutes, but nothing, I didn’t feel anything warm. :(

Maybe I should try to use freezer spray? Although it is quite expensive.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,919
Hm, I was trying to introduce 1.5V and the current went up to 1A. I let it warm up for a few minutes, but nothing, I didn’t feel anything warm. :(

Maybe I should try to use freezer spray? Although it is quite expensive.
With the right safety precautions, you can use isopropyl alcohol instead of freeze spray. You put a little bit of it on the board in the area you suspect lies, and look for it evaporating. Edit: You can also use a thermal camera, if you have one.

Also, you'll need a lot more than one amp. 1.5V times 1A is only 1.5 watts. You'll need more like 20 or 30 watts if you don't have a really good thermal camera. Dial the power supply up to 10 amps and see what happens.

With the coil removed, solder the positive lead to the pad where the coil was - the one you're probing in that second photo. Start at like 5A, and go up to 10 if you don't find anything. Keep in mind that most of the replaceable components in the memory circuit tend to be on the back of the board.
 
Last edited:

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
I suspect the GPU, so I tried the PCI-E 16x connector diode measurement again, more carefully to see if I missed something and I think I found something. On side A, #81 pin is faulty. I attached a picture.

20210424_120343.jpg


So is the GPU shorted?
 

grofmarton

n00b
Joined
Apr 14, 2021
Messages
8
Also, you'll need a lot more than one amp. 1.5V times 1A is only 1.5 watts. You'll need more like 20 or 30 watts if you don't have a really good thermal camera. Dial the power supply up to 10 amps and see what happens.

With the coil removed, solder the positive lead to the pad where the coil was - the one you're probing in that second photo. Start at like 5A, and go up to 10 if you don't find anything. Keep in mind that most of the replaceable components in the memory circuit tend to be on the back of the board.

Yeah, but I have tried so far with 1.5V and set the limit to 2A but it doesn't go above 1A, it doesn't use more than that.

Hm, I was trying to introduce 1.5V and the current went up to 1A.

-> Sorry if I was misunderstood. I meant that I set a 5A limit, but it didn’t go up higher than 1A.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,919
I suspect the GPU, so I tried the PCI-E 16x connector diode measurement again, more carefully to see if I missed something and I think I found something. On side A, #81 pin is faulty. I attached a picture.

View attachment 350471

So is the GPU shorted?
Possible.

Given the mysterious memory rail short (those are the worst), I'd probably move on to an easier case at this point. Keep in mind that it's also possible for the short to occur inside the board itself. In those cases, you're unlikely to find it without an X-ray machine, and you can't really repair it, even if you did.
 
Top