Two not working Sapphire RX460 - Repair for educational reasons

SavaBogdan

n00b
Joined
Nov 6, 2021
Messages
15
Hello everyone.
I want to say that I am a beginner in electronics repair but I have very good soldering skills and access to programmers ( Beeprog, Xprog, etc). I work in automotive repairs, mainly ECU-s where I have to swap different parts between them (cloning them) but no experience in actually determining what the faults are with the broken ones.
Ok, so to keep this short. My main goal is to learn how to diagnose and fix electronic components, and GPU-s are pretty interesting.
I just picked up a few bad GPU-s and among them I found two identical Sapphire RX460. Now, the problems they have are different but are physically identical.
I know that two broken cards are not much help in comparing measurements across each other but if it will get to that I am willing to buy another working GPU so I can have a good baseline.
So to the problems:
One card is completely dead, no power on any rails. I just have 12v, 3.3v and 5v.
The other one is shorted, computer just half spins the CPU fan and then OCP triggers in the PSU.
I started with the shorted one because I thought that if I find the short maybe the card will start, and it will be easier. It wasn't :)
Here are photos of the board and what measurements i took:

20211108_161434 Traces.jpg20211108_161424 Traces.jpgGS7103+GS7133.png
1. 12v to ground - stabilizes at about 6.6 kOhm
2. 3.3v to ground - 29Ohm ( This seems very low, the card with no power has 231Ohm and a good Gigabyte RX460 has 1.7kOhm)
3. ?3.3V - 46Ohm
4. ?Memory - 43Ohm
5. Voltage regulator for 5v - 15kOhm
6. Another voltage regulator for 5v - 2 kOhm
7. Vcore - 2Ohm
8. Pin 6 on the dropout linear regulator - 0.2Ohm (this is short to ground, but on both bad cards?)
9. Pin 8 on the regulator - 2.4 Ohm
I don't understand this part. Pin8 is ground and pin6 is Vout. On one card the chip is GS7133 and the other GS7103. They seem to have the same function.
10. ?Memory controller - 60 Ohm

All these measurements are from the shorted card.
Some values are the same on both cards and some differ. I don't want to over complicate the post so I won't make the comparisons now.

Maybe RazorWind can lend a helping hand. He seems to really know his way around a GPU :)

Thank you
 
Last edited:
Which rail is it that you think is shorted? Just the 3.3V?

4 is probably the aux (memory controller) rail. 10 is most likely the memory.

Can you post a closer photo of the SOP-8 package LDOs that you're having trouble with, so I can see the markings on them?
 
Here are the chips. The one on the shorted card is GS7133.
I suspect 3.3v because of the low resistance, especially compared to the working but different model RX460.
I may be completely wrong though :)
 

Attachments

  • 20211108_195915.jpg
    20211108_195915.jpg
    261.7 KB · Views: 0
  • 20211108_195223.jpg
    20211108_195223.jpg
    275.6 KB · Views: 0
You might try removing the LDO and see if that clears the short. I'm not that familiar with this particular board design, but it's possible that these little LDOs are underpowered for whatever it is they're supposed to power, and that they commonly fail with a short to ground.
 
I desoldered it and checked 3.3v pin to ground resistance and I still have 29Ohm. I plugged it in and started the PC and the short is gone. The problem is that the core gets hot pretty fast. I started the card with my finger on the GPU and in about 2 3 seconds it got pretty hot, so I turned it off. Hope I didn't kill the GPU.
I then soldered the LDO from the other card that was not shorted. Plugged it in and the short is back. I also checked the LDO and there is no short between pin 6 and 8. I will order some chips but I think the problem is not the chip itself. Maybe the short is where the power goes to? That would explain why there is no short with it removed. I also now measured the pad from pin 6 to ground and I have 0.2Ohm. That doesn't seem right no?
 
Did you actually check the resistance to ground on the pad that's supposedly the output from that LDO while you had it off? What is the actual resistance in ohms that you're seeing?
 
Yes, I took the chip off again. The 0.2Ohm are from pad 6 to ground.
I saw on youtube that someone trying to repair a RX460 found a schematic for a MSI RX460. Maybe it has a similar layout, or at least find out what that chip is suposed to power. I never looked at an electronic circuit board schematic before, but I am familiar with car electric systems schematics so I guess I should be able to figure it out.
I just now realized I made a mistake on the image. Nr. 9 is pin 6 and it had 0.2ohm, and nr. 8 is pin 8 and it had 2.4ohm. Pin 6 has same value with or without the chip.
 
Ok, so what I have found today.
I got hold of a schematic for similar boards and I found the GS7103. The problem is I can't figure out where it goes from pin 6. It says it's 1.8v but I don't know what 1.8v is used for.
I did a few google searches and someone said something about memory and BIOS chip. So I looked closer at the BIOS chip and saw a burn mark. A small bump like these chips form when they burn out. So I took off the chip almost certain that was the problem.
Unfortunately the short is still there. I am attaching the PDF with the schematics I found, maybe someone can make sense of them or can give me a suggestion how to proceed in finding the short.
I measured pin 6 from that chip on a working RX460 and the value is in the kOhms so I am pretty sure here is the problem.
Maybe a shorted capacitor somewhere? I unsoldered and checked for shorts the 4 capacitors near the GS7133 and they are all fine.

20211109_190828.jpg 20211109_190813.jpg

GS7103 Schematic.png
 

Attachments

  • 469930019-rx580-rx570-rx480-rx470-RX-MS-V341.pdf
    2.1 MB · Views: 0
Well, not great news. I watched this video:

The GPU core has 1.8v in, and at 1:27:09 he says that when you find a short on 1.8v it's almost certainly a dead GPU core. And this is true for both cards.
I would have to remove the core to be sure but i don't have a reballing table to be able to heat the board enough. I think i might buy one, it's about 400$ so not that bad.
One thing I don't understand. The guy says that 1.8v is the first voltage rail that comes on, and in turn enables the rest. If i desoldered the chip that generates 1.8v how is the core getting 1.1v? It does get hot and I measured the voltage because I wanted to know if it is short from 12v or 3.3v. And it was 1.1v like it should be.
Quite a steep learning curve on this GPU repair business :)
 
Well, not great news. I watched this video:

The GPU core has 1.8v in, and at 1:27:09 he says that when you find a short on 1.8v it's almost certainly a dead GPU core. And this is true for both cards.
I would have to remove the core to be sure but i don't have a reballing table to be able to heat the board enough. I think i might buy one, it's about 400$ so not that bad.
One thing I don't understand. The guy says that 1.8v is the first voltage rail that comes on, and in turn enables the rest. If i desoldered the chip that generates 1.8v how is the core getting 1.1v? It does get hot and I measured the voltage because I wanted to know if it is short from 12v or 3.3v. And it was 1.1v like it should be.
Quite a steep learning curve on this GPU repair business :)

The $400 dollar ones are probably the really low-end models. Good ones will cost 1k+
 
I was thinking about Puhui T862++:

I don't plan on starting a business, just to develop a new skill, more like a hobby.
But I wouldn't mind spending a bit more if it really makes a difference in quality. Can you recommend something?
 
Of course I could't help myself and I removed the GPU. Easier than I thought, but the chip is small. Anyway, the short is gone. So I guess that was the problem.20211110_145922.jpg
 
knowing is half the battle.jpg


The red and blue lasers is finding an RX460 core. I actually have two gigabyte versions of that card. One is in use and one is spare. I had 3 of them, but sparked the best one that ran at the lowest voltage fiddling around with a screwdriver in the case a few years ago. Good luck. Some of those 460s can also be flashed to use additional cores. They were neutered from the factory.
 
I found GPU cores on aliexpress. They are around 70$ so I think I might order one to play around with. I know it doesn't make sense financially speaking but the satisfaction of correctly identifying the problem and actually seeing the card alive again might be worth it. Or maybe I can find another bad RX460 and hope it's not the core again :)
 
I found GPU cores on aliexpress. They are around 70$ so I think I might order one to play around with. I know it doesn't make sense financially speaking but the satisfaction of correctly identifying the problem and actually seeing the card alive again might be worth it. Or maybe I can find another bad RX460 and hope it's not the core again :)
Good luck. Sounds like a fun project if you have the time. I had no idea you would be able to find 460 cores, that gpu is what 5 or 6 years old at this point?
 
Back
Top