Sapphire rx470

Ziko

n00b
Joined
Sep 9, 2021
Messages
23
Hi

I am new to this forum so hi to everyone!!!
Have bad rx470, both 10A from external power 12v blowed, there is short but I tried injection 1.0v with 2A consumption and not any parts getting hot
Mosfets ic is ior355
Also common cap problem from back side looks ok
Any tips
 

Attachments

  • Screenshot_20210909-180922_Gallery.jpg
    Screenshot_20210909-180922_Gallery.jpg
    890.5 KB · Views: 0
  • Screenshot_20210909-180901_Gallery.jpg
    Screenshot_20210909-180901_Gallery.jpg
    875.9 KB · Views: 0
  • Screenshot_20210909-180841_Gallery.jpg
    Screenshot_20210909-180841_Gallery.jpg
    356 KB · Views: 1
Last edited:
What resistance do you have between the spot you soldered that wire and ground? Exact resistance. In ohms.
 
0.9R , when connect both pins instrument shows 0,3-0.4R
But go tell, on 0v8 drove 1.8A, if I put 1v2 then drove more,but I afraid to go higher then 1v2
Tnx
 
Check for continuity between the spot you soldered the wire and the output of the GPU core VRM. If you have zero, then most likely you have a high side FET failure. I'd just start removing them until I cleared the short, and then replace all the FETs I removed.

Use a preheater to remove and reinstall the FETs. You'll destroy them with just a hot air station.
 
You mean to measures on top coil (it's through hole)marked yellow?
To remove fets marked with red arrow,all 3?
If you mean one of 5 phases fets are integrated inside,how can test? And yes on coil outputs is zero ohm,but I don't want to remove all 5 ior3553 but can see any mark on any of them like over heating or ...
About desoldering, you are right,I tried hit air to max temp,all parts around hot but fet not moves at all.
Pre heater not have but will try something

Tnx
 

Attachments

  • 20210909_221521.jpg
    20210909_221521.jpg
    358.6 KB · Views: 0
Last edited:
I inject 1v max and gpu slightly warm,not any other element can feel warm,maybe 12v damaged gpu
 
It's possible that being subjected to 12V damaged the GPU, but there's a pretty good chance that it didn't. I've seen a lot of cases where the GPU worked just fine after clearing the short.

You need to use your knowledge of how the VRM circuit on the board works to figure out which FETs are connected to the blown fuses. Then, you'd start removing the components containing those FETs until you clear the short. You might be able to use isopropanol with injection to figure out of in any of the FET packages are getting warmer than the others.
 
I did with alcohol, not any looks like hot , all VIn ang gnd are shorted,will have to do one by one.
For this card there are 3 videos,same fises blown,one was controller (suspicions to me) on the left side of pcb,BUT as you told to me, I no idea how to desolder can't remove any parts,when i put leaded solder it liquid after hot air but ori solder on bord ,no way,hard like stoune
Tnx
 
I did with alcohol, not any looks like hot , all VIn ang gnd are shorted,will have to do one by one.
For this card there are 3 videos,same fises blown,one was controller (suspicions to me) on the left side of pcb,BUT as you told to me, I no idea how to desolder can't remove any parts,when i put leaded solder it liquid after hot air but ori solder on bord ,no way,hard like stoune
Tnx
Gotta get the board hotter. Either warm it up with a heat gun first or use a real PCB preheater. It makes a huge difference.
 
  • Like
Reactions: Ziko
like this
I warm with heat gun from bottom, max temp from top side,I hardly removed 2 fets, it's crazy, pcb hot can't touch but part doesn't unsolder
 
If there's no visible damage, it's kind of unlikely, but there is a mode of failure where the FETs get so hot after they fail that their terminals actually weld themselves to the copper in the board. In these cases, no matter how hot you get them, you can't remove them without also lifting the copper pads they attach to. I don't think that's what's happened here, but it's possible. This particular failure is super common on the MSI custom design of the 980 Ti. Usually, there's also some catastrophic damage to the board itself, such that you don't need to guess which one failed. In the case of the MSI 980 Ti, I think it's so common because they used an unusually thin outer copper layer.

You could probably remove those parts if you heat them for long enough with your hot air station, but you'll have to heat them for a considerable amount of time several minutes, at least, and that means they'll probably be destroyed, and you'd need new ones to put back on (with leaded solder) to get the card working again.
 
I have 2 hot-air stations
One Atten 850A+
Other not know name,I put max temp,lcd shows 480c but with thermometer on multimeter shows 330c max that's why can't melt unleaded solder
My job is car electronic and not had problem,but with laptops and graphic card must get better hot air station

Thanks for Your time

P.S. I follow one servicer on YouTube, he repair laptops, graphic card and never seen he use preheater, he has Quick 861DW hot air station.
Also, can you tell me what kind of preheater,I see there are with plate and heater, other with hot air.
 
Last edited:
One tips works more easy to remove fets,I use solder paste on fet and can remove.
Here us image where i gave shorts, 0.8R
On vin pin of IR3553M is short,on top 4,botom is ok
Is this meant all 4 chips bad??? Or those 4 top are in parallel, bottom one is ok,i guess it takes 12v from pci slot
Smd caps from back removed but short still present
 

Attachments

  • pcb.jpg
    pcb.jpg
    74.5 KB · Views: 1
Last edited:
P.S. I follow one servicer on YouTube, he repair laptops, graphic card and never seen he use preheater, he has Quick 861DW hot air station.
Also, can you tell me what kind of preheater,I see there are with plate and heater, other with hot air.
I like the radiant ones, myself, like the Aoyue 866. I've never seen one that used forced hot air like a heat gun. The hot plate type ones probably wouldn't work that well for graphics cards.

I've repaired a few graphics cards using just hot air, and smaller cards like the 470 may be a little easier to do that way, but the best way to do this is with a preheater. Especially on bigger, more powerful GPU designs, there's so much copper in the core VRM area that you have to use a ton of heat for a long period of time to get it hot enough to melt. A preheater gets you around this.

What you're seeing is that all of VRM phases are connected to each other, hence why the resistance is the same on each phase. You probably only have one of the power stage ICs that's shorted, but if current injection didn't work to figure out which one it is, then you'll probably need to just start removing them until you find the one where removing it cleared the short.
 
  • Like
Reactions: Ziko
like this
I found 3th IR3553M from top caused short,now not short on 12v,but when measuring resistance on vcore coils to gnd its very low 0.6R
Is that mean gpu also shorted,ir how to test
I bring 0.7v on vcore coils and draw 900mA
 
0.6 ohms is a bit low for a GPU this small, but not necessarily a short. Now that you've cleared the short, the card should probably work, but I'd advise replacing any components you removed first.
 
Ahh, now see other problem
After put leaded solder and clear with wick I see one pin missing (GATEL) Tried to clear solder and now all connected like its one same pad
 

Attachments

  • Screenshot_20210915-155248_Gallery.jpg
    Screenshot_20210915-155248_Gallery.jpg
    433.5 KB · Views: 1
Yes china's amtech, that first pins are shorted to ground but from datasheet also can be both pins to gnd,if someone has this board to check for me
Pins are csin- and csin+

Is ani chance to fix this,to scrach or what?
 
Use more flux - get better flux if you need to - and a soldering iron to separate the pins.

Clean the tip, heat each pad and soak up the solder, and then clean the tip. Repeat a bunch of times, being careful not to overheat the pads.

As I recall, Amtech flux is supposed to be made in New Jersey. If what you have says it's made in China, it may be counterfeit.
 
All merged pins are so smooth like I dud with paper sand, not helping adding solder and flux,they are so flat like glued with sokder
 
The big pads in the center are probably supposed to be connected, since they're both ground. It's pretty common to see that pads for pins that do the same thing are actually connected like that, and then the coating on the board is used to make them look separate, and get the solder to make nice little bumps.

I'm not sure what else to tell you besides "solder better." You should be able to dilute the lead-free solder with leaded solder and eventually clean it all off. You may need to use more heat as well. If you're not getting the old solder hot enough to actually melt, it won't help, and you've got a pretty big glob there.
 
I think as you says,maybe by heating i removed coating, but need help if someone has this card to check for me pins marked red and green,are they connected to GND
First 2 pins from corner
TNX
 

Attachments

  • arrows.jpg
    arrows.jpg
    265 KB · Views: 1
Last edited:
All these 4 pins by datasheet ir3553
REFIN,IOUT,CSYN- and CSYN+ are connected to ground if current sens amplifier not used
So maybe all to gnd
Also i bring 0.7v to coil for gpu,and it draws 1A5
 
I tried with 2 fases to test but no vcccore, I found no ENable pin to driver ir3567b,I bring 3v3 directly to EN pin and then vcc cire is 1.0v
May I short 3v3 to Enable pin?

Tnx
 
I repair card detected rx470 but when I try furmark, pc goes off and get message from image.
After second try card dead,ir3553 that I change dead again. Just to ask,card has 5x IR3553M, I used 4x because one is damaged pins on board. Can be reason less power to gpu to crash? Actually I used only 4 phases
 
Back
Top