How likely is it my card is beyond repair?

[Goose]

I have a Vega 64 reference card that I installed a Byskski water block to for my first water cooled system and everything went fine to begin with. I had the system up and running for about 2 weeks with everything working just fine, temps were good. I decided i wanted to rearrange my loop, largely for aesthetics but also for better access, draining etc

.

When i did this i also removed the block from GPU in order to clean it out as i could see some debris in it (it was acetal from the threads on my EK pump/res combo). I reassembled the block and pressure tested it to make sure it was air tight and it was. I reassembled my loop, pressure tested it and then leak tested for 12 hours just to be sure. There were no leaks and everything seemed fine.



When i tried to boot the system for the first time i was getting no display at all - i tried with different monitors, DP cable, HDMI cable, different power supply - nothing. So once again i drained the loop, removed the graphics card, removed the block, reattached the stock air cooler and then tried again to boot (in a different system).



I still got no display, the fan would not spin but the 'Radion' logo on the cooler DOES light up. I have no idea what i could have done to kill the card but since the logo does light up it's clearly getting some power. I appreciate this is a hard question to answer but if i send it off to a company that repairs GPUs what do you think my chances are of it being fixed?



Thanks

 

[Poneros]

Could be that your PSU gave out, got another GPU to test with?

 

[Goose]

Could be that your PSU gave out, got another GPU to test with?

I tried the GPU in a whole other PC so i'm 99.9% sure its the card thats borked.

 

[tyler_jrb]

Depends on what’s wrong with the card I guess.

I’d say with what you have said it probably has a reasonable chance of being repaired.

There are basic tests with a multimeter you can do if you have one to figure out what is wrong.

It becomes a pain and a real slim chance when a component blows completely and blows a hole through the pcb.

Everything else on the pcb can be diagnosed and replaced pretty easily by someone who knows what they are doing, has the right tools etc.

 

[Goose]

If it was a cracked die from too much pressure would that be visible to the human eye?

 

[tyler_jrb]

If it was a cracked die from too much pressure would that be visible to the human eye?

If it cracked you would notice as it’s like cracking a piece of glass rather than just a clean snap. You would see it in the reflections on the die pretty easily. It does take a lot of force to break one as well.

Likely more than any waterblock would do just because of how they mount and the amount of warpage in the pcb.

 

[Kaapstad]

If it cracked you would notice as it’s like cracking a piece of glass rather than just a clean snap. You would see it in the reflections on the die pretty easily. It does take a lot of force to break one as well.

Likely more than any waterblock would do just because of how they mount and the amount of warpage in the pcb.

It is not difficult to damage cards using HBM where it joins to the GPU core when fitting waterblocks. Unfortunately you can not see the damage and the first thing you know about it is when you start the PC and get a black screen or massive artefacts.

 

[Goose]

It is not difficult to damage cards using HBM where it joins to the GPU core when fitting waterblocks. Unfortunately you can not see the damage and the first thing you know about it is when you start the PC and get a black screen or massive artefacts.

Would that cause the fan on the cooler to stop working though?

 

[tyler_jrb]

Would that cause the fan on the cooler to stop working though?

Could do if the gpu die itself was internally damaged and or shorted. Any short on the 12v lines will stop it I’d have thought. I’ve seen that before with a shorted gpu chip.

But it’s hard to say without a meter, as that would show resistance values and what voltages are where when you power the card.

It is not difficult to damage cards using HBM where it joins to the GPU core when fitting waterblocks. Unfortunately you can not see the damage and the first thing you know about it is when you start the PC and get a black screen or massive artefacts.

You are probably right being a hbm based card I completely forgot. I’d imagine they can be quite fragile.

 

[Vince]

I have a Vega 64 reference card that I installed a Byskski water block to for my first water cooled system and everything went fine to begin with. I had the system up and running for about 2 weeks with everything working just fine, temps were good. I decided i wanted to rearrange my loop, largely for aesthetics but also for better access, draining etc

.

When i did this i also removed the block from GPU in order to clean it out as i could see some debris in it (it was acetal from the threads on my EK pump/res combo). I reassembled the block and pressure tested it to make sure it was air tight and it was. I reassembled my loop, pressure tested it and then leak tested for 12 hours just to be sure. There were no leaks and everything seemed fine.



When i tried to boot the system for the first time i was getting no display at all - i tried with different monitors, DP cable, HDMI cable, different power supply - nothing. So once again i drained the loop, removed the graphics card, removed the block, reattached the stock air cooler and then tried again to boot (in a different system).



I still got no display, the fan would not spin but the 'Radion' logo on the cooler DOES light up. I have no idea what i could have done to kill the card but since the logo does light up it's clearly getting some power. I appreciate this is a hard question to answer but if i send it off to a company that repairs GPUs what do you think my chances are of it being fixed?



Thanks

Got a multimeter? Eli Tech on YouTube shows how to troubleshoot Vega in a video he did. No display is not always the chip so you could get lucky.

 

[Goose]

Got a multimeter? Eli Tech on YouTube shows how to troubleshoot Vega in a video he did. No display is not always the chip so you could get lucky.

I dont and to be honest knowing nothing about electronics i dont think i'd be able to achieve much with one anyway, even after watching the Eli Tech video.

I've phoned three companies this morning to ask about repairs - one has recently stopped doing GPU repairs, one said "buy a new one" and the other said all they do is a reflow and nothing else.

I'm thinking that i'm out of luck here. This was my first ever custom loop and i've managed to do the one thing i wanted to avoid which is kill the gpu. I think i'm going to have to sell all my water cooling gear i've bought, no doubt at a huge loss, and stick to air cooling. I cant risk killing another card.

 

[Vince]

I dont and to be honest knowing nothing about electronics i dont think i'd be able to achieve much with one anyway, even after watching the Eli Tech video.

I've phoned three companies this morning to ask about repairs - one has recently stopped doing GPU repairs, one said "buy a new one" and the other said all they do is a reflow and nothing else.

I'm thinking that i'm out of luck here. This was my first ever custom loop and i've managed to do the one thing i wanted to avoid which is kill the gpu. I think i'm going to have to sell all my water cooling gear i've bought, no doubt at a huge loss, and stick to air cooling. I cant risk killing another card.

If you're totally stuck I've got a thread in GD on board repair. I'm no expert but think I'd be able to say if it's a dead phase (power related) or if the core is a gonner. The most common problem on these tend to be shorts on drivers for power phases.

Question does your one have a moulded or non moulded core?

 

[Goose]

If you're totally stuck I've got a thread in GD on board repair. I'm no expert but think I'd be able to say if it's a dead phase (power related) or if the core is a gonner. The most common problem on these tend to be shorts on drivers for power phases.

Question does your one have a moulded or non moulded core?

I'm afraid i wouldn't know the difference but it's a reference board.

 

[Vince]

I'm afraid i wouldn't know the difference but it's a reference board.

There are 3 separate core types on the reference design (depending on where it was assembled). The non moulded is the one where you could potentially destroy the core pretty easy as there isn't a layer of epoxy between the core and hbm. Only way to find out is to strip it down and see.

A moulded core will look like this... my old Vega...



A non moulded one will have gaps between the chips. The moulded ones are very difficult to break by over tightening.

 

[Goose]

@Vince

I have a picture of mine in another thread here - Could some help me out with thermal pad placement on Vega 64?

 

[Vince]

Ok it looks non-moulded to me. They, unlike the moulded core, dont like over tightening at all. Still you cant be sure it is that at all until you attack it with a multimeter to see if all of the phases come up.

 

[Goose]

Ok it looks non-moulded to me. They, unlike the moulded core, dont like over tightening at all. Still you cant be sure it is that at all until you attack it with a multimeter to see if all of the phases come up.

Sorry for being thick but the picture you posted of a moulded core in post #14 to my eyes looks the same as my chip with a gap between the core and the HBM chips?

 

[Vince]

Sorry for being thick but the picture you posted of a moulded core in post #14 to my eyes looks the same as my chip with a gap between the core and the HBM chips?

Sorry dude perhaps i didnt explain, one has a gap and on the other the gap is filled with a resin as per the below:



Ignore the designation which tells you what hbm and what vega it is for. my 56 had a resin moulded core with samsung hbm. I abused my 56 trying to put a broken aio on it and the resin core can take abuse. I mean i'm pretty sure that you wouldn't have damaged it kind of abuse (this is coming from first hand experience with many vega cards).

I cant be 100% sure from your pics but if it's a resin core id sway towards more repairable than if it's not a resin core as in my experience the resin ones are not what I would call fragile scale of 1 to 10, 10 being least fragile, they would be a 9. The resinless ones are like a 2 and can be very fragile.

For me id at least take some readings off of the card with a multimeter to see what it's doing and to see what the resistance is to the core and memory, with just a couple of readings you should be able to know if the core is bad or not.

 

[Goose]

@Vince

Mine definitely has resin between the core and the hbm chips, it looks just like your one. The picture i took of mine that i linked you to earlier was very poor quality so maybe it looks like there is no resin?

How expensive a multimeter would I need to buy and would you be able to guide me on how to actually check my card?

Thanks

 

[Vince]

@Vince

Mine definitely has resin between the core and the hbm chips, it looks just like your one. The picture i took of mine that i linked you to earlier was very poor quality so maybe it looks like there is no resin?

How expensive a multimeter would I need to buy and would you be able to guide me on how to actually check my card?

Thanks

You could probably get away with even a cheap 15 quid multimeter or perhaps even cheaper. I could also probably quite easily guide you through taking core and memory resistance after that though when you know the core is good/bad what next? Your likely going to be tracking a short and will have to probably replace at least one smd component. To find the short you will need a desktop power supply, then you need the right tools to replace the smd. Some spare GPU's to borrow parts off of etc etc etc...