Need help diagnosing AIO and gpu temperature issues.

[Bread Pitt]

Tldr: I fear I may have broken my brand new AIO, my old GPU, or both slightly. And I don't know where I went wrong.

The story:

Spoiler

Due to [the world we live in] a new GPU as outside of my budget for the foreseeable future. I am sure this is not a new story. I've recently become more interested in consuming tech/computer related media and that's put me on a track to get some new life out of older parts.

I own an RX 480, could be worse, could be a lot better. It's a reference design too and I only now realise how bad the cooling is on it. A good clean and new paste brought my gaming down from a toasty ~92 to ~88 degrees. Eventually I struck inspiration and figured I should change the cooling block on my card. So I bought a Kraken G12 kit and a Fractal Celsius+ 360 AIO on sale. I could even repurpose the AIO on my CPU come the time I am able to buy a new GPU.
The surgery was a bit more involved than planned. The Kraken kit was designed for bigger cards so I had to keep the blower fan to get some air directed towards the gddr and vrm. Some work with a hacksaw later and Frankenstein was complete. I was now idling on ~6-7 degrees over ambient, and the toughest gaming load was heating me up to 45 max. Success!

I bought a little power cable adapter (CRJ I think it's called, to standard pwn fan) so I could let the GPU set a fan curve for the AIO. It was running on the pump header on the motherboard and that's a little noisy.
I was, however, worried that my little old RX 480 wouldn't be happy supplying power for a pump and 3 fans. So I figured I'd be smart and hook the fans up to the powered fan hub in my case and only use the built in unpowered hub on the Celsius to relay the pwn signal from the Pump to the fans. 

And that is where my problems appear: The AIO seems to be underperforming and the fans don't actually speed up. It'll fire up a game, but a more demanding one crashed it instantly. Mounting pressure seems fine, I can't figure it out. So I swap back to the setup that worked to see if I can replicate the error step by step. And now the pump won't turn on at all without being connected to the powered fan hub in some way, which is really weird. Even when the pump is on and I can hear it whine, no cooling is happening and the gpu reports wild temperature spikes in even the most mundane tasks. Doing anything remotely demanding, or even swapping between internet tabs too quickly, crashes my gpu.

- I'm not entirely sure that I didn't somehow damage my gpu. Nothing gets hot at all, to the touch. So if anything it might be hotspots on the chip that instantly cause protections to kick in? Or maybe it is reporting incorrectly? Any tips on diagnosis? (Both GPU-Z and Radeon center report the same temperatures and speeds, etc, if that means anything)
- Did connecting the AIO, indirectly, to the powered fan hub burn it out or something? How can I tell, could I have known?
- Next steps to take? I'm going to contact my vendor about the AIO, see if it falls under the guarantee. With a new one I reckon I can go back to the working setup but now I'm afraid it was just pure luck that it worked. Any thoughts on recreating the setup without it blowing itself up?

[Founders]

My guess would be your GPUs VRM is overheating. Do you see that heatsink towards the front of your card? That's your cards MOSFETs. If they don't have cool air over them they heat up quick under load. Your GPUs stock HSF has a shroud for a reason, it keeps air circulating over the fins of your GPUs heatsink and the heatsink on those MOSFETs. Another thing to note here, GPUs don't have an IHS, it's just direct transfer from your cooler to your GPUs die. You need to make sure that every part of your GPUs chip is covered with thermal paste. Any uncovered areas can lead to hotspots that can kill the GPU. My suggestion here is get rid of this aftermarket AIO and install your stock cooler back on with some Thermal Grizzly or any comparable thermal compound. Side note: That's a blower fan at the end of your card, without a shroud installed it does absolutely nothing. 

[Bread Pitt]

9 minutes ago, Founders said:

My guess would be your GPUs VRM is overheating. Do you see those heatsinks towards the front of your card? That's your cards MOSFETs. If they don't have cool air over them they heat up quick under load. Your GPUs stock HSF has a shroud for a reason, it keeps air circulating over the fins of your GPUs heatsink and the heatsink on those MOSFETs. Another thing to not here, GPUs don't have an IHS, it's just direct transfer from your cooler to your GPUs die. You need to make sure that every part of your GPUs chip is covered with thermal paste. Any uncovered areas can lead to hotspots that can kill the GPU. My suggestion here is get rid of this aftermarket AIO and install your stock cooler back on with some Thermal Grizzly or any comparable thermal compound. 

Ah, I covered all of those issues in the part I spoilered (cause it's a bit tldr and not part of the problem, maybe that was a poor decision).
Suffice to say, I already did a repaste previously and it didn't help much. 
I also did a 'touch' test on pretty much every part of the gpu as soon as the power was off, nothing is actually even remotely hot. (Except for the die itself I assume, but I can't touch that)
I kept the blower fan and scratchbuilt a new partial shroud to direct that air as best I could towards everything not covered by the AIO. I could've added a picture of that.
It had been running perfectly for a week or so, including under heavy load, and only the mentioned further changes started showing problems immediately.

Edit: I am going to put the stock heatsink back on just so it can run. But I'll need to get creative with the shroud as that did not end up surviving the transplant entirely....

Thanks for your reply though

[Founders]

5 minutes ago, Bread Pitt said:

Ah, I covered all of those issues in the part I spoilered (cause it's a bit tldr and not part of the problem, maybe that was a poor decision).
Suffice to say, I already did a repaste previously and it didn't help much.
I kept the blower fan and scratchbuilt a new partial shroud to direct that air as best I could towards everything not covered by the AIO. I could've added a picture of that.
It had been running perfectly for a week or so, including under heavy load, and only the mentioned further changes started showing problems immediately.

The only way to fully diagnose this issue is going back to the stock cooler to get your answer. 

• Install the stock cooler > If the card works you know it's something with the way you've configured the AIO or it's dead.
• Install the stock cooler > If the card no longer works then you know it didn't fair well with the AIO.

I'm sort of scratching my head on why you spend this much converting an RX480 to liquid cooler. If it was just a fun project fair enough, but the performance gain if any at all would be pretty marginal. 

[Bread Pitt]

Well, I did put the stock heatsink back on. The shroud is literally hanging together with duct tape...

It's alive, so that's a positive. Just the AIO bricked. But I still wonder why.

Going to have to look around for better tips on how to get a fan curve on the AIO fans. But maybe will just have to set something manually in bios and not have the gpu decide. 

Would still love some feedback on that, and if I somehow shorted, or burned out the AIO by connecting it to the powered fan hub. I had thought those connections would be one direction?