So ever wondered what the AIO of the new R9 Fury X looks like? Honestly – I didn’t. I just wanted to include the R9 Fury X into my own water cooling loop inside the EVGA Hadron Hydro. If you don’t know the case, have a look here and understand why you probably have to do it to fit everything in there.
So I’ve found a teardown of CPU AIO watercoolers. My idea was to get the impeller out of the pump, hook up some tubes and that’s about it.
And I thought – it’s from Cooler Master, so it won’t be much different – will it? Oh I was wrong, very wrong.
1. As pointed out by anddill some things were misinterpreted by me, I fixed that. Mainly: It’s not an IC under the flex PCB but the connector (obviously) and the black stuff is not for sealing the pump but for fixating the windings of the stator.
2. My card still lives, I killed my motherboard.
3. It seems like my custom tubes are leaking near the pump entry – I don’t know if I can fix that, but I’d strongly advice against doing it as I did.
As I took this pictures during my build, I didn’t spent a lot of time to get perfect pictures, sorry about that. And just as a word of warning: don’t do this, there is nothing to gain…
Let’s get started with something everyone should have seen by now:
So you just remove 4 screws with a little hex key and you can see the internals in all their glory, not much to see here. To get further down, you remove all the cables, one for the RADEON illumination, one for the fan and one for the pump. Next you turn around the card and remove the backplate with a small phillips screwdriver (size 0 if I remember correctly). And you will see this:
Alright, to get further down you have to remove the four screws of the spring holding the cooler on the die, all the screws on the back and also two screws on the bracket to release the frame of the cooler. Then you get to the mighty?, well surely enormous, die of the Fury. Looks like they tried fitting 8 stacks of HBM on there – at least there is some space left, but maybe that is needed for routing all the signals into the chip.
Taking the cooler further apart requires, I guess, criminal energy. They have used one way slotted screws to hold the cover on top of the pump and some nasty triangle shaped ones (I do not tamper with devices often these days, otherwise I would have had a correct bit for it I guess).
So the one way screws were no match for my laser-tipped Wera slot drivers, just apply enough pressure and they will come out again. So here is what’s below:
As you can see, they have soldered the PCB with the electronics directly onto the motor. The labels indicate a I²C communication to the main PCB and also some alert output for whatever purpose, what you don’t see yet is the flex electronics which goes down to the copper cooling plate – I suspect some sort of additional temperature sensor, and I have no idea why they would bother doing that – except maybe the temperature everyone is reading isn’t actually the core but only the copper plate, that would make the temperatures look much better than they are.
Next up is of course getting out those nasty triangle screws to see what the cooler looks from the inside. But wait, there is still water in the system. So first I have to get that out – trying to remove a tube without destroying it proved to be a fruitless effort, so I cut one near the radiator. A neon colored liquid came out of it – surely no simple water. Actually if you do some research, the R9 Fury X user guide from AMD suggests that it is some sort of „a mixture of monoethylene glycol / ethanediol“. Well I won’t drink it and you also shouldn’t get it on your hands either. A note maybe, the bowl I used is a very old one and the emaill has chipped off at some places, the liquid is (as it should be) clear and nice.
Okay now, ready to remove the triangle screws. As written, I don’t have a bit at hand for them, my brother thought that with a right sized hex key it should work, but none of them had the right size. So I used a Torx size 5 bit to remove four screws, then the bit broke. So for my next try I was lucky enough to have a true round bit from Hazet, it was marked as a PH2 bit, but is was really round, they forgot something during production I guess. What I did next was to use my trusty Proxxon mini multi tool and grind a triangle bit out of the round one. It worked reasonably well, if you turn the bit two hex sides between the steps you can get quite good triangles out of it. So with a self made 1.5mm triangle bit and a lot of pressure (yet again) I was able to remove all of the screws (I felt a bit proud). Out comes this:
It’s a rather simple design, very thin fins a small recess in the middle where the water is pressed in, around the fins a small groove is milled into the copper so the water can get to the exit.
On the other side you get the pump block from the bottom side:
I’ve left in all the sealing rubbers. One is as expected a square one around the whole block. The other one covers the top of the fins, so that the water has to go through the fins and can’t take the easy way out. The slot in the middle is where the water is pressed in and the small rectangle at the bottom is where the water leaves the pump block. You can also see the flex PCB, which I suspect to be a temperature sensor.
I haven’t found any way to open up the block any further – I have no idea what they have done, maybe it is welded together. To complete the tear down I tried to get the PCB off of the top side. This turned out to be a bad idea (I think). After using a hot air gun and some force with the screw driver the PCB came off – in a bad way. Seems like the motor got so hot that in the end I pulled the contacts out of the motor and not out of the PCB.
So what’s there to see? Based on the research done by anddill over at the 3DCenter forum, the winding is loosely put into a hole which is then filled with some sort of glueish stuff to hold the winding in place. Failing to do so reliably which is also a cause for the noise and whining people are hearing from their pumps. Anyways I probably just have killed my pump.
What’s the PCB looking like? Not much to see on the bottom:
You can see the pins I pulled from the motor, and the mounting of the flex PCB – there is also an IC under that black blob where the flex PCB is mounted.
Yeah and that is that. No way to remove the impeller from the pump, well I didn’t want to go to the extreme level and drill out the motor as I want to have it sealed. My suggestion for anyone trying to get the stock cooler integrated into a custom loop is to use the pump of the Fury as pump for the loop.
The good news is, that you can easily fit new tubes to the cooler:
Standard 8/10 mm tubes will fit just on the stock tube mounts, use some tube mounting things of which I don’t even remotely know how they are called in English and you are done. It is sealed tight and no water came out. Until I ran a burn in test with furmark and everything got hot, then it started to leak, so I advice against doing this – maybe I can get something better.
So big question: Does it still work? I don’t know, after finishing my build – which looked really nice, my machine wouldn’t start. It just shut itself off in 3 seconds. Now with everything removed my mainboard cycles on / off by itself every 3 seconds. Not sure if my Fury killed my mainboard because I killed the Fury or if I killed my mainboard during the installation of the water cooler. Surely I have killed something here. (*sadface*)
As a small update on this: I killed my mainboard, the card is working fine, even if you leave the pump electronics unconnected (good to know for other coolers like from EK).
Now it’s looking like this:
Temps are horrible, especially my CPU is burning (literally @100°C), I’ve probably used too little pressure after killing my first board with too much pressure. And of course a custom build inside the Hadron is a bad idea as well (another story maybe).