GPU Part IV: Dr. Frankenstein's Laboratory

Previously when researching how to bend heatpipes, I found this advice pretty funny:
https://hardforum.com/threads/possible-to-bend-heatpipes.1191767/
What an idiot. That's the last thing I would ever do.
Oh the irony...

This is Plan C.
Right. So anticipating a face full of copper shrapnel, I brought out the second undamaged Accelero I had and heated the bracket until I could see the solder melting off. I was able to gently remove the bracket with pliers.


The heatpipes did not rupture. Actually they suffered no visible damage at all this time.

I did some light sanding to smooth out the melted solder.
Here is the first mockup. The position of the Accelero is ideal here, because it to the right of the case:

The heatpipe order here is not ideal though. There is no way for the last two pipes to be routed without more bending.
Either the long pipes have to be bent, or two of the Accelero's pipes will have to be bent, like this:

Here is how that went:

The heatpipe bender tool couldn't really be used here. Too cramped and the bend radius was too large to be useful. I used a round metal bar to help shape the curves, but really it was done by hand.

The last one was really tough to get in. I wanted it on the end, but there was no other way to do it without bending the pipe too tightly or having it stick out to the side or top where it will collide with other things. I did not want it in the center slots above the core... too risky to put only half of a heatpipe there.
Some of the bends look a bit sharp, but the diameter is above 4mm for the most part, so they are still functional.

Preparing to put it together, another problem surfaced.
In the CAD model, I only cared if things fit but not how they fit together, and I hadn't given any consideration to assembly.
First thing, the blocks are not able to create any clamping pressure without drilling out one side. I drilled out the top block.

Then, there wasn't an obvious way to attach the blocks to the card - The Accelero's fins block the screw holes on one side, and also the GTX 1080 blocks them from the other side. So I switched to machine screws and then countersunk the holes so it would not cause a clearance issue against the GTX 1080.

And, because the card is being mounted upside down in the case, unlike the HDplex H5, it is not possible to adjust heatpipe positioning after the assembly is mounted to the card because the screws are inaccessible.
I used the existing heatpipes as a guide for which to get the assembly positioned without needing to mount the card first.

The other block could be mounted now.

This fun problem was discovered next while checking alignment:


The PCIe riser was slightly blocked by one of the heatpipes.
Then this fun problem:

I had shims under the Accelero to support the weight and they were too high. The cooling assembly is about 40mm thick, but the card height is 43.5mm from the floor of the case and the GPU silicon package is about 2mm thick, so I made some 1.5mm shims to place under the Accelero to keep the entire card supported and not stress the PCB.
That was the last of the mounting problems. Moving on

I used Liquid Ultra this time:

And the card was assembled.. Here are a few last shots before it went in:




All rigged up:
