High Performance computing and parallel computing amateur linux test lab

I dabble in trying to build high performance applications, and parallel computing stuff. I don't do a good job at it, but I would like more practice.   I tried to conceive of the type of hardware I would like for my continued practice. These are their stories:

I know from bitcoin mining that some instruction sets are better at somethings, so I'm pretty sure I want two servers, with dual graphics cards. One nvidia, and one ATI (or maybe I'm just an ATI guy but nvidia is kind of sort of blowing them out of the water). I also know I'm probably going to want two Parallella's.

So to choose my gpu's I'm probably going to head over to phoronix, and take a look at some of their break downs. So I'd probably want to get two Radeon R9 290's for my ati box, unless I wanted to pretend it was a little ways back when I was making bank bitcoin mining, then I'd get a couple of 6950's. And for my nvidia box I would probably get two GTX 680s.

I actually have a parallella right now and I highly reccomend them. The defualt os for parallela is ubuntu. You know in case you need to like hashcat something with like 16 cores.

Ok so if your getting into high performance hardware you also want to make sure your coding to use it. With the graphics cards you can go with the universal opencl or the nvidia specific CUDA. For parallel programming you want to go with, erlang most likely, or GO.