Preparing to watercool this "mining-special" MI50. The graphite thermal pad is a nightmare, extremely difficult to fully remove, and you get conductive dust everywhere. I cleaned the package with a brush, IPA and compressed air, hopefully the GPU isn't going to explode after I reapply power... :woozy_baa:
@niconiconi How did you come up to develop a scientific application at home, when you could do it on a server or a supercomputer?
@niconiconi I mean, why foot the bill for research yourself? 
@fulkrum@qoto.org Unfortunately nobody else seems to be interested in advancing this particular area of free and open source electronics design automation, so obvious answer is coding it yourself...
@niconiconi in my last job, I worked on open hardware and we were doing open designs for some of our products. We were working on a under the desk form factor - the specs are actually on github. (I didn't do the work, a friend/colleague did) Sadly the company was sold to another company and they shut down our division.
@fulkrum@qoto.org It's more like HPTC than HPC, and most of the time there's no "HP". It certainly can be used for research and large-scale simulations, but most problems are daily engineering problems and they are solved on a single desktop PC.