Lawrence Livermore National Laboratory has always been one of my heroes when it comes to developing advanced computing technology using some out-of-the-box thinking. One of their departments focuses on highly scalable, highly parallelizable programming, often for the purposes of particle interaction simulation.
But it's always been a challenge to get time on one of their really huge parallel boxes to even learn how to deploy your code on such a beast. After all, learning requires failure and went every minute one of the big boys is turned on they eat 1.21 gigawatts of power and get ready to travel back to the future, it's a pricey proposition to put your code out there with the expectation that it's going to fail.
Enter the BitScope Cluster Module.
I know what you're thinking, crypto-folk: "Can it mine Bitcoin?" I'm sure it can; after all, it's running a modified Linux kernel, pumping four glorious 1.2 GHz CPUs, a mostly competent graphics processor, and costs 1/10 or less of what one Nvidia 1070 would run you.
Lots of people have had this idea before at smaller scales, of course.
Let's be honest, we never had the expectation that even nearly 1000 nodes of Raspberry Pi power was going to make a dent in the highly specialized field of Bitcoin and altcoin mining, right?
Imagine the other possibilities for what is nearly desktop size highly parallel supercomputing. Imagine what you could do with your own local neural network server if the field of antagonistic neural networks tickles your pickle. With the right switches, imagine what kind of web serving throughput that you could get out of one of these things with the proper Elixir configuration. Imagine how much you can learn about distributed databases on your own.
$20,000 is literally a small price to pay for a midrange supercomputer under your desk. Not just a midrange supercomputer, but one which consumes a perfectly reasonable amount of power. In fact, at 4000 W max it consumes less power than a couple of high-end gaming desktops (and certainly less than four). That's a pretty useful trade-off.
If you've always wanted a dream machine, maybe it's time to give a few nocturnal thoughts to Lawrence Livermore's supercomputer in a box.