Check out Volunteer your Raspberry Pi to IBM’s World Community Grid, an article in the Raspberry Pi blog.

---

Check out Volunteer your Raspberry Pi to IBM’s World Community Grid, an article in the Raspberry Pi blog.

The pi is very inefficient for crunching data.
It's not only very slow (being an arm architecture), but it's a 40 to 28nm design, which is about 2 to 4 times bigger than other cpus on the market today.

For a while I toyed with the idea of making a pi cluster, but found AMLogic TV boxes to be faster, and yet still super slow compared to a 1.7Ghz X86 Atomic pi.
While the atomic pi is 14nm, it's price and ppd/watt comes very close to a 7nm ryzen 3000 series system.
It's just not as fast, but can easily be doubled, quadrupled, ... to match the performance of a ryzen system, while still staying below the tdp and cost of such system.

I would totally be interested in running boinc on the pi architecture, if their design was at least 14nm (or smaller), and had quadruple the cpu cores (at least 16), but preferably 100 CPU cores in a board or more.

A fitting arm server for boinc, could easily run 300 cpu cores, and still fall within normal cpu tdp ratings, technically limiting itself to water cooling systems starting from 400 to 600 cpu cores, depending on the design.