Not that wimpy. I've been using the manufacturer's EVB for this chip since March and it goes really well. Streets ahead of anything else in the RISC-v world at

Not that wimpy. I've been using the manufacturer's EVB for this chip since March and it goes really well. Streets ahead of anything else in the RISC-v world at the moment. And dual- and quad-socket boards are coming soon. Even on things that don't parallelise well it's still pretty good. For example on building riscv-gnu-toolchain (binutils, gcc, libc) for which the stupid GNU recursive makefile buildsystem means it only averages using about 9-10 of those 64 cores, it takes 34 minutes. This compares to 11 minutes on a 32 core Threadripper, 20 minutes on a 6 core Zen 2 laptop, 32 minutes on a Thinkpad X1 Carbon with i7-8650u, and 102 minutes on a Pi4. On things that use all the cores, it's comparable to a 12-16 core current generation x86 -- which costs the same or more. It's a lot powerful than the Graviton 1 that became available on AWS in November 2018. The vector unit work great, that's been my main reason for testing one. There will be much much faster RISC-V SoCs with lots of cores in the next couple of years, but this lets you get software ported on a realistic machine right now.