On Sat, 2 Sep 2023 16:38:31 +1200, I wrote:
He says in the video that he was going to spend a week trying to do all his computing on RISC-V machines, with a report on the results due out on Sunday UK time (I suppose Monday our time).
That video is now out <https://www.youtube.com/watch?v=Na6PT4npsMg>. He was using about two different RISC-V boards, and to start with, had a mix of problems with video playback: one browser would do video smoothly but not audio, another one would do audio but not video, luckily the third try managed both OK (at 720p video resolution). So he was able to spend a Sunday evening catching up on YouTube videos, as he normally does. And also some Paramount Plus streaming service as well. LibreOffice worked fine, and so did Google Docs. But he normally had YubiKey two-factor authentication enabled on the latter, which he had to turn off for this experiment because the key simply would not work. Other things that would not work were Kdenlive (video editing), and Blender (3D/compositing) — both threw up errors trying to open a suitable OpenGL context. But he was able to compile and run SuperTuxKart. Finally, for a performance comparison, he ran the multithreaded GIMP “Lava” filter on a blank 1920×1080 canvas; this took 33 seconds on the Lichee Pi 4A and 67 seconds on the VisionFive 2, compared with 47 seconds on a Raspberry Pi 4. Worth noting that both boards are still only considered “development-quality”, not “consumer-quality” (yet). All in all, desktop/consumer-oriented RISC-V has progressed by leaps and bounds over the last 18 months. Would be fun to repeat the experiment in another year or so ...