Towards the end of the video I show a profile of the sample music. For comparison, here's a single frame of the UI on a release build. The UI is locked at 60fps; so can't go faster than that, but you can also see a couple of audio frames on this screenshot (which don't necessarily align inside frames), along with the FFT thread at the bottom. Here 512 frames of audio is being generated in about 1.2ms - this is the equivalent of about 11ms of audio. Although I keep an eye on this stuff, and build things sensibly to avoid bottlenecks, I haven't actually spent too much time optimising anything yet. But it's good to know there is headroom. I also forgot to mention in the video that one of the main things the profiler brings to the table is debugging; it already showed me a place where I'm doing unexpected work, and a mutex lock that should not be happening. For the small effort it took to build this feature, I think it was well worth it.