This demonstration showed real-time video decoding using ARM Neon vector instructions on an Apple CPU at an unspecified frame rate (presumably lower than that achieved on the x86 CPU) and real-time video decoding at 1080p/24 fps, using AVX2 vector instructions on an x86 CPU.

While this is not an impressive performance (compared to 4k/60fps or better for codecs with HW suport), it is still much, much faster than the early AV1 decoders.

I believe that this was precisely the intention of this demonstration, to show that AV2 is already much more mature performance-wise than AV1 ever was in a similar point of its lifetime, as the main early complaint against AV1 was its slowness in the absence of dedicated hardware support.