So it's a http2 vs http1 benchmark. Not a http vs https benchmark?

Right. HTTP/2 is only available with TLS, due to browser policies, so it necessarily requires TLS.

This means you need HTTPS if you want the performance wins of HTTP/2.

This doesn't mean that HTTPS is a performance win if you're still using HTTP/1.1. However, the "performance overhead" for TLS is negligible in practice.

Yes, so there are two major changes you'd need to get this performance benefit, not one, and if you just make the TLS change, your performance will (probably marginally) suffer, not improve. It's a bogus comparison.

Didn't you know that air fresheners make cars go faster? Just look at my Ferrari F450 with an freshener vs this Nissan Micra without one!