Ray tracing is a light propagation model that treats light as rays (a stream of photons). So the algo itself is single-threaded and built around relatively simple operations.
Now, rendering is a task of running a ray tracing method many times independently. These runs can be done in parallel, as you said.
But the original issue was about poor HT effectiveness in this benchmark. I won't really comment on whether it's actually that bad (there's not enough data yet). But I wouldn't be surprised.
Remember how HT (SMT) works. A core in your CPU can hold and execute 2 processes at the same time... up to a point when they need a core element that's singular (e.g. ALU).
Some tasks leave a lot of computation potential unused (because the CPU spends more time on just working with data and communicating outside).
Ray tracing is very intense computationally.
I've seen a few tests showing that rendering doesn't benefit from HT as much as some other applications do.
Here's an example (quite extreme, as the site name suggests):
https://www.extremetech.com/computi...e-effects-of-hyper-threading-software-updates
You should be able to find a more recent one to check if things improved drastically since 2012 (but they haven't ;-)).
