This is a very good idea and I am surprised that your post was viewed 100 times before me and no one post a comment.
I can take the h with application to centrifugal pump flow sometimes at the end of the month. I am with 12x2=24 CPU and 24 GB RAM.
Regarding your test. How many physical CPU did you use? - If only one and two the speed-up with hyperthreading should be higher but if you use 4 or 6 then a speed op of a couple of % should be expected.
I have contacted my VAR for input on what would be a suitable benchmark. I am not sure if different simulation features (i.e. radiation, fans, porous media, thermo-electrics, etc) impact the solution time differently. I do not want to propose a simulation that inaccurately reflects the potential time savings.
That improvement of 2% was using 1 core of my i7 9650 with and without hyperthreading. Using all 4 cores was a 12.7% improvement over 1 core (both w/o hyperthreading) (I always forget to increase its utilization from default 1CPU, hence this number came later).
Here's a link with a little history relevent to your question. Take a look near the bottom for Ian Hogg's posting with a comparison chart.