For houdini I extended my tests and used multiple setups with the help of the great community of Odforce, the results can be found here.
To what degree our 3d packages are multi threaded, is adding a mighty 56 thread machine really worth it. Okay I thought of sharing my experience and results with you guys comparing :
1- i7 6800K 6/12 @ 3.8GHz
2- 2 x E5-2683v3 28/56 @2.5
3- 2 x E5-2696v2 24/48 @ 3.25 (not official xeon version works with overclocking, I will not use it with all the tests)
Programs to be tested:
-Houdini (Pyro,Grain,ocean) Files
-Vray (official vray benchmark)
-Realflow (Dyverso, Hybrido)
Note: open cl and cuda weren't used and also viewport wasn't active while simming, trying to make it purely cpu test.
At first I must say that I was sooooo disappointed and shocked by the results, the i7 just kicked the dual xeon butt in most of the situations. Concluding that a fast core is much better than a multi slow ones, but that wasn't really very true, my problem was that I tested the machines with very simple and low res setups finding out that the dual xeons only shines when the res is increased and the things get complex.
I will start by talking about Houdini:
Pyro: I created a sphere and off the shelf fire, i7 was faster. I created 4 copies of it and still i7 wins until I cranked the pyro spacing and made it 0.02 instead of 0.1 and that's when the xeons just doubled the i7 performance.
Grain: same thing xeons only shines with high res sims
Ocean: for flip the story was a bit different, increasing the res only brought them close to each other and cpu utilization was fluctuating compared to 90-100% with the other tests.
I used the official vray benchmark and xeons easly win the competition. In real life scenarios it depends on how much of your frame is covered with geometry, I had a project with particle rendering (particle cached, high sub sampling and most of the frame was empty) in this case both machines were the same rendering speed.
Command line simulation was used here, and increasing the resolution didnt benefit the xeon in this case except when I used Dyverso (Granular) particles. So if you want to use Dyverso liquids a fast i7 and some cuda devices will make your system fly.
Below is the results chart (time in seconds)
In the following chart I added the third processor results to the mix.
There is no perfect solution were you will have the perfect machine that will excel at everything. So having a fast i7 is so great for everyday work, viewport performance, testing, particle caching and when working with Cuda because GPU mostly benefits a fast core to feed it the data according to Linustech (Cores vs clockspeed for video encoding).
So the middle ground here if you can get a xeon system running all cores at 3GHz or above that will be perfect like the 3rd option we had above.