There are some Threadripper benchmarks on Geekbench - assuming they're legit, the results don't bode well. In multi-threaded performance, the 16 core 1950X is only about 20% faster than the 6 core Xeon in my current system;
(https://browser.primatelabs.com/v4/cpu/compare/3108993?baseline=2517429)
Perhaps this is a result of Geekbench failing to properly utilise so many cores? The single core result is pretty encouraging, given that it's pretty much the same as the Xeon at 3.6GHz with a 200MHz clock speed deficit.
The multicore benchmark is lower than a 1800x. I think an accurate deduction is that it is not representative of the final product and there is some serious issues with the software. The reason i say it is the software is because of the Blender demo they showed, where they had improved multi-core performance over the 1800x.