It's all the same silicon. Threadripper systems have a lot more I/O than Ryzen systems, but Ryzens already have far more I/O than required in a compute node. Hence, a Threadripper system can never do more work at less power than a Ryzen system, if the latter is configured properly. (Exceptions: Workloads which are I/O bound, not compute bound.)
Edit: On second thought, never say never. More cores within the same power envelope will do more work, and therefore it may be possible to squeeze out somewhat more work even in a slightly smaller power envelope, when there are more cores per system. But the higher power consumption of the two fatter I/O dies of Threadripper systems (the one in the processor and the one in the southbridge) will make this hard to achieve.