Computation time and cores

Is there an internal limit on the number of cores in EE8.5 and 10.2? Because I found that my model reached maximum speed and take the shortest computation time only with 16 cores, while my computer has 64 avaliable cores with128-thread (see figure). When using greater or less than 16 cores, the computation time will significantly decreases to some extent.

There is no internal limit to the number of cores that may be used in EFDC+. However, as you can see from the testing we did in earlier versions of EFDC+, beyond 6 cores there is no appreciable gain in speed.
https://eemodelingsystem.atlassian.net/wiki/spaces/EEREF/pages/107020295/Speed+Gains+with+OMP
Depending on your model configuration it is likely you will need to use MPI to further increase in run speed.
https://eemodelingsystem.atlassian.net/wiki/spaces/EK/pages/1095270402/Guiding+Principles+for+Running+MPI

We discussed with some of our technical team, and one cause of your issue could be CPU Parking. This is where Windows OS restricts the number of cores being used to save power. This is not related to Intel or EFDC. You can see some options for removing this limitation here:
https://www.coderbag.com/programming-c/disable-cpu-core-parking-utility