Whenever you use hardware over software you can expect significant performance increases. Have you had a chance to review our benchmarking page here:https://www.wolfssl.com/docs/benchmarks/
You can see for example on the STM32F2 (which has onboard acceleration like the STM32F437 but unlike the STM32F427):
Software Crypto: wolfCrypt Benchmark, Normal Big Integer Math Library
AES 1024 kB took 0.822 seconds, 1.22 MB/s
STM32F2 Hardware Crypto: wolfCrypt Benchmark, Normal Big Integer Math Library
AES 1024 kB took 0.105 seconds, 9.52 MB/s
Have you also considered using the STM32F437 that has on-board acceleration?
You asked specifically about setup time and resources freed up on the main MCU, If you were to use the ATECC508A AND you were using ECC based cipher suites then absolutely you could expect speed up in connection time and also reduced run-time memory as the ECC keys could be offloaded to the key store in the ATECC508A module.
wolfSSL has not yet added support for the ATSHA204 module but that is something we could do. If that is something you would like to see added I would encourage you to shoot an email to firstname.lastname@example.org or email@example.com to discuss how that might be achieved!