I did a fresh install of 3.1.4.
Did you use the ROM option or the RAM/LoadModule option? In the latter case, it may happen (depending on your system) that parts of the modules went into ChipMem, and hence it gets slow. "MuProtectModules" can help in this case. Also, depending on your setup, the exec.library may end up in chip memory. MuFastZero in combination with MuMove4K can help in this case.
I used the default mask and max transfers. Perhaps that is the culprit? What are the best settings for performance to use? Does block size matter I used 1024 for the SSDs?
There is nothing you can do concerning the mask or maxtransfer to improve the performance. Mask=0xffffffffe is fine, and maxtransfer can be as large as you want (3.1.4. does not require setting a maxtransfer value anymore). For SSDs, I would recommend block sizes of 4096 bytes (4K).