Well, one thing UniPCemu is that those emulators aren't is that it's a cycle-accurate emulator. And as is known, accuracy takes power. A more fast mode is already implemented using IPS mode(which makes the CPU function much like Dosbox, but with maintained mechanics for CPU-BIU communication for emulator compatibility with both modes).
There are currently two parts of emulation that are very demanding on the CPU: CPU emulation(interpreter emulation at ~30% CPU usage on Intel i7-4790K@4.0GHz) and (S)VGA video card emulation is measured at ~15%(CGA is faster due to using the base 14MHz bus clock like other hardware). CPU can be slowed down without visual effect, but slowing down the (S)VGA actually lowers the framerate, leading to even worse performance, from a user point of view. The basic accuracy of both is required to properly function, so the only way ahead is just to optimize further, which still takes (more) time.