switching losses
A perfect switch wastes no power: when off it carries no current, when on it has no voltage across it, and in either state the product (voltage times current) is zero. The trouble is the in-between. During the brief moment a MOSFET transitions from off to on or back, it passes through a state where it has both significant voltage across it and significant current through it at the same time. That overlap dissipates energy, and the energy lost on every single transition is called switching loss.
Crucially, switching loss is paid once per transition, so it grows with frequency: total switching power is roughly proportional to the energy per switch times the switching frequency. A rough estimate is P = 0.5 times V times I times (transition time) times frequency. For example, switching 100 V and 10 A with a 50 ns total transition at 100 kHz gives about 0.5 times 100 times 10 times 50e-9 times 100e3 = 2.5 watts, lost purely in the act of switching, on top of any conduction loss.
This is why total MOSFET loss has two parts that trade off against each other. Conduction loss (I^2 times R_DS(on)) dominates at low frequency and high current; switching loss dominates at high frequency. You attack switching loss by making transitions faster (a stronger gate driver, lower gate charge), but faster edges create more electrical noise and ringing, so there is always a balance. Pushing switching frequency higher to shrink the inductors and capacitors is paid for in switching loss: nothing is free.
A converter at 100 kHz has 1 W conduction loss and 2.5 W switching loss; double the frequency to 200 kHz to shrink the magnetics and switching loss roughly doubles to 5 W while conduction stays at 1 W.
Switching loss scales with frequency; conduction loss does not.
Faster switching cuts switching loss but injects more EMI and ringing. The right edge speed is a deliberate compromise, not as fast as possible.