Digital computer being compared to the human brain must be normalized and/or specified. Eg, 32 bit and 64 bits have different power; supercomputers and desktops , can successfully compute a different set of functions (even regardless of time taken).
Darendra Modhi, lead of IBM's SyNAPSE project, estimates the brain has 38 *petaflops* of computing power . [1] TaihuLight, the world's most powerful supercomputer, now has 93 petaflops. [2] They have radically different energy requirements and operate at radically different speeds.
Definition of computational power: the ability to compute a function. there is a function that A can compute and B can't => A and B don't have the same computational power. (A also can compute every function that B can compute => A has more computational power than B)
Given a fixed (finite) amount of time, both theoretical Turing machine , digital computer, brain can only manipulate a finite amount of data. Given this, all Turing complete machines != equivalent computational power (per paren'ts definition).
A more accurate model of computational power for modern generation digital computers is the Random-access stored-program machine [1]. Analog, recurrent neural networks may be represented by non-deterministic TM, or something even more complex [2] > different power than RASP.