Suppose we write a program in `C` which requires 32 bit array, sets an input state, operates on the array, and produces an output state. This program is the logical model, here requiring say 32bits and 10 operations/bit average. So we have 320 logical changes to the array (which btw, wont be purely parallel or serial).
Suppose the most efficient CPU we have requires E_thermo energy for this whole process (setting the 32bit input, operating, reading the output).
Now, using somewhat disputed ideas about physical limits, E_thermo say implies we've used 320,000 'bits' thermodynamically, ie., somehow the CPU has produced an equivalent of 320,000 "energetic changes".
So the CPU has an efficiency, in some suspicious sense, of 100,000 thermo-bits / logical-bit. For each single on-average change to the array we see in the `C program`, we theoretically measure an equivalent heat of 100,000 changes.
Now, how efficient is the brain at computation? Well since we have no logical model of the computation its performing, there's absolutely no way of answering that question.
So why do people say it's efficient? Well because any apparently equivalent algorithm we come up with (in, eg., `C`) to do even basic things that the brain does, requires vast amounts more energy. Eg., processing images on a GPU requires, say 300 W, and our whole brain uses (a claimed) 20 W.
I think the resolution to this problem is that the brain(-environment-body) system implements computation in a radically different manner than anything like a CPU, GPU, etc. I suspect that computation happens at every scale: molecular, sub-cellular, cellular, neuronal, inter-neuronal, nervous-system, body, body-environment, etc. And across the whole body, and using energy in the environment to maintain state.
However, I find the use of the term "computer" and "computational" mostly just productive of pseudoscience. It is a basically meaningless term that only produces confusion. The only useful area where it helps is when handling logical models of algorithms -- models largely absent from the whole of empirical and theoretical science.