Wouldn't that depend on if there is a time limit? It seems under time constraints, a human grandmaster aided by a computer would do better than the computer alone, since the human can also use heuristics.
An new iPhone can run the latest chess engines at a level that should beat all but the absolute best players (https://www.hiarcs.com/iphone-chess.html).
Apple doesn't publish power stats for their A-series chips that I could find, and I couldn't find a direct comparison of power draw running a chess engine, but Anandtec's numbers running synthetic benchmarks indicate between 3-5 watts when under single thread load: https://www.anandtech.com/show/16983/the-apple-a15-soc-perfo....
Estimates for human brain energy consumption are vague, but generally accepted somewhere between 12-25 watts, with 20 being the most used average guess (https://www.quora.com/How-much-energy-does-a-human-brain-con..., https://www.scientificamerican.com/article/thinking-hard-cal..., https://www.munichre.com/topics-online/en/digitalisation/int...)
Given these numbers, not infeasible to assume that a computer restricted to the energy usage of the human brain would win every game when not playing Magnus?