‘Energetics of Computing’ asks why computers use so much energy
Zoom in on a digital computer, and you’ll see electrons moving in extraordinarily complicated, time-varying patterns. Those patterns of electrons are the physical form of the logical patterns of 1s and 0s that represent information in binary code; the dynamics of those patterns is the physical form of the information processing performed by the computer as it runs an algorithm.
It requires energy to move these electrons around according to those patterns — and using that energy to move all those electrons produces waste heat, so much so that about 5% of all US and European energy consumption currently goes to the generation of heat by running computers.
Despite the huge monetary and environmental cost of digital computation, experts still only have a vague understanding of how the heat produced by a computer depends on the dynamics of the patterns of 1s and 0s within the computer. However, recent breakthroughs in nonequilibrium statistical physics hold the promise of revolutionizing our understanding of the heat produced by information processing — whether it be processing the patterns of 1s and 0s in digital computers, or the analog information processed inside biological computers, such as living cells and human brains.
The Energetics of Computing in Life and Machines, edited by David Wolpert, Chris Kempes, Peter Stadler, and Joshua Grochow, lays out recent advances that are driving a new “thermodynamics of computation,” addressing questions such as:
What are the fundamental equations relating the amount of energy required to run a computation to the details of that computation? (Chapters 1 and 9)
How — and what — do biological systems compute? (Chapter 7)
How could we automatically reduce the energy consumption of software? (Chapter 10)
Written by and for scientists working across disciplines that include computer science, physics, cellular biology, and neurobiology, the book provides “a snapshot of the current state of the field,” where, according to Wolpert, “these ideas are starting to explode, in all directions.”
By presenting what is now known, and unknown, about the relationship between energetics and the microscopic processes that occur during computation, the editors hope to entice a new generation of scientists to work on questions that could have far-reaching consequences for how we understand all of our computers, be they conventional, digital computers, or biological computers. On a purely practical level, this research could lead to cooler-running, more powerful digital machines, like exascale computers and even tiny swarm robots. It could also impact the sustainability of computing technology — for instance, can biological systems serve as inspiration for designing computers with minimal thermodynamic cost?
The fourth volume in SFI Press’s seminar series, The Energetics of Computing in Life and Machines publishes new findings from the Institute’s ongoing Thermodynamics of Computation research project and the associated wiki for collaborators.
The book is available for purchase on Amazon.