Reversible computing is described as a model of computation where the computational process is time reversible, meaning that the output of any computation can be used to reconstruct its input. The concept relies on the principle that no data is lost during the computation process, which differs from traditional computational models where some information is discarded, such as erasing bits.
In today’s world, computers consume large quantities of energy for operating and cooling, and AI is increasing that consumption considerably, and nearly tripling demand. At the same time, Moore’s Law is already faltering, and nearing the limit of how many transistors can be doubled, while the demand for increased power continues to soar, along with the increased heat that follows.
To overcome Moore’s Law, alternative architectures are needed. While advancements in Quantum Computing have been made, none exist as a viable option for replacing current architectures, at least not any time soon. Another potential solution comes in the form of reversible computing (a key component of quantum computing). In reversible computing, the operations that process logic can be reversed, which is done by having the outputs enable determination of the original inputs while recapturing the expended energy. The process helps minimizing the heat associated from lost data.
In contrast, typical processors use logic gates to process data using 1s and 0s. The process results in lost information both from the electrical operation and the loss of that information as well, which is expended as heat. This is why PCs and data centers rely on efficient cooling methods to continue efficient operation, and why Moore’s Law is the fast-approaching roadblock that needs to be overcome with a different architecture if computers are to continue to advance. But reverse computing is more theory than reality at this point.
Analyst Jack Gold of J.Gold Associates breaks that notion down:
“Reversible computing is really more theoretical than real. The basic premise is, every time you change the state of a logic gate and you drop a bit of data which is an electrical signal, you lose that power (the lost data has to go somewhere and laws of physics mean it goes to heat). Do this with a lot of data bits in a computer and you generate a lot of heat. So, researchers came up with a notion of lossless’ data logic, where you never really lose any bits of data, and everything can be reversed, and where the input and out are interchangeable.
"In practice, this really doesn’t work for traditional silicon-based logic circuitry very well. But people are now looking at applying some of these principles to quantum computing. Because of the way quantum bits work, it’s possible to apply some of the principles, which would mean that in theory at least, you won’t generate any heat – which is a good thing since quantum computing needs to run at near absolute zero.
"Will this happen anytime soon? Real quantum computing is just starting to get to a point of usefulness with 10s to 100s of bits. But the need for millions of ‘qubits’ means that we are still a long way from production level machines that can solve really big problems.”
With that said, there are two types of reversible computing models – physical reversibility and logical reversibility. Physical reversibility is a model of unconventional computing that results with no increase in physical entropy, meaning it's an isentropic process that's frictionless and doesn't transfer heat. Logical reversibility involves computing in a way that it always remains possible to efficiently reconstruct the previous state of the computation from the current state. This model allows for thermodynamically reversible computing, which generates no (or very little) new physical entropy and makes it energy efficient.
Heat, or rather the lack of it, is the defining factor between both models, something engineers are keen to capitalize on when designing new chips. With the rise of AI, energy and heat efficiency become pressing concerns. The demand for next-gen hardware that can handle AI applications is driving up energy bills, as a vast number of servers require extensive cooling to keep them running. Vaire Computing is looking to solve that heat issue while boosting processor performance by developing chips that take advantage of reversible computing.
Vaire was recently granted $4M in seed funding from 7percent, a venture capitalist firm that decided to invest in the company to overcome Moore's Law and create an entirely new architecture of near-zero energy chips. "Every so often, something comes along which will be transformative for the whole of humanity: jet engines, transistor microchips, or quantum computers," states 7percent. "VAIRE Computing is pioneering exactly that. We believe it could be the Intel or Nvidia of the future."
According to Vaire, the company is preparing to launch an initial version of its silicon in early 2025, with chips designed for edge deployments. Within four or five years, Vaire expects it will have chips robust enough for use in data centers. To help the company achieve its goals, Vaire is currently hiring "people who can achieve what the industry deems impossible. That through their research and development can force a paradigm shift in computer science." Perspective candidates must have an electrical engineering background with experience in chip design, be a graduate of a top university (or researcher with exceptional results), and at least have a couple of relevant papers under their belt.
Vaire CEO Rodolfo Rosini told Fierce-Network in The Five Nine podcast recently that the first chips for edge developments will provide “extreme energy efficiency” but with the commitment to greater power efficiency in later generations for use in the data center. “We believe that every chip in 15 years will be a reversable chip. It’s inescapable. There’s no alternative to this,” he said. “It’s going to be the foundational architecture of how we build computers…
“Our goal is not to be acquired by Nvidia or any other company. We think there’s an opportunity to build the next greater chip company. We think reversable computing is the platform to do it. We think it’s a multi-trillion-dollar opportunity. Someone will build a great company on this technology. We think it should be us. Right now we have the first mover advantage. We have the patent and talent advantage.”
RELATED: The Five Nine: Reversing the GPU heat problem with Vaire Computing
RELATED: Vaire Computing secures $4M in seed funding for reversible computing bet
RELATED: Watch out Nvidia: AI startup bets big on reversible computing