Lightelligence, a Boston-based photonics company, has revealed the world’s first small form factor, photonics-based computing device, which means it uses light to perform computational operations. The company claims the unit is “hundreds of times faster than a typical computing unit, such as the NVIDIA RTX 3080.” 350 times faster, to be exact, but that only applies to certain types of applications.
Also in the demonstration stage, the Photonic Arithmetic Computing Engine (PACE) integrates electronics and photonics into a single chassis, performing operations at the speed of light and thus achieving extreme acceleration for certain computational workloads. of AI, deep learning and machine learning.
PACE speeds up some matrix acceleration applications much faster than NVIDIA’s RTX 3080 due to the very nature of its compute elements. It’s pretty easy to understand: Latency, which is the amount of time between when an event is supposed to happen and when it actually happens, is much, much lower on the Lightelligence system. This is the advantage of data traveling at the speed of light.
To achieve this, Lightelligence knew it would have to focus not only on the optical capabilities of PACE, but also on traditional semiconductors and the software solutions bridging the two. The company thus presents itself as a supplier of hardware and software; the company has also engineered specially designed algorithms to solve some of today’s most fundamental computing problems in a photonic environment.
“One of the unique advantages of Lightelligence over other companies developing optical computing is our ability to co-design many different areas together,” said Erwan Di Vita, chief engineer for PACE at Lightelligence. “Our photonic engineers work alongside analog, digital, packaging and software engineers to design the chips, which are then fabricated and integrated into 3D systems by our post-silicon teams. Without these innovations from our packaging team opto-electronics, none of that would be possible. “
PACE is a somewhat narrow engine when it comes to the exact workloads it can perform. But, as the company says, “PACE effectively seeks solutions to many of the most difficult mathematical problems in computation, including the Ising problem and the Max-Cut and Min-Cut graphics problems, illustrating the real potential of computing. photonics integrated into computations. ”From this perspective, we can classify it as a kind of ASIC (Application-Specific Integrated Circuit): it does very very good very little (or only one thing).
As Yichen Shen, Ph.D., Founder and CEO of Lightelligence, said, “These problems belong to a large class of intractable mathematical problems known as NP-complete that have baffled mathematicians for the past 50 years. “, did he declare. “The algorithms for NP-complete problems are important because these problems can be mapped to each other, and they have hundreds of practical applications in areas that include cryptography, power grid optimization, and data analysis. Advanced picture. The progress we have made on NP-complete Combinatorial optimization problems illustrate the potential of our technology to transform computing. “
However, PACE achieves this coveted specialization through an additional computing area – which not only makes the system faster, but makes it incredibly more efficient. While traditional semiconductor systems face the problem of excessive heat resulting from current passing through nanoscale characteristics at sometimes ridiculous frequencies, the photonic system treats its workloads with zero ohmic heating – it there is no heat produced by the current resistance. Instead, it’s all about light.
Lightelligence is built around the doctorate of its CEO. thesis – and the legitimacy it confers. Indeed, when “Deep Learning with Coherent Nanophotonic Circuits” was published in Nature in 2017, CEO and Founder of Lightelligence, Yichen Chen, had already planned a way for optical circuits to be at the forefront of computing learning efforts. Automatique. By 2020, the company had already received $ 100 million in funding and employed around 150 employees. A year later, Lightspeed produced a demo product that it says is “hundreds of times faster than a typical computing unit, such as the NVIDIA RTX 3080”. 350 times faster, to be clear.
PACE’s debut aims to attract enough capital to comfortably meet its goal of bringing an AI accelerator pilot product to market in 2022. This is still only an ambitious goal in the company’s vision, however, its objective is to develop and distribute a mass market. , a photonics-based hardware solution from 2023, targeting the cloud AI, finance and retail markets. Considering how Lightelligence has managed to improve the company’s COMET 2019 design performance by a factor of a million with PACE in the span of two years, it will be interesting to see where their efforts take them. regarding the launch.
