Tech News

Optical Chip Promises 350x Speedup Over RTX 3080 in Some Algorithms

Lightelligence is a photonics startup based in Boston has unveiled the world’s first compact form-factor, photonics-based computer system, which employs light to execute computations. According to the business, the device is “hundreds of times quicker than a standard computing unit, such as the NVIDIA RTX 3080”. To be precise, it is 350 times faster but only for particular types of applications.

The Photonic Arithmetic Computing Engine (PACE), which is still in the prototype stage, merges electronics and photonics in a single chassis, conducting operations at the speed of light and therefore achieving exceptional acceleration for particular AI, Deep Learning, and Machine Learning compute workloads.

Because of the nature of its computational elements, PACE speeds some matrix acceleration workloads substantially faster than NVIDIA’s RTX 3080. It’s simple to understand: Lightelligence’s system has substantially reduced latency, which is the amount of time between when an event is requested to happen and when it really happens. That is the advantage of data flowing at the speed of light.

Lightelligence realized that in order to achieve this, it would have to focus not only on the optical capabilities of the PACE, but also on traditional semiconductors and software solutions that connect the two. Thus, the firm advertises itself as a hardware and software provider; the company has also developed algorithms specifically geared to solve some of today’s most fundamental computing challenges in a photonic environment.

“Our ability to co-design many different domains together is one of Lightelligence’s unique advantages over other firms creating optical computing,” said Erwan Di Vita, Lightelligence’s principal engineer for PACE. “Our photonic engineers develop the chips alongside analog, digital, packaging, and software engineers, which are then fabricated and assembled into 3D systems by our post-silicon teams.” None of this would be feasible without the ideas of our opto-electronic packaging team”.

When it comes to the specific workloads that the PACE can execute, it is a relatively limited engine. However, according to the business, “PACE efficiently searches for solutions to many of the most difficult computational math problems, such as the Ising problem and the graph Max-Cut and Min-Cut issues, demonstrating the real-world potential of integrated photonics in advanced computation.” In that sense, we might consider it an ASIC (Application-Specific Integrated Circuit): it does very few things (or a single thing) extremely well.

“These difficulties belong to an important class of intractable mathematical problems known as NP-complete, which have stumped mathematicians for the last 50 years”, stated Yichen Shen, Ph.D., founder and CEO of Lightelligence. “Algorithms for NP-complete problems are essential because they may be transferred to other problems and have hundreds of practical applications in disciplines such as encryption, power grid optimization, and advanced image analysis. The progress we’ve made on NP-complete combinatorial optimization issues demonstrates our technology’s potential to change computing”.

Credit: tomshardware

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button
Click to Chat!