Photonic computing technology startup Lightmatter today detailed Passage, a “wafer-scale” programmable chip interconnect fabric that uses light to boost performance while reducing energy usage. Lightmatter says that Passage, which allows arrays of heterogeneous chips including processors, graphics cards, memory, and AI accelerators to communicate with each other, operates at “unprecedented speeds” compared with existing solutions.

The technology underpinning Lightmatter’s platform — photonic integrated circuits — stems from a 2017 paper coauthored by Lightmatter CEO and MIT alumnus Nicholas Harris that described a novel way to perform machine learning workloads using optical interference. Such chips require only a limited amount of energy because light produces less heat than electricity. They also benefit from reduced latency and are less susceptible to changes in temperature, electromagnetic fields, and noise.

According to Lightmatter, Passage offers a fully-reconfigurable connection topology between chips, ostensibly reducing the cost and complexity of building heterogeneous systems. The interconnect packs forty switchable photonic lanes into a space that traditionally supports one optical fiber, enabling chip-to-chip speeds exceeding 1Tbps at a latency of 5 nanoseconds across an array of 48 chips. (Lightmatter claims these figures were measured on hardware in its lab — not merely hypothesized.) Today, most server processors communicate with one another via optical fibers at speeds around 400Gbps.

Lightmatter Passage

Above: Lightmatter’s Passage interconnect enables high-speed chip-to-chip data transfers.

Image Credit: Lightmatter

Lightmatter cofounder and CEO Nick Harris notes that Passage is chip-agnostic, so customers could put AMD and Nvidia hardware on top of Passage and see a 100 times increase in compute performance. He also asserts that because Passage significantly reduces the energy needed to power processors and datacenters, it could theoretically power the entirety of Google Search in just one server rack.

ALSO READ  Apple is Offering Discounts on Some Refurbished iPads

“Lightmatter is leading a necessary paradigm shift in computer architecture needed to power the next giant leaps in compute technology, while also reducing the negative impact on our planet of rapidly-growing state of the art, yet inefficient, compute and communications solutions,” said Nick Harris, co-founder and CEO at Lightmatter. “Modern compute workloads call for system-level performance. With Passage, we’ve created a photonic rack-on-chip solution that enables ultra-high bandwidth interconnection between different kinds of chips while simultaneously reducing cost, complexity, and energy consumption and capable of supporting the future of computing.”

Lightmatter’s hardware, which is designed to be plugged into both standard servers and workstations, isn’t immune to the limitations of optical processing. Speedy photonic circuits require speedy memory, and then there’s the matter of packaging every component — including lasers, modulators, and optical combiners — onto tiny wafers.

That may be why companies like Intel and LightOn are pursuing hybrid approaches that combine silicon and optical circuits on the same die, such that parts of the model run optically and parts of it run electronically. These companies are not alone — startup Lightelligence has so far demonstrated the MNIST benchmark machine learning model, which uses computer vision to recognize handwritten digits, on its technology. And LightOn, Optalysis, and Fathom Computing, all vying for a slice of the budding optical chip market, have raised tens of millions in venture capital for their own chips. Not to be outdone, Boston-based Lightmatter has raised a total of $33 million from GV (Alphabet’s venture arm), Spark Capital, and Matrix Partners, among other investors. Lightmatter says its current focus beyond hardware is ensuring the test chip works with popular AI software, including Google’s TensorFlow machine learning framework.

ALSO READ  Google Nest Audio Review: Big Muscles, Strong Results

The audio problem:

Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here



Please enter your comment!
Please enter your name here