Skip to main content
  1. Home
  2. Computing
  3. News

This startup claims to have unlocked the secret to speedy CPUs

Add as a preferred source on Google

As technology grows more advanced, there is only so much more that manufacturers can do in order to up the ante without running into massive issues like overheating, increased power consumption, and even just the sheer size of high-end chiplets. Lightmatter, a startup, claims to have found a solution.

Through the use of photonics instead of electrical wires, Lightmatter could potentially get rid of bottlenecks and allow for the creation of even more powerful chiplets. This might be the secret to the creation of some of the best CPUs in the future.

Lightmatter's Passage interconnect.
Lightmatter

Some chiplet technologies, such as Intel’s Foveros 3D, allow for pairing multiple chips in the same substrate. These chiplets then have to be connected through the use of electrical wires. However, this results in flying electrons that then cause higher temperatures and consume more power. However, Lightmatter is approaching this in an entirely different way by replacing the wiring with photonics. The startup described its findings during Hot Chips 2022.

Recommended Videos

Electricity is quick, but light is faster. To make it work, Lightmatter introduced Passage, a “wafer-scale, programmable photonic interconnect.” Through the use of Passage, Lightmatter wants to interconnect different chips with nanophotonic wave guides instead of electrical wiring. The interconnected chiplets could include CPUs, GPUs, memory chips, or ASICs.

“Passage is diced from a 300mm Silicon Photonics wafer that includes lasers, optical modulators, photo detectors, and transistors all side-by-side integrated in the platform,” said Nicholas Harris, the founder and CEO of Lightmatter. “Because Passage has integrated lasers and transistors, the co-packaged chips don’t have to deal with any of the complexity of the transmit, receive, or circuit switching photonics elements.”

The benefits of using this technology are plentiful, including low signal loss, the ability for each tile to house several different chips, and a massive increase in bandwidth across the board.

Lightmatter teased that Passage allows it to provide up to 96Tbps of bandwidth to each die. Communication between Passage and other systems through fiber arrays maxes out at around 16Tbps. As Tom’s Hardware reports, AMD’s Infinity Fabric peaks at around 800Gbps — that’s a huge difference.

Passage also could prove to be easier to use for manufacturers, replacing proprietary technologies introduced by Intel and AMD. All it takes is to drop the device into a photonics-powered Passage and it will provide an interconnect between them. However, silicon-based transistors will still require electrical communication.

The technology is certainly one way that the future of computing could be achieved, and the potential bandwidth boost is massive. However, Lightmatter is not the only company that explores photonics as a way to battle Moore’s Law, so it remains to be seen whether the giants will turn to it or stick to their own solutions.

Monica J. White
Monica is a computing writer at Digital Trends, focusing on PC hardware. Since joining the team in 2021, Monica has written…
You can now link your favorite apps to AI Mode in Google Search to get things done
AI Mode now works with Instacart, Canva, and YouTube Music inside Search.
google-search-ai-mode-connect-apps

Google is making AI Mode in Search more useful by letting you connect third-party apps. Starting this week in the US, you can securely connect some of your go-to apps directly to AI Mode, letting Search actually complete tasks for you instead of just answering questions.

This update builds on a similar trick Google already pulled off inside the Gemini app, and now it is landing in Search itself. The initial rollout includes three launch partners, Instacart, Canva, and YouTube Music, with Google saying more app integrations are on the way.

Read more
You can now edit videos in Google Vids by simply describing the changes
Gemini Omni powers Google Vids’ new editing tools, and personal avatars are joining too
Google Vids gets Gemini Omni

Google is bringing Gemini Omni and personal avatars to Google Vids, expanding the app’s AI-powered video creation tools for paid users. Gemini Omni can now generate and edit clips through natural language, while personal avatars let users appear in videos without recording themselves on camera.

Vids already offered Veo-powered video generation, AI presenters, screen recording, and tools for turning Slides presentations into narrated videos. Omni expands that setup into a more complete editing workflow, where users can keep refining a clip through conversation instead of rebuilding it after every change.

Read more
NotebookLM just got a new name and a serious upgrade for Google AI Pro subscribers
Gemini Notebook, as it’s now called, will roll out the features that debuted on the Ultra tier last month to Pro users in the coming weeks.
Gemini Notebook branding on a MacBook

Google is retiring the NotebookLM name, and the AI research tool is being rebranded to Gemini Notebook, folding one of the company's most useful products deeper into its main AI brand. Alongside the rebrand, Google is expanding one of the tool's most powerful features to more users, which was previously limited to those on the Google AI Ultra plan.

The upgrade Pro users have been waiting on

Read more