IBM researchers have demonstrated a way to bring optics’ speed and capacity inside data centers.In a technical paper, IBM introduced a new co-packaged optics (CPO) prototype module that can enable high-speed optical connectivity. Researchers have pioneered a new process for CPO to enable connectivity within data centers at the speed of light to complement existing short reach electrical wires. This technology could significantly increase the bandwidth of data center communications, minimizing GPU downtime while drastically accelerating AI processing. This research innovation, as described, could dramatically improve how data centers train and run generative AI models, IBM said.By designing and assembling the first publicly announced successful polymer optical waveguide (PWG) to power this technology, IBM researchers demonstrated how CPO will redefine the way the computing industry transmits high-bandwidth data between chips, circuit boards, and servers.Today, fiber optic technology carries data at high speeds across long distances for commerce and communications. Data centers use fiber optics for external communications networks, but racks inside those data centers still predominantly run communications on copper-based electrical wires. These wires connect GPU accelerators that may spend more than half of their time idle, waiting for data from other devices in a large, distributed training process which can incur significant expense and energy, IBM noted. This new approach, IBM said, could result in lower costs for scaling generative AI through a more than 5x power reduction in energy consumption compared to mid-range electrical interconnects, while extending the length of data center interconnect cables from one to hundreds of meters.In addition, it could speed up AI model training, enabling developers to train a large language model (LLM) up to five times faster -- reducing the time it takes to train a standard LLM from three months to three weeks, with performance gains increasing by using larger models and more GPUs. And it could dramatically increase energy efficiency for data centers, saving the energy equivalent of 5,000 U.S. homes’ annual power consumption per AI model trained, IBM researchers said.As always, drop me a line at [email protected] if you have news to share or want to say hi!Grab your coffee. Here's what you need to know today.Right of Boom, Feb. 19-21, 2025, MGM Grand Resort, Las Vegas, Nevada CompTIA Communities & Councils Forum, March 10-12, 2025, Chicago Channel Partners Conference & Expo and MSP Summit, March 10-12, 2025, The Venetian, Las Vegas, Nevada N-able Empower, week of April 5, 2025, Berlin, InterContinental Hotel NerdioCon, April 7-9, 2025, La Quinta Resort and Club, Palm Springs, California Kaseya Connect, April 28-May 1, 2025, MGM Grand Resort, Las Vegas, Nevada
You can skip this ad in 5 seconds