Intel Guarantees Higher AI Efficiency, Longer Battery Life with Upcoming “Lunar Lake” Chips

Intel has introduced its “Lunar Lake” chips for on-device machine studying and synthetic intelligence on laptops and compact computing programs — that includes a 48 TOPS neural coprocessor and the promise of a 40 % energy discount in comparison with the corporate’s earlier era.

“AI is driving one of the vital consequential eras of innovation the business has ever seen,” claimed Intel’s chief government officer, Pat Gelsinger, throughout a keynote presentation on the Computex commerce present this week. “The magic of silicon is as soon as once more enabling exponential developments in computing that can push the boundaries of human potential and energy the worldwide financial system for years to come back.”

To energy that period, Intel has introduced its Lunar Lake processor structure. This, the corporate says, delivers a 40 % enchancment in vitality effectivity and as much as a 60 % enhance in real-world battery utilization — however, extra importantly, comes with a fourth-generation neural processing unit (NPU) designed to speed up on-device machine studying and synthetic intelligence workloads, together with generative AI (gen AI).

The Lunar Lake chips are constructed from three distinct parts, with the “compute tile” delivering a mix of energy-efficient and high-performance 64-bit x86 CPU cores, Intel’s newest Xe2 “Battlemage” graphics processing cores, and the devoted NPU 4 accelerator — delivering a claimed 48 tera-operations per second (TOPS) of minimum-precision compute for AI and ML, on prime of the up-to-67 TOPS out there from the GPU.

A “platform controller tile” provides safety features and a connectivity suite together with Wi-Fi 7, Bluetooth 5.4, PCI Categorical Gen. 5 and Gen. 4, and Thunderbolt 4; lastly, as much as 32GB of on-package reminiscence is designed to drop latency and allow quicker entry to information — serving to to push down the system’s total energy consumption, the corporate claims.

Intel’s newest NPU-equipped chips observe sizzling on the heels of these from rival AMD, as the 2 compete to observe Qualcomm into powering transportable units constructed to Microsoft’s Copilot+ specification — which mandates a minimal NPU efficiency and the presence of a devoted keyboard key for launching Home windows 11’s Copilot AI assistant. AMD, nevertheless, seems to have crushed its longstanding competitor within the NPU efficiency stakes, promising 50 TOPS of efficiency from its Ryzen AI 300 collection NPU to Intel’s 48 TOPS.

On the identical time, Intel unveiled its newest {hardware} for AI on the information heart stage, providing pricing for its Gaudi 2 and Gaudi 3 accelerators — with a “commonplace AI equipment” primarily based round eight Gaudi 2 accelerators priced at $65,000, a claimed two-thirds low cost over “comparable aggressive platforms,” with an eight-accelerator Gaudi 3 equal hitting $125,000 — and unveiling its Intel Xeon 6 E-core “Sierra Forest” and P-core “Granite Rapids” chips.

The Lunar Lake chips will probably be out there in retail programs within the third quarter of the 12 months, Intel guarantees, “in time for the vacation shopping for season.” Pricing and full specs had not been disclosed on the time of writing.