Intel ‘Lunar Lake’ Core Ultra Laptop CPUs, Xeon 6 Debut at Computex 2024

Intel has formally announced its much-hyped “Lunar Lake” CPU architecture, which will come to market as the Core Ultra 200 series, for thin-and-light laptops starting in Q3 2024. Intel promises groundbreaking advances in not just performance but also graphics, AI acceleration, and power savings. The launch comes at a particularly significant time, as new competition from Qualcomm and other Arm-based CPU manufacturers is set to disrupt the laptop space for the first time since Intel established its x86 PC standard decades ago. With Microsoft announcing its CoPilot+ software platform for AI-enabled PCs, and with Windows 11 now able to run on both X86 and Arm CPUs, Intel is saying this generation of CPUs will change the way we think about laptops.

Intel CEO Pat Gelsinger, who presented a keynote speech at Computex 2024 on Tuesday, AI PCs will make up over 60 percent of all PCs sold by 2027. Every device will become an AI device and every company will become an AI company, he said, adding that Intel can supply the entire continuum of products needed to power that, especially with its Intel Foundry capabilities.

While exact CPU specifications will be revealed closer to the actual launch day, Intel has announced several high-level details of the Lunar Lake architecture. Starting with CPU cores, these CPUs will feature four new “Lion Cove” Performance cores and four “Skymont” Efficiency cores, both of which have been significantly overhauled with a focus on power efficiency. E-cores take the primary role, with P-cores only kicking in when needed by demanding workloads. One huge change is that Hyper-Threading, which allowed two instruction threads to be scheduled per core, has been dispensed with. Intel says its physical E-cores are more efficient, and more power can be saved by simplifying thread scheduling.

Another huge change is the use of on-package RAM for the first time. By placing RAM chips directly on the CPU package, laptop OEMs can reduce total motherboard footprint, allowing for larger batteries or lower weight. This also simplifies electrical trace routing and reduces power use. Lunar Lake CPUs will be available with 16GB or 32GB of dual-channel LPDDR5X-7500. 

The integrated NPU is said to be capable of 48 TOPS of throughput, making it able to speed up generative AI tasks such as image generation, live captioning, and Microsoft’s CoPilot+ experiences. Overall platform power efficiency is up to 40 percent better in specific tasks. 

Graphics power gets a boost with a new 2nd Gen Intel Xe integrated GPU. Intel claims up to 2X performance in games compared to the previous generation, even on thin-and-light laptops. Moreover, a significant amount of AI processing happens on the GPU, and with a capacity of 67 TOPS on the GPU alone, Lunar Lake can achieve 120 TOPS total AI throughput across its components. Lunar Lake CPUs can power up to three 8K displays at 120Hz. 

Gelsinger also announced new Intel Xeon 6 processors with up to 144 E-cores, delivering significant power savings and the ability to consolidate space to datacentres. Intel promises up to 4.2X throughput for video transcoding, 2.6X performance per Watt, and “game-changing” power savings. Later this year, Intel will introduce a 288-core model made up of two 144-core dies on a single package. 

Talking about datacentre AI acceleration, Gelsinger announced how its Gaudi 2 and Gaudi 3 datacentre AI accelerators can be used to power multimodal AI enterprise applications and retrieval-augmented generative AI which can help deal with sensitive data securely and on-premises. Gaudi 2 and Gaudi 3 AI kits will be available for US$ 65,000 and $125,000 respectively, which Gelsinger said is up to 2/3rds more affordable than the competition while offering up to 2X faster inferencing and 40 percent lower time-to-train.

Disclosure: Airfare and accommodation for the visit to Taipei were sponsored by Intel