Igor's Lab this week published an image of Intel's Lunar Lake MX processor, which proves that PC makers are hard at work developing notebooks with LNL CPUs inside. The picture also confirms that Lunar Lake uses a multi-chiplet design and comes with on-package LPDDR5X memory to maximize its performance.
Intel expects its Lunar Lake processors to offer significant improvements in the performance efficiency of its platforms due to microarchitectural innovations. The Lunar Lake MX's compute tile reportedly packs eight general-purpose cores — four high-performance Lion Cove cores and four energy-efficient Skymont cores — to offer the right balance between performance and power consumption. The compute tile also reportedly packs an integrated GPU featuring 64 Xe2 Execution Units (EUs) based on the Battlemage architecture. The system-on-chip also has a six-tile NPU 4.0 AI accelerator and a system-on-chip tile that connects to the main tile using Intel's Foveros technology and packs various I/O features.
A notable shift in Intel's approach, in the case of Lunar Lake MX, is the use of TSMC's N3B manufacturing technology for the compute tile. This move marks a significant change from Intel's traditional reliance on in-house manufacturing for its high-end chips. It is noteworthy that Intel originally planned to use its 18A (1.8nm-class) fabrication process for Lunar Lake-series CPUs, but 18A will not be ready for production until late 2024, whereas TSMC's N3B is used for mass production today.
Interestingly, Igor's Lab further claims that Intel's Lunar Lake MX processors have been co-developed in close collaboration with Microsoft, focusing on enhancing the interaction between software and hardware. This collab indicates that users could expect a better use of CPU capabilities, particularly in systems running the latest and upcoming versions of Microsoft's operating systems.
Intel's Lunar Lake MX is targeted mainly at thin-and-light laptops and is expected to feature 16GB or 32GB of LPDDR5X-8533 memory directly integrated into the package. This design choice aims to minimize the platform's size and enhance performance. According to Intel's projections from slides that were leaked several months ago, the Lunar Lake MX configuration is estimated to save between 100 to 250 square millimeters of space compared to conventional designs where the memory is separate from the CPU package.
Depending on the power target, the Lunar Lake platform will support 8W fanless and 17W – 30W actively cooled designs.
