Intel has described Lunar Lake as a "radical low-power architecture," a statement that reflects its commitment to power efficiency. This new focus allows Intel to discard features from previous generations that don't contribute to this goal, including the need for in-house chip manufacturing.
A major shift in Intel's strategy is the introduction of a new core structure. Lunar Lake features a hybrid architecture like previous Intel generations, but prioritizes Skymont efficient (E) cores over Lion Cove performance (P) cores. Intel identifies Skymont as the primary engine driving performance in Lunar Lake.
Unlike earlier models that offloaded low-power tasks to E-cores, Lunar Lake's E-cores will handle most tasks, with P-cores activating only when additional performance is required.
Intel's performance targets have evolved significantly. While previous E-cores could manage basic tasks like streaming video, more demanding applications, such as Microsoft Teams, would activate P-cores. This won’t be the case with Lunar Lake.
To facilitate this approach, Intel has developed a more powerful E-core. The company claims that Skymont can match the performance of Meteor Lake’s E-core while consuming just one-third of the power. Additionally, at peak performance, it is expected to double single-core performance.
Moreover, Intel asserts that Skymont achieves a 2% increase in Instructions Per Clock (IPC) on average compared to Raptor Cove. This performance leap positions Intel's new E-core for laptops against its last-generation P-core for desktops.
This commitment to efficiency permeates all aspects of Lunar Lake’s design. For example, Intel uses separate voltage rails for the P-cores and E-cores, allowing for complete shutdown of P-cores when they are not in use. There are also enhancements such as increased cache for E-cores (4MB) and improved access to on-board memory.
While Skymont takes center stage, the Lion Cove cores still emphasize efficiency. A notable—and potentially controversial—change is the elimination of Hyper-Threading, which Intel explains as "not free."
Intel's latest strategy also marks a shift in its manufacturing approach. Traditionally, Intel has designed and produced its own CPUs, but with Lunar Lake, it has chosen to outsource chip production to TSMC, known for working with Apple, AMD, and Nvidia.
The compute tile at the heart of Lunar Lake will utilize TSMC’s N3B process, while the platform tile leverages the N6 process. N3B is a leading-edge node that Intel does not currently have a comparable process for; instead, the company is focusing on developing its 18A node, set to launch next year. Though TSMC will manufacture Lunar Lake chips, Intel emphasizes that it doesn’t rely on one manufacturer, aiming for global production capabilities.
Now, focusing on specifications, Lunar Lake is based on an eight-core CPU divided into two clusters: four E-cores and four P-cores. Intel has indicated it can scale this design but will maintain the core cluster of four.
Significant improvements come in the form of the redesigned neural processing unit (NPU) tailored for Copilot+ PCs. It can now handle up to 48 Tera Operations Per Second (TOPS), significantly surpassing Meteor Lake and slightly exceeding the Snapdragon X Elite.
In addition to the NPU, Lunar Lake boasts over 120 platform TOPS, with contributions from the CPU (5 TOPS), NPU (48 TOPS), and an impressive 67 from the GPU featuring Intel’s new Xe2 graphics architecture.
This Xe2 architecture, also known as Battlemage, will drive both Lunar Lake and future Intel desktop graphics cards. Intel claims a performance increase of up to 50% compared to the previous generation, building on the substantial advancements already seen with Meteor Lake’s GPU.
This performance leap results from several architectural changes. For instance, it now natively supports the ExecuteIndirect command, one of the most common in DirectX 12 games, potentially boosting performance by more than 12 times in specific scenarios. Intel has also implemented a new compression method and improved cache clearing to enhance efficiency.
Lunar Lake functions as a complete system-on-a-chip (SoC), featuring a display engine supporting DisplayPort 2.1 and HDMI 2.1, alongside a media engine that handles codecs like AV1 and the new VVC codec. Additionally, Lunar Lake can support up to 32GB of LPDDR5X memory.
While we await the arrival of Lunar Lake devices to make definitive assessments, many factors such as clock speeds and power targets remain uncertain. This new direction for Intel contrasts sharply with the approaches of competitors like AMD and Qualcomm, emphasizing the importance of seeing how Lunar Lake performs in real-world applications.
One thing is clear: Intel is embarking on a transformative journey. Although Meteor Lake appeared inconsistent, it fits more coherently within the context of Lunar Lake. With a streamlined core configuration, an emphasis on efficiency, and collaboration with leading foundries, Lunar Lake represents not just a new generation, but a significant reinvention of Intel.
Ranking
