Intel has introduced Lunar Lake, its latest x86 processor design, claimed to be the most efficient to date. This chipset stands out with a whopping 120 tera operations per second (TOPS) AI performance, indicating a significant leap in processing power for artificial intelligence applications.
The processor’s compute tile encompasses CPU clusters, the new Battlemage Xe2-LPG integrated GPU (iGPU), and a 48 TOPS neural processing unit (NPU). Lunar Lake adopts a hybrid 4P+4E core design, prioritizing superior single-threaded performance. Each of the four Lion Cove Performance-cores (P-cores) is equipped with 2.5 MB of L2 cache and shares a 12 MB L3 cache. Interestingly, these P-cores do not use hyperthreading but Intel is confident that the eight available threads can effectively compete with AMD’s 8-core, 16-thread processors. The implication here is that hyperthreading’s absence may only apply to Lunar Lake specifically, and not necessarily to the Lion Cove cores in general.
In contrast, the Energy-efficient cores (E-cores), named Skymont, share 4 MB of L2 cache. Recent leaks suggest that these Skymont cores may offer instructions per cycle (IPC) performance on par with AMD’s Zen 3 architecture. Intel’s internal metrics suggest that Skymont E-cores can reach up to 80% efficiency of the P-core at full power consumption, highlighting impressive efficiency gains.
A significant addition to Lunar Lake is the integration of Intel’s Battlemage Arc Graphics in the form of the Xe2-LPG. This marks its first appearance before a discrete version is available. The Xe2-LPG boasts eight Xe cores, eight ray tracing (RT) cores, the new Xe Matrix eXtensions (XMX) AI Engine, and an 8 MB cache. XMX, Intel’s answer to Nvidia’s Tensor cores, is already present in desktop Alchemist Arc cards and its integration into the iGPU allows for improved quality settings in Intel’s Xe Super Sampling (XeSS) technology.
Intel touts that the Xe2-LPG iGPU on their Lunar Lake Reference Validation Platform (RVP) delivers up to 50% faster performance than the Xe-LPG in Meteor Lake in benchmarks like 3DMark Time Spy, although details regarding the RVP’s specific configuration or specifications were not disclosed.
XMX technology not only enhances the graphical prowess of the iGPU but also boosts the total AI performance by adding another 67 TOPS. Combined with the CPU and NPU capabilities, Lunar Lake promises an impressive total of 120 TOPs AI compute performance across the platform. Moreover, Lunar Lake introduces a new media engine that supports AV1 encode/decode and VVC/H.266 decode, signaling advancements in media processing capabilities.
With these innovations, Intel’s Lunar Lake sets a new standard for power-efficient and performance-driven computing, particularly in the context of artificial intelligence and integrated graphics technology. This chipset is poised to be a formidable competitor in the CPU market, offering enhanced processing capability for both general tasks and complex AI-driven applications.
