Value and Conclusion

Often people think Intel's main competition is AMD Ryzen, when it's really Apple M-series and Qualcomm Snapdragon Elite. The latter two are powering super-efficient ultraportable notebooks with smartphone-like battery life and snappiness, and that's something you simply can't counter with a bulky processor with huge CPU cores. This is why Intel pivoted to heterogeneous multicore (hybrid), and has since focused not just on minor single or low-double digit IPC gains for its P-cores, but major performance leaps for its E-cores.

You have to understand that E-cores were developed by "removing things," from a typical core and are a frugal product of reduction, while the P-cores are developed by "adding things" to a typical core, and are a product of addition. This is why Intel has a vast canvas to add capabilities and performance to its E-cores; and this is precisely why Skymont is a breakthrough.

Just the single slide where Intel showed IPC similarities between Skymont and Raptor Cove (a P-core), should send shockwaves across the industry, and the likes of Apple, Qualcomm, and AMD, will take note. Skymont is able to match the IPC of Raptor Cove at a fraction of its transistor count and hence a fraction of its power. Intel now has the freedom to flood its processors with Skymont clusters, and a handful Lion Cove P-cores, to create a formidable hybrid processor that can square off against high core-count AMD processors that only have high-performance cores, or their compacted versions.

Lunar Lake isn't the only microarchitecture combining Lion Cove and Skymont, there's also the upcoming Arrow Lake. While Lion Cove puts the Skymont cores on a low-power island, detached from the P-core ring, Arrow Lake will have multiple Skymont clusters share the ringbus and L3 cache with the Lion Cove P-cores, so you can only imagine the carnage Intel can unleash on the competition in multithreaded performance benchmarks.

The Xe-LPG iGPU powering graphics for Meteor Lake is impressive enough, as it beat the Radeon 780M RDNA 3 iGPU of competing Ryzen 7040 Phoenix processors of its time; and here, Intel is promising a 50% generational gain in iGPU performance on the back of the new Xe2 Battlemage architecture, which not just packs updated ray tracing hardware, but also the full XMX AI accelerators that were previously only found in Intel's discrete GPUs; so AI-based upscaling technologies such as XeSS run without impacting the GPU's SIMD performance (if you recall, XeSS used a DP4a path for Xe-LP and Xe-LPG). Intel has also taken the opportunity update the display- and media engines, including VVC hardware decode, eDP 1.5, DP 2.1, and HDMI 2.1.

NPU 4 doesn't just quadruple the AI inferencing performance over NPU 3, thereby meeting the requirements for Microsoft Copilot+, but also does so without a linear increase in power. This is not just thanks to the updated foundry node (from 5 nm to 3 nm), but also a host of architectural improvements, and improved memory bandwidth.

Lunar Lake is exclusively an MoP (memory on package) microarchitecture, it was designed with a specific mix of accelerators, each with its unique performance and power profile, so Intel could take on the likes of the Apple M3 and Snapdragon Elite X, beat them in performance, while matching them in PCB footprint, power and battery life. This could only be possible by bringing the LPDDR5X memory to the package, giving Intel greater control over the hardware stack.

For this reason, there likely won't be Lunar Lake in conventional U- or P- packages. If you want to see Lion Cove, Skymont, Xe2 Battlemage, and NPU 4 in a more familiar package, you should look out for Arrow Lake, which not just covers other mobile form-factors, but also desktop.