Intel details Panther Lake for next-gen laptops

Intel has presented the full technical overview of Panther Lake, the next-generation Core Ultra Series 3 platform. The design centers on Intel’s 18A process node, bringing together new Cougar Cove performance cores, Darkmont efficiency cores, and an updated Xe3 graphics architecture in a chiplet-based package built with Foveros-S stacking. The platform also introduces a fifth-generation NPU5 for AI acceleration and will officially debut around CES 2026, with notebook systems expected in the first quarter of the year.

Source: Intel

Panther Lake is organized around three main tiles: the compute tile on Intel 18A, the graphics tile manufactured on either Intel 3 or TSMC N3E, and the platform controller tile made on TSMC’s N6 node. Each configuration is mounted on a common base tile using Foveros-S packaging, combining CPU, GPU, and I/O dies into a compact SoC layout.

Panther Lake: Scalable AI PC Performance Built on 18A:

• Lunar Lake-level power efficiency and Arrow Lake-class performance.1
• Up to 16 new performance-cores (P-cores) and efficient-cores (E-cores) delivering more than 50% faster CPU performance vs. previous generation.2
• New Intel® Arc™ GPU with up to 12 Xe cores delivering more than 50% faster graphics performance vs. previous generation.3
• Balanced XPU design for next-level AI acceleration with up to 180 Platform TOPS (trillions of operations per second).4
• Beyond the PC, Panther Lake will extend to edge applications including robotics. A new Intel Robotics AI software suite and reference board enables customers with sophisticated AI capabilities to rapidly innovate and develop cost-effective robots using Panther Lake for both controls and AI perception.

Source: Intel

The compute tile integrates three types of cores. The Cougar Cove P-cores evolve from Lion Cove with improved memory disambiguation, larger TLBs, and a more accurate multi-level branch predictor. Each P-core includes 3 MB of L2 cache and 256 KB of L1 cache, with a small L0 data cache for low-latency access. The Darkmont E-cores are a wider, faster version of the previous Skymont design, supporting a 9-wide decode, a larger out-of-order window with 416 entries, and 26 dispatch ports. Each four-core E-cluster shares 4 MB of L2 cache. Panther Lake also adds a four-core Low Power Efficiency cluster, built on the same Darkmont architecture and located directly on the compute tile, handling background or lightweight workloads without waking the larger CPU complex.

Source: Intel

Intel claims single-threaded performance improves by roughly 10% over Lunar Lake and Arrow Lake at the same power level, or can achieve the same performance with forty percent less power. In multi-threaded workloads, the company reports more than fifty percent higher performance than Lunar Lake under equal power conditions and up to thirty percent lower power at Arrow Lake’s performance level. These gains are attributed primarily to the 18A process efficiency and the balanced mix of P, E, and LP-E cores.

Source: Intel

The redesigned memory subsystem supports DDR5-7200 and LPDDR5X-9600 , offering higher bandwidth and capacity than previous generations while abandoning the memory-on-package approach used in Lunar Lake. The compute tile shares up to 18 MB of L3 cache between core clusters and connects to an 8 MB memory-side cache, which reduces DRAM traffic and latency.

Source: Intel

On the graphics side, Panther Lake introduces the Xe3 architecture , offered in two tile configurations: a small version with four Xe cores built on Intel 3, and a larger one with twelve Xe cores produced on TSMC N3E. Both include increased L1 and L2 cache capacity, improved anisotropic filtering and stencil rates, and an enhanced ray tracing unit with dynamic ray management. Intel cites around fifty percent higher GPU performance compared to Lunar Lake at the same power level. The platform also marks the debut of XeSS 3 with Multi-Frame Generation , a technology that produces multiple interpolated frames for smoother rendering. Intel plans to add a Frame Generation Override control to its graphics software, letting users force specific frame generation modes.

The integrated NPU5 achieves around 50 TOPS , a modest increase from Lunar Lake’s NPU4 but with much higher area efficiency. It supports new AI formats such as FP8 and INT8 , doubling MAC throughput and cutting power consumption by more than forty percent. When combined with the CPU and GPU accelerators, the total platform compute performance reaches approximately 180 TOPS .

Power and scheduling are coordinated by an updated Thread Director , which adapts to changing workloads. Under heavy GPU use, tasks are first assigned to E-cores to free up power for graphics processing, a method Intel says can improve frame rates by about ten percent in gaming workloads. A new system utility called Intelligent Experience Optimizer can automatically switch between Windows power modes depending on demand, delivering up to twenty percent better performance within the same power limits.

Source: Intel

Panther Lake CPUs will be offered in three die combinations: an 8-core version with four P-cores and four LP-E cores; a 16-core model with four P-cores, eight E-cores, and four LP-E cores; and a flagship 16-core version paired with a 12-core Xe3 GPU. Connectivity options include up to twenty PCIe lanes, Thunderbolt 4 integration, and optional Thunderbolt 5 via discrete controllers. Wireless support is upgraded to Wi-Fi 7 Revision 2 with the Whale Peak 2 BE211 module and Bluetooth Core 6.0 with LE Audio and Auracast features.

Panther Lake will begin ramping high-volume production this year, with the first SKU slated to ship before the end of the year and broad market availability starting January 2026.

— Intel

Intel’s upcoming mobile platform thus combines its first 18A compute tile with a modular GPU, an AI-optimized NPU, and renewed focus on efficiency. While architectural changes to the cores themselves are moderate, the unification of packaging, higher performance per watt, and expanded AI and graphics capabilities mark a clear progression toward Intel’s next generation of mobile chips, expected to ship as Core Ultra 300 in early 2026.

Source: Intel