At the Intel Tech Tour 2025, we had the opportunity to explore the groundbreaking 18A chips, Panther Lake and Clearwater Forest, along with various platforms incorporating these cutting-edge technologies.
Unveiling Intel 18A: A First Look at Panther Lake & Clearwater Forest
The highlight of Intel Tech Tour 2025 was the guided visit to Fab 52, the primary facility dedicated to the production of next-generation chips utilizing the 18A process node. During our tour, we witnessed the manufacturing process of Panther Lake and Clearwater Forest wafers, providing critical insights into Intel’s innovative production capabilities. These chips are anticipated to be released in 2026, with Panther Lake targeting consumer markets, while Clearwater Forest will focus on the server segment.
Here’s a brief glimpse into the exciting activities happening at Fab 52:
In addition, we were also shown the silicon wafers being produced:
Insight into Intel Panther Lake with 18A
Among the star attractions were two unique configurations of Panther Lake, along with the comprehensive 18A Compute Tile wafer. The Panther Lake 12Xe features an 18A Compute Tile and a graphics tile fabricated using TSMC’s N3E process node. Meanwhile, the more compact Panther Lake 4Xe configuration employs the same 18A compute tile but combines it with a graphics tile based on Intel’s 3 process technology.
The smaller 4Xe configuration maintains a 4-tile setup, comprising the compute tile, graphics tile, I/O or SoC tile, and a filler tile. Much like its 12Xe sibling, this configuration employs Intel’s advanced Foveros 2.5D packaging technology.
Panther Lake RVP Motherboard Overview
On display was Intel’s Panther Lake RVP (Reference Validation Platform) motherboard, designed for testing and validating Panther Lake CPUs before their official launch. This motherboard has been featured in other prominent technology showcases, including Computex 2025.
This motherboard is equipped with a BGA 2540 socket, compatible with Panther Lake-H and Panther Lake-U CPUs. A key innovation is its use of LPCAMM2 memory instead of conventional SO-DIMM modules.
The RVP motherboard incorporates several features, including an 8-phase VRM design, two PCIe slots (x4/x1), multiple USB ports, voltage readpoints, and two USB Type-C ports. Furthermore, it includes additional headers for troubleshooting, diagnostic features, and DEBUG LEDs for ease of use. Powering the motherboard requires either a single 12-pin EPS connector or a DC-In jack.
Intel also featured Crucial’s LPCAMM2 modules used in the RVP, with a capacity of 64 GB and speeds reaching 7500 MT/s. The module bears the part number “CT64G75C2LP5X. M48C1”.Furthermore, Intel has projected LPCAMM2 support for Panther Lake CPUs to go as high as 9533 MT/s and capacity limits of 96 GB.
The RVP motherboards are integrated into a Reference Validation Platform (RVP), a cubical system equipped with a display and various diagnostics and validation functionalities.
The demo systems featured slightly different RVP motherboards, utilizing two DDR5 SO-DIMM slots, each with capacities that also reach 64 GB and operate at speeds of 5600 MT/s (CL46 @ 1.1V).Additionally, the CPU was connected to a high-efficiency copper heatsink complete with a 9000 RPM active cooling fan.
Panther Lake Development Kit & Robotics Edge Module
Another standout was the PTL DevKit, a compact Mini PC-like device featuring an SFF motherboard equipped with LPCAMM2 memory, allowing developers the flexibility to optimize their applications for the Panther Lake architecture. This efficient machine boasts a wealth of I/O options.
Finally, Intel revealed its Panther Lake module tailored for robotics and edge computing platforms. This compact PCB features the Panther Lake CPU along with DRAM memory embedded into four sites. The specific DRAM components utilized were sourced from Micron, bearing the part number “4ZC42 DBFVB”.
The remaining area of the PCB is devoid of components, providing a preview of how the final iteration will appear; the back will accommodate additional circuitry as well as a connector.
Clearwater Forest: Unveiling Advanced Server Solutions
Continuing with its innovations, Intel showcased its Clearwater Forest chip, also known as the “Xeon 6+”, built with 12 compute tiles utilizing the 18A technology. This sophisticated chip utilizes advanced packaging techniques with Foveros 3D and EMIB to consolidate its various chiplets.
Though the front displays only four chiplets, the central three chiplets house four compute tiles each, while the outer two cater to I/O capabilities. There’s also a base tile, providing additional cache integration.
Alongside the chip itself, the Clearwater Forest Compute Tile wafer was displayed, featuring numerous small chiplets, each incorporating 24 Darkmont E-Cores.
Server Solutions Derived from Clearwater Forest
Intel’s Clearwater Forest “Xeon 6+”CPUs are designed to fit into LGA 7529 sockets, which are identical to those used for Xeon 6900P chips, enabling seamless upgrades for existing Granite Rapids-AP environments.
Among the showcased servers supporting the Xeon 6+ solution was Lenovo’s CSP HD350 V4, a single-socket configuration offering 12 DIMM slots. We can also expect dual-socket (2S) configurations featuring as many as 24 DIMM slots around the launch period.
Live Demonstrations of Panther Lake & Clearwater Forest
Intel conducted several live demonstrations at its product showcase section, with numerous Panther Lake OEM and ODM laptops engaging in various benchmarks against Arrow Lake and Lunar Lake CPUs.
The first demonstration was a Multi-tasking Low-Power Island test, comparing the performance of its low-power island featuring 4 Darkmont E-Cores against those of Arrow Lake and Lunar Lake. This feature is instrumental in minimizing power consumption during light workloads involving applications like Microsoft Teams and YouTube.
The testing revealed that the Panther Lake system exhibited equivalent or lower power consumption compared to Lunar Lake and approximately 35% lower than Arrow Lake, averaging around 7.5-8.0W. In contrast, the Arrow Lake CPUs operated in the range of 10.5-11.5W, showcasing Intel’s impressive stride in efficiency.
Additionally, a gaming demo showcased the performance of Painkiller on an Intel Panther Lake 12Xe CPU with a 45W TDP configuration, running at 1080p “Epic”settings. The game performed admirably at native resolution with 50-60 FPS, but with XeSS 3 set to “Ultra Quality Plus”and MFG at 4x mode, the FPS skyrocketed beyond 200, offering an immensely smoother experience. Intel mentioned that an early version of the XeSS 3 MFG model was utilized, indicating that further enhancements can be expected upon official launch.
Another notable demo introduced the “Smart Power HDR”mode, showcasing how Intel’s Panther Lake CPUs can significantly reduce power consumption while offering HDR capabilities. This mode provides a roughly 20% reduction in panel luminance, translating to an extended battery life of 30-40 minutes under HDR conditions for SDR content. This is pivotal in addressing battery drains associated with HDR viewing, allowing for dynamic adjustments in voltage based on displayed content through an algorithm integrated within the SoC.
For Clearwater Forest, attendees experienced various demonstrations, including a 5G Control demonstration, where Clearwater Forest provided over twofold performance improvements compared to Sierra Forest CPUs for 5G Core workloads. In general compute tasks, Clearwater Forest showed remarkable improvements thanks to a 17% IPC uplift with Darkmont E-Cores and the doubling of core count (288 vs 144).
This experience provided tremendous insights into the potential of these upcoming technologies. Panther Lake is poised for its official unveiling at CES 2026, while Clearwater Forest is set to emerge around mid-2026, promising even more exciting developments and insights in the upcoming months.
Related Articles:
Radeon RX 9070 XT Outperforms RTX 5070 Ti in Call Of Duty: Black Ops 7, Achieving 31% Higher Performance at 1440p
15:55
AMD Launches Next-Gen RDNA GPU Architecture with Radiance Cores, Neural Arrays & Universal Compression for Enhanced Ray Tracing, Improved Upscaling, and Reduced Bandwidth Requirements
13:51
Intel’s 18A Node Hits Record Low Defect Density, Ready for Internal and External Customers
13:40
Leave a Reply