Intel's 18A Process Makes Its Debut: In-Depth Look at Panther Lake - New Architecture to Embrace the AI Wave
Located in Phoenix, Arizona, USA, stands Intel's most advanced wafer fabrication plant. In the last few days of September, Lei Technology was honored to be invited by Intel to participate in its annual technology event, the ITT (Intel Technology Tour). We flew across the Pacific Ocean to this beautiful desert oasis city to visit and learn about Intel's latest technological achievements.
Image source: Lei Technology
During the visit to the Intel factory, Intel senior chief engineer Arik Gihon, Intel Fellow Carlos Cordeiro, Intel vice president Chris Auth and others explained to us Intel's upcoming Panther Lake architecture, revealing the technological secrets behind it.
Panther Lake Makes a Splash: The Debut of Intel 18A Process
Panther Lake is the first consumer - grade product platform manufactured using Intel's Intel 18A process. As Intel's most advanced mass - production process at present, the Intel 18A process has two brand - new features: RibbonFET gate - all - around transistors and PowerVia back - side power delivery.
Image source: Intel
Among them, RibbonFET gate - all - around transistors are a new technology prepared by Intel for processes of 2nm and below. It replaces the traditional FinFET architecture with a new transistor architecture, avoiding problems such as leakage and unstable performance caused by process upgrades.
Many friends may have questions about this: Why do process upgrades lead to leakage and unstable performance? You can think of the CPU as a city, and transistors as the buildings that make up this city. There are channels between the buildings for current to pass through and reach the corresponding transistors.
However, as the process technology is upgraded, the gaps between the buildings become narrower (the transistor density increases). At this time, there is a probability that the current will take a shortcut directly, jumping over the building to the adjacent channel. This not only causes the transistors that are supposed to receive the current to lose power supply, but also causes a surge in the current in the adjacent channel, which in turn affects the normal operation of the corresponding transistors and may even directly break down the transistors, resulting in CPU damage.
Image source: Intel
To solve this problem, Intel designed the RibbonFET gate - all - around transistor technology. It makes the current channels into long and thin "nanowires/sheets" and then surrounds the current channels 360° with the gate (the current switch valve) to control the moving direction of the current to the greatest extent and significantly reduce problems such as leakage.
In addition to RibbonFET, Intel also introduced the PowerVia back - side power delivery technology in Panther Lake. It moves the power supply module originally placed on the front side of the chip to the back side. This technology not only reduces the heat dissipation pressure on the front side of the chip, but also allows the current to enter the transistors more efficiently.
Image source: Intel
This revolutionary technology can significantly reduce power supply losses, enabling the CPU to run at full power with lower power consumption. At the same time, it can also bring a higher performance ceiling (because the power supply is more sufficient and the heat dissipation pressure is reduced). Coupled with the upgrade of the RibbonFET gate - all - around transistors, it enables the Panther Lake processor to achieve higher performance with lower power consumption and the performance is more stable.
These characteristics make it very suitable for use in mobile platforms, and also make the Intel 18A process a key part of Intel's "5 nodes in 4 years" strategy. This is why the Panther Lake architecture has attracted so much attention, as it will mark the beginning of a new era for mobile PCs.
Panther Lake Becomes a Milestone: New Architecture, New Cores
Panther Lake brings not only the Intel 18A process but also a brand - new architecture design. The new architecture shifts from the traditional monolithic design to a multi - chip design, that is, dividing the processor into multiple functional chip modules and then integrating them onto a single package substrate using Foveros 2.5D packaging technology.
The modular design allows Intel to design and manufacture chips like building Lego blocks. It saves costs and also enables engineers to customize the processor design according to the usage requirements of different products. For example, for processors in thin and light laptops, a smaller - scale GPU module can be used to reduce overall power consumption, while for high - performance laptops, a larger - scale GPU module can be used to improve overall performance.
Image source: Intel
The flexible design of Panther Lake will have great potential in the future AI PC market, as it can meet the computing needs of various types of devices. From personal servers to handheld terminals, the most suitable architecture can be built through modular design, and the design cost is also more controllable.
CPU: Higher Performance, Higher Efficiency
In addition to adopting the new - generation modular design, Panther Lake also comprehensively upgrades the big - little core architecture, introducing brand - new performance cores Cougar Cove and efficiency cores Darkmont. Cougar Cove is actually the performance core of the next - generation Nova Lake architecture. Compared with the previous large cores, it has significant improvements in IPC, clock frequency and other aspects. You can look forward to the performance of this performance core in high - load applications such as games.
Image source: Intel
It is worth noting that Panther Lake also introduces a brand - new technology - Memory Side Cache. Simply put, this is a high - speed cache located between the memory and the CPU. It is like adding a "small warehouse" beside the processor. It specifically caches data that is frequently used recently, relieving the pressure on the memory bandwidth.
This new technology is the key to the efficient operation of an integrated architecture like Panther Lake. On the one hand, it can improve the utilization rate of the memory bandwidth, ensuring that the cores of the CPU and GPU do not slow down due to memory read - write bottlenecks when working together. On the other hand, it optimizes latency, reducing the waiting time for the computing cores to access data and enabling the processor to respond more quickly to various normal requirements.
In addition, Panther Lake also brings an upgraded Intel Thread Director thread scheduling technology. Through hardware monitoring and AI algorithms, it can more precisely identify the nature of the currently running load and intelligently assign threads to the most suitable units such as performance cores, efficiency cores or even the NPU for execution.
Image source: Intel
Compared with the previous generation, it has a deeper understanding of the performance/power consumption characteristics of different cores and also adds scheduling optimization for multi - chip and multi - heterogeneous units. It allows the processor to balance performance and energy efficiency, maximizing battery life without sacrificing the actual experience.
To further improve battery life, Panther Lake also introduces global power management. For example, when the GPU enters high - load rendering, the processor will intelligently reduce the power consumption budget of unused units (CPU, NPU) and allocate power to the GPU, and vice versa.
Image source: Lei Technology
This SoC - level energy scheduling, combined with more advanced processes and packaging heat dissipation, will be the key for Panther Lake to have longer battery life and better heat control. In the past, we often said that performance and battery life were like fish and bear's paws, and you couldn't have both. But with Panther Lake, we may really be able to "have both".
GPU: A New Brain for AI and Gaming
In terms of the GPU, Panther Lake is equipped with the brand - new Xe3 Iris Xe graphics (codenamed Celestial), which is the first appearance of Intel's 3rd - generation Xe architecture graphics unit. Compared with the previous generation Xe2 (Battlemage), Xe3 has undergone a major upgrade in architecture, with an increase in the number of cores and the frequency ceiling.
Image source: Intel
It is reported that the Xe3 Iris Xe graphics will come in two versions. The top - end version will have 12 Xe cores, a 50% increase in the number of cores compared with the previous top - end version. At the same time, the graphics performance has increased by more than 50% compared with Lunar Lake. These improvements enable the performance of the Xe3 Iris Xe graphics to rival that of entry - level discrete graphics cards. Actual tests show that after enabling XeSS/RT, it can even play large - scale 3A games such as "Cyberpunk 2077" and "Black Myth: Wukong" at high frame rates.
In addition to improving gaming performance, the AI performance of the Xe3 Iris Xe graphics has also increased significantly. The high - end version of the Xe3 GPU can provide up to 120 TOPS of AI performance, which alone exceeds the total computing power of the previous generation's CPU + GPU + NPU.
Image source: Intel
With the support of the AI performance of the Xe3 GPU, games and graphics rendering that require AI participation, such as real - time ray tracing (Panther Lake's integrated graphics support hardware - accelerated ray tracing) and AI - based image quality improvement technology, will have better effects and frame rates.
In addition, AI requirements such as high - quality graphics recognition and video enhancement no longer need to rely on the NPU or cloud computing power. The GPU can handle them directly, which is not only more efficient but also saves the NPU's computing power for other AI tasks.
NPU: Keeping AI Always On
Finally, let's talk about the NPU. As an important part of the new - generation processor, the fifth - generation NPU in Panther Lake has a significant increase in computing power, up to 50 TOPS. Although this is not high compared with the GPU's computing power, it is still several times that of the previous generation.
Image source: Intel
Moreover, the NPU consumes much less energy than the GPU, which enables the NPU to better support always - on AI tasks. For example, functions such as local voice recognition for voice assistants, real - time background blurring and noise reduction in video conferences, gesture operation recognition and user presence detection, security encryption acceleration, and even running a simplified generative AI model locally can all be handled by the NPU. This allows AI PCs to truly achieve both battery life and AI functionality.
It can be said that an AI PC powered by Panther Lake is a true complete form. It can not only keep AI functions always on but also has up to 180 TOPS (CPU + GPU + NPU) of AI computing power to assist users in handling various tasks, making AI truly a part of daily experience.
How Does Panther Lake Deeply Optimize for High - Frequency Scenarios?
When many people buy laptops, they basically only focus on CPU and GPU performance and often ignore parameters such as Wi - Fi and connection protocols. In fact, these seemingly insignificant parameters are sometimes the key to determining your daily usage experience.
Take the most commonly used Wi - Fi function as an example. Currently, Wi - Fi 7 is gradually becoming popular, but many laptops still only support Wi - Fi 6. Don't underestimate the difference of just one generation. There are significant differences between the two in terms of stability, speed, and latency.
Let me put it this