2024.04.02 12:51
I'm PortAI, I can summarize articles.

Track Hyper | Yang Yuanqing: Lenovo is about to be the first to release AI PC

Upstream and downstream accelerating resonance

Author: Zhou Yuan / Wall Street News

With the advent of ChatGPT, the practical application of AI technology has taken a big step forward.

GenAI (Generative Artificial Intelligence) has become a market driver for companies upstream and downstream, truly showcasing the charm of GenAI to the public, and igniting a rare consensus and enthusiasm for technological advancement and commercialization in the industry.

On April 1st, Yang Yuanqing, Chairman and CEO of Lenovo Group, announced at the Lenovo Group 2024/2025 fiscal year oath-taking ceremony that on April 18th, the Tech World Lenovo Innovation Technology Conference will be held in Shanghai, where the truly meaningful AI PC with 5 major features will debut in the Chinese market.

What is an AI PC? What does it mean to have a dedicated AI PC processor? What are the differences between an AI PC processor and a traditional PC CPU? Why are chip manufacturers adopting a new architecture?

Yang Yuanqing stated that AI PC has five major characteristics. From embedded personal large model compression technology to using natural interactive intelligence, to personal knowledge bases; from local heterogeneous AI computing power containing CPU, GPU, NPU, to user data and privacy security barriers, supplemented by an open AI application ecosystem, all have relevant advantages.

Yang Yuanqing previously told Wall Street News that AI PC is a completely new category that is not compatible with existing PC systems. How do downstream terminal vendors such as HP, Dell, Apple, Lenovo, and Asus react to the technological changes in the hardware upstream of this new category? Why has the PC become the pioneer of intelligent terminals in the GenAI era?

XPU Architecture: The Source of AI PC Computing Power

In October 2023 at Lenovo Tech World, Yang Yuanqing first showcased the Lenovo AI PC and its powerful computing capabilities. The personal large model with a local knowledge base—AI Twin—became the focus of the industry.

On April 1st, 2024, at the Lenovo Group 2024/2025 fiscal year oath-taking ceremony, Yang Yuanqing emphasized, "The personal AI Twin will be able to store massive personal data and perform reasoning, greatly expanding our 'memory' and enhancing our thinking, judgment, and decision-making abilities. The AI Twin will surpass the physical body, permanently existing in the world, in a digital sense, 'immortality' may no longer be a fantasy."

The "immortality" feature is built on the foundation of the powerful computing power of AI PC. This computing power comes from the unique processor structure of AI PC. In fact, the upstream of PC terminals—chip designers—had already taken "collective" action by the end of 2023.

On December 14, 2023, Intel released the new generation Meteor Lake processor for AI PC; on December 15 of the same year, Intel launched the AI PC processor "Core Ultra" platform, with computing power 8 times that of the previous generation, currently adapted to over 10 domestic AI large models The computing architecture of Intel's Core Ultra adopts the XPU strategy: that is, the collaboration of CPU, GPU, and NPU.

Starting from Meteor Lake, Intel has introduced Al on a large scale in the PC field and provided software models and tools support for technology developers. Meteor Lake integrates NPU for the first time, enabling PC terminals to process AI workloads with lower power consumption, in line with the development trend of AI PCs.

With the built-in NPU AI engine, the Core Ultra processor can achieve low-latency AI local computing. This means that users can enjoy faster response times and better user experience when using AI applications such as speech recognition, image processing, and body language perception on their PCs.

As a long-time rival of Intel, AMD took the lead. On December 7, 2023, AMD released the Ryzen AI software platform, which adopts the new XDNA architecture for AI PC engines.

The Ryzen AI engine has been modified for traditional x86 architecture processors. This involves adding an independent NPU AI engine hardware unit to x86 processors. It is the AMD Ryzen AI engine that has created the new concept and application of AI PCs. This NPU unit has been integrated into the AMD Ryzen 7040 series PC processors released in early 2023.

The Ryzen 7040 adopts x86 architecture. In addition to the CPU (Zen4 architecture), it also includes a GPU (RDNA3 architecture) and the NPU of the Ryzen AI engine, which is consistent with Intel's Core Ultra computing architecture.

The reason why the Ryzen AI engine can enable laptops to perform AI workloads locally, without relying on the network and cloud, thereby reducing latency and protecting privacy; and can also run in edge-cloud hybrid scenarios, distributing tasks between cloud servers and local notebooks for collaborative acceleration, is due to its newly designed XDNA architecture.

On December 7, 2023, AMD announced the software integration package for the Ryzen AI engine, providing acceleration functions for dedicated creative design software; and also expanded the downstream application scope of this AI engine's ecosystem, including Adobe, Microsoft, Avid, Bori SFX, OBS Studio, iQiyi, and ByteDance, among others.

More importantly, AMD's launch of the Ryzen AI software platform also supports TensorFlow, PyTorch, and ONNX, the hottest AI large models at present: developers can fully leverage the powerful performance of the Ryzen AI engine to deploy AI applications and accelerations such as generative models with almost no barriers.

The core computing unit of AI PCs uniformly adopts the XPU architecture, which IDC refers to as heterogeneous hybrid computing: the main purpose is to provide strong parallel computing capabilities for PCs through heterogeneous computing collaboration Heterogeneous hybrid computing, using computing units with different instruction sets and system architectures to form a local computing system, matching corresponding computing units to different AI workloads. As a result, the leap in parallel computing power enables PCs to easily perform complex AI model inference tasks, demonstrating higher levels of intelligence.

This is the fundamental feature of AI PCs. Unlike traditional PCs with separate CPUs and GPUs, the computing structure of AI PCs lacks an NPU.

PC: The First Terminal of AI Technology Universal Period

The essence of AI PCs is still PCs. So, what value does a PC have in the future AIGC era? What role will it play? Is a PC more important than XR or smartphones?

"On June 9, 1975, I pressed a few keys on the keyboard and was shocked!" Apple co-founder Wozniak said, "Those letters appeared on the screen." Wozniak's action on that Sunday was considered a milestone moment in PC history.

Later, Wozniak said, "That was the first time in history." The keyboard was an accessory of the then-nascent Apple I, and the true significance of PCs originated from the birth of the first-generation Apple I.

In April 1977, Apple Computer, which was established on January 3 of the same year, launched the Apple II, not only moving Apple from Steve Jobs' garage to a luxurious office in a bustling area but also propelling Apple to its first peak after establishment; more importantly, it "incidentally" pioneered the PC industry.

Since then, every major product upgrade of PCs has been accompanied by significant technological innovations. In this process, PCs have played a pioneering role and become the best carrier of technology popularization.

From the perspective of user experience, PC terminal manufacturers have effectively integrated various innovative technologies, innovated products, and delivered them on a large scale, further accelerating the popularization process.

Bill Gates, the founder of Microsoft, proposed his famous grand ambition 48 years ago (in 1975) - "to have a computer in every home." This was indeed Bill Gates' personal business vision or life ambition, but history later proved that this vision met the objective needs of technological development and has truly become a reality today.

Looking back at the nearly 50-year history of semiconductor technology development, PCs are not only the core technology carrier terminals of the Internet era - even after Jobs introduced the first iPhone on January 9, 2007, ushering in the era of mobile Internet - but also have not been replaced in their core terminal status by the popularization of semiconductor technology.

Standing at the same historical moment and looking to the future, what hardware do humans, who have been immersed in the technological popularization of PCs for nearly 50 years, truly need, that is epoch-making and capable of embracing "super AI"? Is it smartphones? PCs? XR? Laptops? Or something else? Based on practical considerations, the future that can be seen now, the super AI terminal that humanity needs, with universal significance, should be a creative tool that supports human imagination to create the future and explore endless possibilities; it should also be a new intelligent device that can fully protect human data privacy, stimulate AI capabilities, and have a restraining effect.

From this perspective, smartphones do not represent the future, and even with AI capabilities at the edge, they are not the mainstream intelligent terminal of the future. Smartphones, like PCs, belong to the same era. The real value of smartphones lies in supplementing the lightweight application needs of PCs in mobile scenarios, while laptops can better meet heavy mobile workload demands than smartphones.

AI intelligent technologies represented by "GenAI" are about to or are already opening a new era of information technology.

Just as history has shown in the past, in this new era as the carrier of universal technology, from the perspective of cost, implementation path, and efficiency, the industry will naturally search for that slightly dusty pearl from the accumulated achievements of technological history - the PC, although it may not look as attractive on the outside, and may even seem a bit old-fashioned, but internally, it has undergone a transformation.

This is why, in the rapidly approaching future, the PC remains the primary form of AI universal technology terminal