The advent of neural processing units (NPUs) may be hailed as a game-changer for personal computers (PCs), but finding the right balance in product design remains a challenge to give consumers what they want. It is important.
The growing demand for artificial intelligence (AI) tools, including generative AI (GenAI), in the workplace requires industry players, such as chip and device manufacturers, to ensure their products are ready for this change.
Also, what is an AI PC? (And should you buy one?)
In particular, as enterprises move AI workloads from the cloud to client devices, PCs need to keep pace. This migration improves performance by eliminating the need to send AI workloads to the cloud and back to users' devices. Businesses can also increase data privacy and security by keeping data on the device rather than moving it to and from the cloud.
AI tasks that run locally on a PC are typically performed by the CPU (Central Processing Unit), GPU (Graphics Processing Unit), or both, as neither is optimized to handle AI tasks. , your device's performance and battery life may be reduced. That's where AI-specific chips, or NPUs, come into play.
User requirements continue to focus on hybrid work
During the global pandemic, user requirements revolved around the ability to work remotely and support a geographically dispersed workforce. This improves built-in cameras and video conferencing capabilities, said Tom Butler, Lenovo's executive director of commercial portfolio and product management and head of the hardware maker's commercial notebook portfolio.
This has led Lenovo to focus on improving camera and video quality, as well as building microphones that support noise segmentation. Butler told ZDNET that users were also looking for good laptop performance, including long battery life.
Related article: Nvidia makes AI PC claims at CES 2024
With the majority of the world's workforce still working remotely or hybrid, user requirements continue to focus on performance, battery life, and video conferencing capabilities. The rise of AI PC has led market players to consider offloading some workloads such as noise canceling and background blurring functions to his NPU.
Butler said finding the best way to distribute workloads is a key goal now that PCs have three core engines, and Lenovo's new commercial laptops all come with an NPU. He added that “We are effectively moving into the era of AI PCs,” he said.
AMD CEO Lisa Su also described AI as “the evolution of PC gaming” in her keynote speech at SXSW this week.
“This is the beginning of significantly increasing our productivity,” Su said, adding that with products on the market today, the AI PC era has already arrived. He also said AMD is working to optimize its chips for AI, including building “better software” to improve productivity.
Related article: A new test of speed comes in the age of AI PCs
Asus Chief Operating Officer Joe Hsieh said in an interview with ZDNET that 2024 is the year customers will start asking whether their devices are AI-enabled. Like Lenovo, Xie said ASUS is also working to ensure new products include core chip components for NPU and AI workloads.
ASUS is also focused on developing the software engines and tools that users need to train their own AI models, Xie said, noting that most large-scale language models are currently only trained on public data. he pointed out. Asus believes that as personal devices handle his AI workloads, users will want these applications to use their data rather than public data.
Optimizing hardware and software for AI
When asked what requirements are the most difficult to balance as demand for AI PCs increases, Butler said that in addition to longer battery life and increased performance, the desire for devices to be thinner and lighter I pointed out the usual trade-offs between.
Improving performance will inevitably impact battery life and vice versa, he said. “With NPU, [now available]However, this allows you to offload some of the workload that would traditionally tax the GPU or CPU. ” he said. For example, the noise canceling function can be transferred to her NPU. Software He said vendors are considering: How to optimize your code to take advantage of NPU.
Meanwhile, Asus' vice president and co-head of the AIoT business group, Albert Chang, said Asus wants to provide tools to help developers choose the right computing resources. An application developer needs to be able to determine whether an AI tool needs to be equipped with a CPU, NPU or integrated GPU, Chan said.
The coming wave of AI PCs
Forrester predicts that by 2024, 60% of enterprises will be delivering GenAI-powered applications to their employees and customers.
Research companies also predict that demand for PCs will increase due to AI. IDC predicts that AI PC shipments will increase from 50 million units in 2024 to more than 167 million units in 2027, accounting for nearly 60% of total PC shipments. I predict that it will. IDC defines an AI PC as a personal computer with system-on-chip capabilities built to perform his GenAI tasks on the device.
Also: Nvidia's new AI chatbot runs locally on your PC and is free
IDC also predicts that 265.4 million PCs will be shipped worldwide in 2024, a 2% increase from 2023 due to the introduction of AI PCs. Shipments of these systems are expected to reach 292.2 million units in 2028, with a compound annual growth rate of 2.4% from 2024 to 2028. “The presence of on-device AI capabilities is unlikely to lead to an increase in the PC installed base, but it certainly will lead to higher average selling prices,” says IDC's Global Mobile and Consumer Research for his device tracker. He is the manager, says Mr. Jitesh Ubrani.
Gartner predicts that by the end of the year, 54.5 million AI PCs will be shipped and the number of GenAI smartphones will reach 240 million. These AI devices are expected to account for 22% of all PCs and 22% of basic and premium smartphones by 2024. The research firm defines an AI PC as a PC with dedicated AI accelerators or cores, NPUs, accelerated processing units (APUs), or tensor processing units (TPUs) to optimize and optimize AI tasks on the device. Designed to be fast.
Also: Microsoft confirms the following Windows, Surface, and AI events.Here's what you can expect
IDC classifies AI PCs into three groups. The first is a PC with an NPU that clocks up to 40 tera operations per second (TOPS) and enables certain AI capabilities in applications running locally on the device. Chipmakers currently shipping such NPUs include AMD, Intel, Apple, and Qualcomm.
According to IDC, the second category consists of next-generation AI PCs with NPUs that deliver 40 to 60 TOPS and have “AI-first” OSes that support “persistent and pervasive” AI capabilities. It is said that there is. AMD, Intel, and Qualcomm are expected to start releasing these chips in 2024, and Microsoft is expected to announce updates to Windows 11 to take advantage of these NPUs.
The final category of advanced AI PCs, which offer NPU performance in excess of 60 TOPS, are not currently in any vendor's announced product pipeline.