AI on your PC: Intel unveils new line of Core Ultra processors with NPUs for localised AI processing

Intel has unveiled its highly anticipated Intel Core Ultra processors, code-named Meteor Lake, during the Intel Innovations 2023 event held in San Jose, California.

This release marks a significant milestone as it introduces Intel’s very first integrated neural processing unit (NPU), designed to supercharge AI capabilities on personal computers while prioritizing data privacy.

Intel has officially confirmed that the Core Ultra processors are set to hit the market on December 14.

AI becomes personal, localised
The upcoming Core Ultra processors are set to usher in a new era of AI-powered personal computing. These chips are meticulously engineered to deliver low-latency AI computing capabilities, irrespective of connectivity issues, and they come equipped with advanced data privacy features that are increasingly crucial in today’s digital landscape.

One of the most noteworthy innovations lies in the integration of an NPU directly into the silicon. This NPU is specifically tailored to offer entirely fresh and transformative PC experiences.

It is particularly well-suited for workloads that previously relied on the CPU, either to enhance quality or efficiency and for tasks that would typically be offloaded to the cloud due to the lack of efficient client-side computing resources.

Intel’s Core Ultra represents a pivotal moment in the evolution of client processors, thanks to its distinctive client chiplet design enabled by Foveros packaging technology.

Move over CPUs, NPUs are here
Beyond the introduction of the NPU, these processors also incorporate significant advancements in power-efficient performance, made possible by Intel 4 process technology. Additionally, the Core Ultra processors boast graphics performance akin to discrete-level graphics, thanks to the incorporation of onboard Intel Arc graphics.

To delve into the specifics, let’s clarify what an NPU, or Neural Processing Unit, is. It is a specialized hardware component meticulously designed to accelerate and execute artificial intelligence (AI) and machine learning (ML) tasks with exceptional efficiency.

What sets it apart from general-purpose processors like central processing units (CPUs) and graphics processing units (GPUs) is its purpose-built nature, optimized for handling the intricate mathematical calculations inherent in neural networks, including matrix multiplications and convolutions.

These operations form the backbone of deep learning algorithms, which find applications in tasks like image and speech recognition, natural language processing, and recommendation systems.

What makes NPUs the next step in mass computing
Efficiency: NPUs excel in the execution of AI workloads, often consuming significantly less power compared to CPUs or GPUs when performing similar tasks. This efficiency is particularly invaluable in battery-powered devices such as smartphones and laptops.

Low Latency: NPUs are meticulously designed for low-latency processing, making them ideal for real-time and time-sensitive AI applications, including autonomous driving and robotics.

Specialization: In stark contrast to general-purpose processors, NPUs are finely tuned for AI tasks. This specialization empowers them to deliver superior performance and energy efficiency when handling neural network computations.

Parallelism: NPUs are outfitted with multiple processing cores or units capable of handling parallel computations, a fundamental requirement for tasks related to neural network training and inference.

Inference Acceleration: NPUs are commonly employed for AI inference, where pre-trained models make predictions based on input data. They accelerate these inference tasks, enabling faster and more responsive AI applications.

NPUs find integration across various types of devices, spanning from smartphones and smart home appliances to data centre servers and edge computing devices. Their pivotal role lies in democratizing AI by enhancing accessibility and efficiency across a wide spectrum of applications.