Background

AI's Presence In Many Computer Workloads Is Overwhelming CPUs, GPUs, And The Cloud, Said Intel

05/05/2023

AI is the hype of technology, and according to chipmaker Intel, the technology is a burden to existing hardware.

Ahead of the 2023 Computex, where Intel is used to make a big splash at the annual conference, Intel Veep and general manager of Client AI John Rayfield said that dedicated AI silicon is needed because AI is now present in many PC workloads.

Video conferences, he said, feature lots of AI enhancing video and making participants sounds great – and users now just expect that PCs do brilliantly when Zooming or WebExing or Teamising.

Furthermore, games use lots of AI, and so do generative AI models, which are already popular on PCs and available as local executables.

It was during this moment, that Intel announced what it calls the 'VPU', the technology it acquired along its acquisition of Movidius in 2016.

Intel VPU
​Intel touted its upcoming Meteor Lake client PC processors as the first PC platform from Intel featuring a built-in neural VPU, a dedicated AI engine integrated directly on the SoC to power efficiently run AI models. (Credit: Intel Corporation)

By upgrading all of its forthcoming Meteor Lake client CPUs to utilize the so-called 'Vision Processing Unit', Intel hopes to get to handle “sustained AI and AI offload."

With it, Intel hopes that AI services, either quietly embedded in apps or running using dedicated apps, can running locally.

Both CPUs and GPUs shall remain tasked with the heavy liftings.

CPUs for example, should still be asked to do simple inference jobs with low latency, usually when the cost of doing so is less than the overhead of working with a driver to shunt the workload elsewhere, whereas GPUs should still get to do jobs involving performance parallelism and throughput.

VPUs on the other hand, are only meant for AIs to utilize.

Intel wants to use VPUs to only process AI-related works.

The company introduced VPUs as solution because shifting AI workloads to the cloud is expensive, and also impractical because buyers want PCs to perform, not their internet connectivity.

With the advent of VPUs, Intel hopes that its Meteor Lake can emerge as SoC that uses Intel’s Foveros packaging tech, to work alongside CPU and GPU.

2023 Computex
2023 Computex in Taipei, Taiwan, held between May 30 and June 2.

Intel announced VPUs in Meteor Lake twice in a week.

First it generalized about the presence of VPUs in the platform at Microsoft’s Build conference. And ahead of Computex, it announced news that all its Meteor Lake product shall include the accelerator.

At 2023's Computex, Intel showcased the importance of the whole PC ecosystem, and how the company plays an integral role in the overall industry.

Intel used the event to further highlight the importance of the industry, especially after PC sales are cratering after the COVID-19 pandemic-induced pop.

But thanks to WHO that called the pandemic off, the 2023's Computex is special because it was the first full-scale physical incarnation of the event after four years.

And following the announcement of Microsoft Windows 11, the overly hype generative AIs, and the high demands in games, Intel is certain that the fragile PC economy shall recover.

Regardless, Intel follows the same ingredient, over and over again, because for more than many times, Intel has said that Meteor Lake is a huge change to everything about the PC that should make the devices even more compelling, make the world a better place, and represent a revolution.