Intel has hosted a virtual preview that takes a look at the AI capabilities that Intel’s next-generation processors, such as Meteor Lake, will bring to laptops later this year.
AI and chatGPT have been hot topics in the tech industry for the past few years, leading some to question what AI can help us do other than improving Snapchat filters and writing our school essays for us (I mean, what?). CGMagazine was invited to a special virtual preview with Intel and their Client AI team to hear about their latest processor coming to the laptop market later this year: Meteor Lake. The main points discussed were how the specific architecture of the Meteor Lake processor has changed with the use of dedicated Vision Processing Units (VPUs) on board its chips to implement its AI features.
The team spoke briefly about the recent success of their Alder Lake and Raptor Lake processors and what they were able to achieve. In the past, they emphasized that they wanted to create a processor that would be able to deliver more processing power with less power consumption. In addition, they wanted a chip that could also improve AI capabilities by having a localized, universal dataset of inference knowledge – meaning that you don’t need the internet to build a dataset for what you need to create with AI models, which also allows for lower latency.
What is Meteor Lake?
Meteor Lake is a re-engineered Intel 4th generation PC processor that features a “chiplet” system-on-chip (SoC) design that allows Intel to deliver advanced intellectual property (IP) and enables users to optimize their PCs by increasing performance where it is needed without consuming too much power. Fovoros 3D packaging technology also makes this design possible, with its “monolithic” design that allows users to manage power and latency within each chip and node.
Meteor Lake is unique because it is Intel’s first PC platform to feature an onboard Neural VPU, a dedicated AI engine integrated directly on the SoC to run AI models efficiently. It is also Intel’s first client IBM 2.0 processor. The use of the VPU and AI accelerators would even help significantly improve the CPU and GPU performance. This could mean less time waiting for graphics to load in Cyberpunk 2077!
What Makes Meteor Lake Better?
The team highlighted three key points that make Intel’s Meteor Lake better than its predecessor, Raptor Lake. Here is a list of them:
- Next-Gen Power Management — Again, this came down to the SoC design with Fovoros 3D Packaging Technology. Having the versatility of utilizing specific chiplets at lower power is a major innovative leap.
- New, Built-in Intel Arc Graphics — Intel Arc GPUs were making waves last year with newer laptops. Now, they were able to implement dedicated software with modern feature sets of graphics, including DX12 Ultimate, ray tracing and Intel Movidius.
- New, Integrated AI Engine — Meteor Lake introduces the first integrated AI engine for local inference on the PC, which the Intel team believes is a new IP that can help reshape the PC user experience.
Bringing Meteor Lake Closer to Home
Intel’s VP and GM of Client AI, John Rayfield, led the rest of the preview by talking about how this next-generation processor would directly help its customers in terms of newly added AI computing efficiencies. He explained the shift from cloud-based AI computing workloads to client-based AI computing workloads. ‘Of course, cloud-based computing has a lot of positives with massive scalability, but it has myriad issues with high latency, privacy concerns, and can be quite expensive.
With a client-based solution using AI computing, this AI system has a massive distributable scale and low latency because it would be local. And because it is local, the likelihood of privacy breaches is reduced. It would also be less expensive to manage than paying for cloud space; it’s all integrated. As Derek Zoolander and Hansel said, “It’s in the computer!” But you might not want to smash the whole laptop or PC on the floor to check out the Meteor Lake processor.
Growing the Client AI Ecosystem
Rayfield also explained how the client AI ecosystem would be evolving as well, especially with the new Neural VPU implementation alongside the CPU and GPU in this chip design. He added how Intel has already been collaborating a lot with bringing better and more tools to enable a broad x86 application ecosystem, and unlock new AI experiences too. They have four tools in the works ONNX, OpenVINO, Web Neural Network API and Microsoft DirectML.
One example of their collaborative efforts was the differences of using a VPU chip versus a one without it, using the Advanced Blur feature in Windows Studio Effects. The team discovered an improvement of Background Concealment and Dynamic Noise Suppression with Advanced Blur and a VPU—which did more computing at lower power consumption. They also explored enhanced features across Adobe apps, providing less time waiting on renders and more time creating.
Rayfield also discussed how their Intel team was working on seeding open-source projects, with names like Blender, Audacity, GIMP, OBS and Unreal Engine tossed into the mix. The example they used with what they found was exciting was finding out how having Meteor Lake’s AI features to generate an image through a text prompt in OpenVINO, using an open-source GIMP plug-in.
More specifically, they used Stable Diffusion with enhanced performance to optimize the best image generation possible. It was funny to see what “an astronaut riding a horse on the moon” looked like through machine learning and AI generation. Still, it was really crisp and looked like a nice nod to claymation or stop-motion animation works like Kubo and the Two Strings or Guillermo Del Toro’s Pinocchio.
Looking to the Future
The Intel team touched on how they envision Meteor Lake as a pivotal step towards their chipsets’ future in designs and performance focuses, along with improving and implementing AI capabilities. They have already teased other chips like Arrow Lake, Lunar Lake, and would only continue to grow. Of course, Lunar Lake was said to still be in the works for 2024 and would be even more power efficient. AI and machine learning can be scary, but Intel shows the benefits of how they can be used to make our lives easier by cutting down computing time and allowing us to see what other enhancements can be made to already great apps.
