Intel executives confirmed Tuesday that the company will make a concerted effort to bring AI capabilities to PCs with its next next-gen CPU cores, codenamed Meteor Lake.
Both Intel Chief Executive Pat Gelsinger and an Intel colleague Rajshree Chabukswar have confirmed that there will be AI capabilities coming with Meteor Lake, which will likely be named the 14th-gen Core chip. Intel formally revealed the enthusiast desktop version of its 13th-generation Core chip (Raptor Lake) this week at its Intel Innovation conference in San Jose.
AI has proven to be a powerful business opportunity within the enterprise, with apparently most corporate chip announcements emphasizing its inference ability. AI hardware can help train algorithms to aid in visual recognition, predictive capabilities, and more. And on smartphones, AI is used to filter out background audio and help distinguish the foreground of a photo when taking pictures.
On PCs, however, its effects are more subtle. It can also filter out background noise when using Windows’ own Alexa or Cortana app – which, unfortunately, consumers apparently rarely do. Applications like Photoshop should take advantage of AI capabilities in terms of photo editing. And the recent Windows 11 2022 update includes some smart effects to help set you apart in video calls, like in Teams.
Still, AI isn’t as much of a priority as, say, performance, battery life, or any other feature. That looks like it might change, though.
‘Essential AI Features’ Are Coming to Meteor Lake
When asked what Intel was doing to make AI a PC selling point, Gelsinger jokingly replied, “I can’t say yet because we haven’t released Meteor Lake yet.”
Gelsinger, however, went on to describe what he believes Meteor Lake will be able to accomplish. “One of the things we’re going to bring up is what I would say are the core AI features,” he said. “I’m sitting on a Zoom or Teams [call]. Do I want to have simultaneous translation or contextualization of this call? Yup. Do I want to have cameras that are following the speaker and performing image enhancements? Yea.”
Gelsinger compared this to the early days of X86 processors, when chips lacked a floating-point unit for mathematical calculations – the distant, conceptual ancestor of today’s GPUs. “People would say, we don’t really need floating point to do this,” he said. “And that just became standard… Well, AI will be like that, standard and delivered on all our platforms.”
The difference, said Gelsinger, is that consumer AI would likely be optimized differently for consumer PCs than it would be in the data center.
Mark Hachman/IDG
In a presentation Tuesday afternoon, Intel’s Chabukswar said that Intel planned a “new AI acceleration engine that is coming soon.” “This engine will focus on delivering amazing AI-powered experiences that will enable great system and application responsiveness,” she said.
Chabukswar confirmed to PCWorld after her presentation that she was talking about Meteor Lake.
Chabukswar also said that the new AI accelerator will be supported by Microsoft, though in what capacity she doesn’t say. However, in a presentation Intel apparently gave on its Tech Tour in Israel, Sweclockers photographed an endorsement from Microsoft’s Vivek Pradeep, who manages the computer vision team at Microsoft’s Applied Sciences. The AI block, however, will apparently be known as the VPU, codenamed “Movidus”.
“Microsoft believes that neural processing units such as Intel’s VPU represent an inflection point in computing and will be key to delivering a new range of delightful experiences to Windows users on their PCs,” Pradeep wrote, adding that Microsoft was “in close partnership” with Intel on the VPU.
Rumors and leaks of various types surrounding an AI block in next-gen Meteor Lake have been swirling for nearly a year, with a driver update seemingly confirming the AI engine, followed by a more detailed leak of the next-gen architecture itself. .
Interestingly, Intel would not be the first to try to popularize AI on the PC. Intel tried to popularize AI with its 10th-gen Core, Ice Lake, followed by Qualcomm’s efforts with its Snapdragon processors. None, however, were successful.