Getty Images/iStockphoto
Intel Core Ultra CPUs with neural processing runs AI on PCs
Intel's Core Ultra CPUs now contain embedded AI neural processing, which adds options for device manufacturers to divide demand for AI processes among different hardware resources.
Intel released Core Ultra CPUs Thursday, the first chips to include onboard neural processing units that allow laptops to take on specialized AI tasks.
The chipmaker said that some manufacturers offer the Core Ultra now; the new chips will be available in more than 230 different laptop models in the next year.
Neural processing units (NPUs) in the Core Ultras, previewed last September and codenamed Meteor Lake, differ from GPUs in that they handle specific AI tasks -- such as background blurring in video meetings, simultaneous translations into multiple languages, or noise reduction in phone calls. GPUs, on the other hand, are more general-use and can handle many different AI tasks.
While NPUs are more like sprinters, they can only handle specific AI loads. GPUs might be slower, but they can take on more varied work. Both will be needed to work together as AI demands more processing power to get jobs done, said Intel CEO Pat Gelsinger at a release event Thursday.
"AI and high-performance computing are coming together in a powerful way to enable this next generation of computing," Gelsinger said. "Faster memory, faster networking, larger memory capabilities [and] computing capabilities…in a sustainable way."
Advice for CIOs, tech buyers
The blast of price and performance information on AI processing power can be overwhelming as tech buyers plan their 2024 budgets. It may seem difficult to determine which laptops and other end-user machines will stay relevant during this period of AI advancement.
Gartner analyst Chirag Dekate offers one rule: Buy the most processing power you can afford, because those who cut corners may find themselves replacing laptops sooner than later.
"They might want to align product roadmaps of multiple vendors to select the right products," Dekate said. "The highest performing [units] would deliver the best value and best longevity."
Chipmakers are quickly innovating to accommodate AI at endpoints. Buyers should take their time to understand the various manufacturer roadmaps, and which of their various processor/GPU combos suit an individual company's operations.
After such an analysis -- that should include direct comparisons on common metrics -- it may turn out that waiting a month for the next round of chips to come out from a particular manufacturer is advantageous.
Don Fluckinger covers digital experience management, end-user computing, CPUs and assorted other topics for TechTarget Editorial. Got a tip? Email him here.