Intel not only has high-performance x86 processors, but also entered the high-performance GPU market this year. At the same time, there is a killer, that is, it has mastered AI chips with multiple architectures. Intel is also paying more and more attention to AI capabilities. The 14th-generation Core will add a VPU unit to greatly improve AI performance and reduce power consumption.
Intel has been attaching importance to Core processor AI for several years. The 10th-generation Core Ice Lake added AI functions, which were implemented through the GNA accelerator at that time and supported a dedicated AI instruction set. The latter 11th and 12th generations of Core Duo AI capabilities are also being continuously enhanced.
Starting from the 13th generation Core at the end of this year, Intel will strengthen AI again and use the capabilities of its own Movidius AI chip, and on the 14th generation Core Meteor Lake next year,Intel will use multi-chip integrated packaging for the first time, taking this opportunity to directly add a new functional unit VPU, which is the visual computing unit.
This VP is based on the third generation Movidius VPU, codenamed Keem Bay,It is specially designed for AI acceleration in low-power scenarios. Previously, Intel data showed that its performance is between 3T and 7.1TFLOPS, which is 10 times that of the previous generation, and the power consumption is only about 6W. .
After integrating the VPU, not only the AI performance has skyrocketed, but it can also liberate the GPU. At present, many AI acceleration work is realized through the GPU.On the 14th generation Core, work such as image recognition can be transferred from the GPU to the VPU, reducing GPU power consumption.