Intel demonstrated the work of a technical sample of a processor from the future Meteor Lake series in a task related to the operation of generative neural network algorithms at the Computex electronics fair.
Future Intel Meteor Lake processors will use a decentralized rather than monolithic design, allowing combinations of different chiplets based on different workflows. One of the main features of the Meteor Lake processors will be a special neural engine VPU (Versatile Processing Unit) based on the third generation of technology from Movidius, which Intel bought in 2016. This neural engine is designed to speed up the work of Meteor Lake processors on tasks related to various AI algorithms and machine learning models.
One of the basic tasks of the VPU in Meteor Lake processors will be the processing of photos and videos and the application of filters in real time. It can apply blur effects, apply auto scaling and replace the background of images. It also gains the ability to recognize user eye movements and gestures in real time. The performance of the Neural Engine should even be sufficient for motion capture using the Unreal Engine.
The VPU will be powerful enough to support generative AI functions, such as the operation of the stable diffusion neural network. The presence of a VPU is not required to work with such models. The main thing is that there must be enough memory. However, by using the hardware capabilities of the CPU, GPU and APU, the work of the AI model can be significantly accelerated. To take advantage of VPU capabilities, Intel is working with GIMP and Adobe to add support for the company’s neural engine to their products.
At Computex, Intel demonstrated a technical sample of the Meteor Lake mobile processor called the Intel 0000, which includes 16 physical cores with support for 22 virtual threads. According to the Wccftech portal, the chip will have a 6+8+2 core configuration and its final incarnation will belong to the Meteor Lake-P series. The configuration of the processor cores is very unusual. The main CPU chip has powerful P-cores and energy-efficient E-cores. And two additional cores are in the SoC die.
The chip ran at a frequency of 0.37 GHz in idle mode. The base frequency of the processor is 3.1 GHz. It has a total of 1.6MB L1 cache, 18MB L2 cache and 24MB L3 cache.
Meteor Lake engineering example demonstration was done in GIMP image editor with support of Stable Diffusion AI model to generate images according to given parameters and then AI plugin was applied on the result to do that increase resolution. To leverage the VPU, the Intel OpenVINO library was used to develop computer vision and artificial intelligence applications.
Meteor Lake mobile processors are expected to be released later this year. The company is not commenting on the latest rumors that the desktop Meteor Lake-S may have been discontinued. At the same time, the manufacturer has not yet confirmed its plans for the release.