Intel Arc GPUs get 2.7 times faster in AI imaging with Microsoft feature

Smart Technology

Intel has released an update for the driver of its Arc video cards that almost triples its performance in Stable Diffusion, an artificial intelligence model that generates images from text descriptions. The technological leap is mainly attributed to Microsoft Olive, a new optimization tool from the software giant.

Olive is software that analyzes an artificial intelligence or machine learning model and finds ways that different hardware architectures can speed it up. With different models manufactured by Intel, AMD and NVIDIA, the optimization process is much faster thanks to the Microsoft tool.

With the help of the software, the manufacturer developed new instructions in its driver that improve the use of the Xe Matrix Extensions (XMX) cores of Arc video cards, which are responsible for accelerating tasks based on artificial intelligence.

The result is an Arc A770 that is up to 2.7 times faster in imaging with Stable Diffusion. Intel has not yet released data showing performance evolution using other video cards in the range, but we know that optimizations can be tried starting with version 31.0.101.4952 of the Arc driver.

Intel says it will continue to work in partnership with Microsoft to “usher in the AI ​​PC transformation with Windows 11 and beyond.” Rumors indicate that the hardware manufacturer will play a key role in the development of Windows 12, a supposed update to the operating system that should bring more aggressive use of AI.

NVIDIA and AMD also optimized the use of their cores for generative artificial intelligence tasks using Microsoft Olive, and through this tool, manufacturers also achieved significant increases, even doubling the performance of their video cards in image generation with the AI model.

This is another step towards the realization of AI PC, a concept that provides for the integration of artificial intelligence into the personal computer. In the future imagined by this idea, notebooks and desktops will be able to generate images, talk to the user and understand commands with natural language without depending on a cloud.