IT Home News on November 20, at the Microsoft Ignite Developer Conference, Intel and Microsoft announced that they will cooperate to optimize DirectML for Intel Arc graphics solutions.
According to experimental results, Intel achieved a 2.7x performance improvement in the AI image generator Stable Diffusion after updating the Arc Alchemist driver
While some of this improvement is due to the Intel driver team with its excellent traditional optimization techniques, the majority of the improvement is due to Microsoft's Olive model optimization tool.
Officially stated that Intel has been cooperating with Microsoft to optimize DirectML for various types of Arc graphics solutions, including independent graphics such as Arc A770 and the Arc core display built into the upcoming Core Ultra mobile processor.
Microsoft said that one issue developers face when bringing AI capabilities to client systems is ensuring that models fit and run well within the constraints of consumer PC system configurations. To help solve this challenge, Microsoft last year released the open-source Olive model optimization tool.
Recently Microsoft has updated Olive, focusing on some new AI models. These include Stability AI’s Stable Diffusion XL text-to-image generator and Meta’s Llama 2 large-scale language model. IT House discovered this information
As an example, Intel found that the Intel Arc A770 can run up to twice as fast as the default model when running ONNX's Stable Diffusion 1.5 with the Olive-optimized version of the DirectML execution provider.
To further improve performance, Intel has also introduced a highly optimized multi-head attention (MHA) meta-command to its driver, which can more fully exploit performance in models like Stable Diffusion
Thus, in the latest version of the driver, Stable Diffusion version 1.5 on the Intel Arc A770 GPU provides up to 36% performance improvement over the previous version, resulting in a 2.7x cumulative performance improvement.
The above is the detailed content of Intel Arc Alchemist driver update significantly improves AI image generation performance. For more information, please follow other related articles on the PHP Chinese website!