Intel Arc Alchemist driver update, AI image generator Stable Diffusion performance ushered in a big improvement
CTOnews.com, November 20, at the Microsoft Ignite developer conference, Intel and Microsoft announced that they will work together to optimize the DirectML of Intel's Arc graphics solution.
The results show that Intel has achieved a 2.7-fold performance improvement in the AI image generator Stable Diffusion after updating the Arc Alchemist driver.
Although part of the improvement is due to the Intel driver team's excellent traditional optimization technology, most of the improvement is due to Microsoft's Olive model optimization tool.
Intel has been working with Microsoft to optimize DirectML for all types of Arc graphics solutions, including unique features such as the Arc A770 and the Arc kernel built into the upcoming Core Ultra mobile processor, officials said.
Microsoft said that one of the problems that developers face when introducing AI capabilities into client systems is to ensure that the model adapts and runs well under the limitations of consumer PC system configuration. To help solve this challenge, Microsoft released an open source Olive model optimization tool last year.
CTOnews.com found that Microsoft has also recently updated Olive, focusing on some new AI models, including Stability AI's Stable Diffusion XL text to image generator and Meta's Llama 2 large language model.
For example, Intel found that the Intel Arc A770 can run twice as fast as the default model when running Stable Diffusion 1.5 of ONNX with the Olive optimized version of DirectML execution provider.
To further enhance its performance, Intel has also introduced a highly optimized long attention (MHA) meta-command for its driver, which can extract more performance from models like Stable Diffusion.
As a result, in the latest version of the driver, Stable Diffusion 1.5 on the Intel Arc A770 GPU provides up to 36% performance improvement over previous versions, resulting in a cumulative performance improvement of 2.7x.
