By Rutvi Trivedi, Freddy Chiu, Harshal Vyas
With the recent launch of Intel® Core™ Ultra, we are embarking on a journey to empower everyone with AI on the PC. This is achieved by integrating Intel® AI Boost into every Intel Core Ultra processor, a powerful Neural Processing Unit to accelerate compute-intensive AI workloads. This enables more efficient AI integration into applications across segments such as creative, collaboration, productivity, and gaming. We are on a mission to democratize AI, one PC at a time. To accomplish this mission, Intel has partnered with Microsoft to deliver software capabilities at AI PC scale, bootstrapping the industry, and partners like Samsung to deliver AI experiences with simple APIs to unleash AI on the Samsung Galaxy Book 4 series AI PC.
With Intel Core Ultra, Intel’s AI Boost technology helps deliver higher quality AI experiences on the PC without compromising on system responsiveness, battery life, or other vital experiences on the PC. This is achieved by using Intel AI Boost for dedicated AI processing needs, freeing up Intel’s Core processors and Intel graphics for other system workloads such as gaming. At Intel, we recognize that hardware innovation must be accompanied by seamless software to deliver the best developer and user experiences. As such, Intel has partnered with the Windows AI team at Microsoft to unlock Intel AI Boost through the DirectML APIs, designing a scalable, yet performant API interface to power an NPU with DirectML for the first time. This powerful API makes it easy for AI application developers to deliver software to take advantage of AI acceleration.
Moreover, we have embraced the ONNX and ONNX Runtime ecosystem, enabling developers with ability to use ONNX models and ONNX Runtime APIs to offload AI with DirectML to the NPU. The powerful combination of Intel AI Boost and DirectML enables developers to deliver AI software at scale in an accelerated timeline, as our partners at Samsung did for their Samsung Gallery app. The following is a high-level Architecture block diagram enabling Samsung AI Workloads via ONNX Runtime & DirectML to offload AI workloads on the NPU:
Looking for the best platform to utilize such a great combination? Check out Samsung’s Galaxy Book 4, launched with Intel Core Ultra processors. Samsung took the opportunity to offload their Gallery application face and object classification features on the Intel NPU utilizing Microsoft DirectML framework, and their efforts paid off with a quick path to production.
What is next? Our journey on AI PC has just begun. The Samsung Gallery application is just one example of many more to come. With continuous integration of Intel AI Boost with DirectML and partner engagements, enhancements and optimizations continue to intensify as we scale to deliver the best hardware and software, in our goal to democratize AI, one PC at a time.
To learn more about “What is an AI PC and How Can It Help Me?”, refer to this article.
Additional Reading: Microsoft DirectML Blog Post