Guest
In this episode of Talking Tech, we’re joined by Intel Fellow Tom “TAP” Petersen, who breaks down advancements in our Xe2 architecture at the core of Lunar Lake's built-in Intel Arc GPU. This scalable architecture will also serve as the foundation of our forthcoming discrete GPUs codenamed Battlemage. Second generation Xe cores are tuned for performance and efficiency, with deeper caches, larger ray tracing units, and enhanced vector engines that implement Xe Matrix eXtensions (XMX) to accelerate AI workloads like Xe Super Sampling (XeSS). We also discuss the challenges of running large AI models locally on AI PCs, along with various means of optimizing them. Learn more about Lunar Lake in this in-depth overview: How Intel Arc graphics are optimized for modern media codecs: Chapters: 0:00 - Intro 0:49 - Considerations for improving graphics performance 1:35 - Breaking down the Xe2’s scalable architecture 5:07 - A look at the current state of AI 7:08 - Running complex AI models on a local PC 9:14 - Using embedding for prompt-based generative AI models 13:41 - What does the future of local AI workloads look like? 14:44 - Teaching AI models with RAG 15:58 - The evolution of transformer models 19:12 – How quantization helps balance size and accuracy 22:05 - What’s new in Xe2 30:13 – Support for new media codecs 31:59 - Conclusion About Intel Technology: Intel has always been at the forefront of developing exciting new technology for business and consumers including emerging technologies, data center servers, business transformation, memory and storage, security, and graphics. The Intel Technology YouTube channel is a place to learn tips and tricks, get the latest news, and watch product demos from both Intel and our many partners across multiple fields. Connect with Intel Technology: Visit Intel Technologies WEBSITE: Follow Intel Technology on TWITTER:
