Inside Intel’s GPU Roadmap - Tom Petersen Talks Xe3, Panther Lake, and Beyond

In this episode of The Full Nerd, the team is joined by special guest Tom Petersen, Fellow at Intel, for an in-depth discussion on Intel’s GPU roadmap, Xe3 architecture, Panther Lake gaming, benchmarking, and more.

Episode Highlights

• Intel–Nvidia Partnership (14:48–16:07)
  Petersen explains Intel’s new partnership with Nvidia, focusing on integrated products for data centers and the potential for Nvidia GPU tiles in consumer SoCs. This strategy leverages chiplets and advanced packaging to improve cost and efficiency.
• Future of Intel Graphics Division (17:46–27:26)
  Intel will continue developing its own graphics IP (Xe3, Xe3P) for both consumer and data center markets. Petersen highlights the importance of discrete GPUs in laptops and the need for a unified driver across all Intel GPU variants.
• GPU Roadmap & Naming Conventions (20:52–26:55)
  From Alchemist (Xe) to Xe2 (Lunar Lake) and the upcoming Xe3 (Panther Lake), Petersen acknowledges the confusion around naming but outlines Intel’s efforts to simplify while maintaining software compatibility.
• Xe3 Architectural Improvements (33:10–35:47)
  Xe3 addresses CPU overhead and low hardware utilization issues that software alone couldn’t resolve in Xe2. Petersen emphasizes micro-architectural corrections that deliver greater efficiency.
• Benchmarking Methodologies (35:55–36:20)
  Petersen discusses the challenges of benchmarking and how Intel uses microbenchmarks to optimize specific architectural components.

Whether you’re a gamer, developer, or simply curious about Intel’s GPU strategy, this episode offers valuable insights into the future of graphics technology.