The launch of AMD’s Ryzen 9000 desktop processors earlier this month didn’t quite deliver the gaming performance improvements that many had hoped for, prompting questions about what went wrong. As it turns out, the discrepancy between AMD’s marketing claims and independent reviewers’ findings stems from a combination of factors that have now been brought to light.
In a community post released last night, AMD pointed to several key issues: differing Windows modes used for testing, VBS security settings, Intel system configurations, and the specific games benchmarked. These variables created a performance gap that puzzled both reviewers and AMD.
David McAfee, AMD’s client channel segment leader, addressed the situation in a special edition of The Full Nerd. One major factor was that AMD’s tests were conducted in “Super Admin” mode, which allowed the new “Zen 5” architecture to utilize branch prediction optimizations not available in the Windows versions used by reviewers. This difference led to inflated performance numbers in AMD’s internal testing.
McAfee explained that AMD’s automated testing framework, designed for large-scale product testing, runs in Super Admin mode. Historically, there was little difference between Super Admin mode and regular user mode, but over time, that gap widened, and AMD failed to account for it in its marketing numbers. This oversight has since been corrected, aligning AMD’s data collection with the way reviewers and gamers actually use their systems.
The blog post also noted how game selection and scene choice within games can affect benchmark outcomes, highlighting the complexity of accurately assessing gaming performance. McAfee further elaborated on how the balance of CPU and GPU demands can vary significantly between game sections, leading to different results.
Crucially, McAfee made it clear that AMD isn’t blaming reviewers for the performance discrepancies. The company acknowledges that differences in testing approaches and configurations were responsible for the divergent outcomes. For those interested in a deeper understanding of the issue, McAfee’s interview offers valuable insights into the factors at play and the lessons learned from this experience.