In context: When AMD launched its new Ryzen 9000 CPUs earlier this month, there was a lot of hype around the promised performance gains across the board, including gaming. AMD’s own pre-release benchmarks showed a tidy 9% generational uplift over Ryzen 7000 in 1080p gaming, and around 6% higher frame rates compared to Intel’s best. But once the reviews started rolling in, not everyone was seeing those kinds of improvements.
This week AMD has provided an official explanation for the discrepancies between their numbers and what reviewers were measuring. The company has outlined a few culprits that they claim likely caused the difference.
First, AMD says that despite its internal test suite including a diverse mix of games, the exact game lineup can significantly sway overall performance results. Second, for the Intel test systems, AMD says it enabled comparable DDR5-6000 memory speeds and default-settings baseline power profile, which can affect performance. Third, AMD says it tests with Windows Virtualization-based Security (VBS) active, which Microsoft recommends for better security, but this can affect frame rates.
Overall, it appears AMD’s reliance on built-in benchmarks for the majority of its automated tests may not provide an accurate representation of real-world usage scenarios.
However, AMD says the biggest difference could be related to branch prediction code optimizations between the Windows 11 versions used by AMD’s labs and many reviewers. The Zen 5 architecture has an expanded branch prediction capacity over previous Zen chips. AMD’s automated test process was run in “Admin mode,” which took advantage of these optimizations – optimizations that may not have been enabled in older Windows 11 builds.
It’s worth mentioning that our own Steve Walton previously highlighted an odd discrepancy in gaming performance when using admin mode. While the cause wasn’t known at the time, it seems like AMD and Microsoft have finally figured it out.
To that end, AMD has shared preliminary data showing that the upcoming Windows 11 24H2 update will fix this bug. The update can boost gaming performance anywhere from 3% to 13%, thanks to the AMD-optimized branch prediction code included in the OS release. The update should start rolling out widely soon but for now it’s available to Windows Insiders.
AMD says it’s “collaborating with Microsoft” to hopefully get these gaming optimizations out sooner via an optional update on Windows 11. As pointed out before by Steven, while Zen 5 CPUs will likely see the biggest boost from 24H2, Zen 4 and even Zen 3 chips should also see some of the benefit.
