AMD explained the failure of the Ryzen 9000 in independent tests and promised to fix everything

AMD has explained in detail why the results of the official tests of Ryzen 9000 processors differ so greatly from the results obtained by the review authors. The company promised that this difference will be leveled out with the release of one of the upcoming Windows 11 updates – not only the Zen 5 family of chips will speed up, but also Zen 3 and Zen 4.

Image source: amd.com

During the announcement of the Ryzen 9000 processors, AMD said that in its own tests the average performance increase in games at 1080p resolution compared to the Ryzen 7000 chips was 9%; the average increase over a full cycle of more than 30 games was 6%. However, in most cases, reviewers were unable to achieve such results, AMD admitted, and there is an explanation for this.

AMD compared its processors with Intel products, using DDR5-6000 memory for tests and standard Intel settings with a basic power consumption limit, at which the processors do not reach their potential, which can affect gaming performance. During testing, Windows VBS (Virtualization-based Security) was also launched – a service that Microsoft recommends turning on to improve security, but it can also affect gaming performance.

AMD also noted that in its gaming tests it uses e-sports games, modern AAA titles and popular old games, the performance of which can depend not only on central processors, but also on graphics processors, and the average performance is highly dependent on the test set. Another reason for discrepancies between official and independent tests was that in games without built-in performance tests, AMD and reviewers could select different scenes and get different FPS figures.

However, perhaps most importantly, the independent tests were actually conducted incorrectly. Finally, Zen 5 architecture processors have an improved branch prediction module. AMD testing was carried out in administrator mode, which allows access to an optimized version of the code for this module, but it is not available in the version of Windows 11 23H2, which the review authors mainly used when testing the Ryzen 9000.

AMD provided a table comparing the performance of some games when running on Windows 11 23H2 and 24H2 – just changing the OS provides a performance increase of up to 13%. In this regard, the company insists that the Ryzen 9000 offers a 10% increase in CPU speed, a 25% increase in AI workloads, and a 5-8% increase in gaming compared to the Ryzen 7000. Note that these figures differ slightly from AMD’s original claims – the company previously announced a 9% increase in games compared to the Ryzen 7000 – but they are still higher than what independent reviewers received.

AMD offers to verify that it is right by installing a preliminary version of Windows 11 24H2, the public version of which will appear in the near future. The company is working with Microsoft to release an update for all Windows 11 users – optimization of the transition prediction module will affect processors on Zen 3, Zen 4 and Zen 5 architectures.