This article tries to explain why some reviewers of the new AMD Ryzen 9000 series Zen 5 CPUs are seeing less improvement over the Zen 4 series and Intel CPUs.
AMD launched all four of its new Zen 5 CPUs in August, and overall it seems like it was kind of a disaster for the company. With Intel's Raptor Lake on the ropes over the ongoing instability issues, AMD was expected to enter the ring and deliver an easy KO to Intel and Zen 4. Instead, reviewers found AMD's Zen 5 chips delivered befuddling results, making the launch a confusing affair. AMD is now attempting to clarify the situation in a new blog post, saying there are myriad reasons why its numbers differed from reviewers' results.
The gist of AMD's blog post is that the PC ecosystem is a complicated place, with a huge swath of different types of games new and old, and a huge variety of hardware involved, so results may vary. AMD states its own results showed decent generation-to-generation improvements for Zen 5 compared with both Zen 4 and Intel's CPUs, depending on the application. However, that's not what reviewers saw, and AMD says there are a few reasons for this.
First, in gaming, AMD says its numbers showed a 9% uplift at 1080p compared with Zen 4 and 6% compared with Intel's CPUs. AMD stipulates this is not what reviewers saw, though, as some saw no actual gains for Zen 5 in gaming compared with Zen 4 and Intel's chips.
The first reason for the discrepancy is that performance varies depending on which games are being tested. The second reason is its internal Intel systems were tested using a baseline power profile, as opposed to a more high-powered profile, which some reviewers used. Its Intel systems also used DDR5-6000 memory, which is slower than some memory kits available for Intel systems. Third, AMD tests with Windows Virtualization-based Security (VBS) enabled—but not all reviewers use this setting, which can affect gaming performance.
Finally—and possibly most importantly—AMD says its automated testing for Zen 5 used an "Admin mode" for Windows 11 that reviewers didn't have access to. This makes no sense at all, but that's what AMD wrote. It says this mode is optimized to take advantage of Zen 5's wider prediction path, but end users will shortly be able to access this mode via a Windows Update.
Given the above caveats, AMD has revised its performance numbers for its Zen 5 CPUs compared with Zen 4 and Intel's Raptor Lake refresh CPUs. For gaming, when compared with an Intel system using faster memory, extreme power delivery, and optimized settings, AMD has revised its numbers from a 6% uplift to zero, or "parity" in AMD's words. Compared with Zen 4, AMD now claims a 5-8% uplift in 1080p gaming versus 9% before. Productivity apps see a ~10% uplift, and AI applications get a ~25% boost.
Finally, AMD says the Admin mode will be available soon to customers running Windows 11 24H2. Currently, it's only available in the preview channel. Admin mode appears to offer some tangible benefits in gaming, ranging from a 13% boost in some games to no change. AMD says the mode can benefit Zen 4 and Zen 3 users, too, but Zen 5 customers will see the biggest benefit. It says it's working with Microsoft to get this to all Windows 11 users as soon as possible.
Overall, we're sure the hardcore PC community that pays attention to this stuff appreciates AMD's candor. But there is no denying it's a bit of a disappointment to see AMD revise its numbers downward post-launch, as it's something we can't ever remember a company having to do before. Even with AMD's clarifications, this launch clearly hasn't gone the way AMD intended, and it will no doubt leave a bitter taste in the mouths of even AMD's most ardent supporters. It's especially disappointing given Intel's current troubles. AMD should have been able to clean its rival's clock, but it ended up stepping on a rake instead.
