AMD Says Windows 11 Preview Branch Prediction Optimizations Backported To 23H2

windows update hero
If you recall, benchmark numbers AMD’s new Ryzen 9000 “Zen 5”-based desktop processors were well behind the expected performance gains compared to the previous generation. Those expectations were set by AMD, and in response to community outcry over the missed marks, AMD gave a number of excuses. One of those was that it tested using the latest Insider Preview build of Windows, and that this build contained significant optimizations related to branch prediction over extant builds.

Putting aside for now the absurdity of testing with a preview build of Windows, we’ve already tested that claim and found that the new build did indeed slightly improve performance on Zen 5. There are lots of reasons someone might not want to run a preview build on their personal PC, though. If you’re in that group yet you still want the best performance for your Ryzen machine, you’re in luck, because AMD shot us a line to let us know that those optimizations have now been backported to Windows 11 23H2.

windows update optional updates
This is the update you’re looking for.

Indeed, if you’re on the current stable version of Windows 11, go to the Windows Update settings page, click “Advanced”, and then “Optional updates”. Under there, you may well see an update numbered “KB5041587” titled “Cumulative Update Preview for Windows 11 Version 23H2”. Install that, and according to AMD, you should be able to enjoy improved CPU performance on your Ryzen processors.

We say “according to AMD” because we spent most of the afternoon testing it on three different machines and actually came up with no real measurable change across three different benchmarks. The three tests in question were Geekbench 6, Cinebench 2024, and a Hogwarts Legacy benchmark in the demanding Hogsmeade town area, and the systems we tested were based on a Ryzen 7 5800X3D, a Ryzen 7 8700G, and an Alienware m18 R1 laptop with a Core i9-13980HX CPU.

benchmarks

Now to be fair, three benchmarks isn’t a lot, but to reiterate, not a single test presented a notable change in performance. For a game, we chose Hogwarts Legacy as it is known to be heavily CPU-limited, and we made sure to disable ray-tracing to ensure that we weren’t hitting a GPU limitation. Even still, using the high-precision CapFrameX benchmark tool and despite average frame rates as high as the 90s, we weren’t able to pick any particular difference between the pre-patch and post-patch state. 
Unfortunately, whatever change made by the new Windows update causes Denuvo to count the machine as a new install, and we were locked out of testing the update in Hogwarts Legacy on the Ryzen 7 5800X3D, but given our results on the other two machines and other two benchmarks, we don’t expect a big difference. Geekbench 6 and Cinebench 2024 showed essentially margin-of-error changes, and given the extremely small differences, we’re inclined to believe that they’re measuring artifacts more than anything else.

If you’re wondering why we tested Zen 3 and Zen 4 processors as well as an Intel chip, it’s because we’ve already tested the impact of the Windows 11 26100 build on Zen 5, and AMD said that this branch prediction update could improve performance for Zen 4 and Zen 3 as well. We were also curious if Intel would benefit, too. However, in our findings, it really didn’t make any difference at all for these other CPUs. Let us know in the comments if you test update KB5041587 for yourself.