The AMD Ryzen 9000 series desktop processors opened to lukewarm reception from reviewers as well as AMD fans and enthusiasts. That is because the company"s Zen 5 architecture, powering these new 9000 series SKUs, has failed to deliver on its promise of the supposed 16% IPC boost that it had claimed. This is sort of a first for AMD"s Zen as typically the company has delivered excellent uplift every time so far since its launch in 2017.
AMD acknowledged that there was a performance issue and that it was dealing with it. The company said that using Windows" hidden Administrator account was a way to work around the problem and certainly testing confirmed that this was indeed the case as there was some performance to be had, not just on Zen 5-based Ryzen 9000 series CPUs, but also on Zen 4-based Ryzen 7000.
AMD also later published a detailed blog post on its Community website explaining what exactly was the problem. The tech giant noted that the Windows OS was unable to take proper advantage of the newer and improved branch prediction on Zen 5.
For those not familiar, branch prediction in CPUs helps minimize branch misprediction penalties and is especially useful in highly pipelined processor designs.
AMD said that the upcoming Windows 11 version 24H2 (note that it is already available for Copilot+ PCs) will address this problem head on and users should find themselves in the middle of improved performance once they upgraded to Windows 11 24H2 from an earlier version.
The company also made some performance claims mainly in the case of gaming, but added UL Procyon Office too.
Today, YouTube channel Hardware Unboxed (HwU) ran multiple tests on games to see how much of an uplift Windows 11 24H2 could bring to AMD"s chips. It measured the Ryzen 7 9700X (Zen 5) and the 7700X (Zen 4) and the two CPUs exhibited an impressive average performance boost of 11% and 10% respectively, vs Windows 11 23H2.
| |
| |
It is noteworthy here that the games were run at 1080p and hence the absolute uplift in performance may be even bigger as I have previously explained in my editorial how low-resolution tests can help accurately measure CPU performance in games.
Meanwhile, Intel did not see any improvement suggesting that an upgrade to 24H2 may not be worth it, at least from the performance standpoint.
Source and images: HwU (YouTube)