Windows 10: Application Benchmarks — The Ugly Side
And then there's the gotcha, at least if you plan on using Windows 10.
We planned on presenting an overall geometric mean of our Windows 10 application testing here, just like we did with Windows 11, but we use several of these applications to generate those results. Unfortunately, as you can see in the charts, these applications absolutely did not respond well to Alder Lake and delivered what can only be described as terrible performance. However, as you'll see below, these same applications ran absolutely perfectly in Windows 11, even beating AMD's comparable chips. That shows that Intel's Thread Director is a powerful tool.
There is an answer for this type of condition, at least in some cases. Intel's reviewer guide cited a similar condition, present in HandBrake x264 (but not x265), where the software developers assigned the program a lower priority that forces it to run on the E-cores only. Intel says that the software developers could update the code to fix the issue, but it appears that this condition (or a similar one) isn't confined to just Handbrake. We saw similar trends in y-cruncher, Corona, POV-Ray, and even Intel's own Open Image Denoise benchmark.
Intel's recommendation is to use the in-built Windows powercfg (opens in new tab) command-line utility to prevent the process from lowering its priority, which resolves the issue. This is a simple process for experienced users, but most average folks would struggle with these types of alterations, and that could be a real downside to using Alder Lake with Windows 10 during the early days. If you tend to use older programs that won't be updated, this type of problem may never be fixed via a software update, but there are third-party software tools that could help. Additionally, we use an expansive selection of benchmarks, but our entire test suite is a speck compared to the universe of different software in the real world. That means these types of errata are inevitably going to pop up with other types of software.
We could have applied the fix and retested the Alder Lake chips, but it's important that you understand that you could encounter this type of reduced performance. Additionally, we also wonder if changing the thread priority on the AMD and last-gen Intel systems would impact performance there, too, meaning the adjustment could give Alder an unfair advantage. Unfortunately, retesting all of these systems to assure a level playing field wasn't possible within the tight NDA timeline. As such, we'll revisit these tests in the future.
For now, let's move on to the Windows 11 benchmarks below, and then see how the remainder of the Windows 10 benchmarks look.
Windows 10: Rendering Benchmarks on Core i9-12900K and i5-12600K
Aside from the tests that refused to cooperate, many of these tests simply repeat the same themes that we've seen in Windows 11. We're including these tests for completeness, but we'll skip over commentary until we hit the power consumption section.