Intel Alder Lake vs. AMD Zen 3: Windows 11 vs. Windows 10 August 2022 results

Video version here.

October 5th, 2021 marked the initial release of Microsoft’s Windows 11 OS. Soon after, many users started to notice lower performance on certain CPUs compared to Windows 10, and fixes were promised. Almost a year later we are here with results from our benchmark suite utilizing the latest Windows 11 update to see how Alder Lake and Zen 3 (Vermeer) perform. The update is dubbed Windows 11 Build 22000.918 or KB5016691 Preview released on August 25th, 2022. It is the latest build and an optional update. Come September’s 2022 patch Tuesday, the latest non-optional update will include many of the changes in the 22000.918 build. We will test again come the release of the September patch Tuesday update, but we wanted to give our readers a sneak peak of what’s to come. The holiday season is only a few months away, and AMD already announced Ryzen 7000 series will be on shelves on September 27th.

You can find our AMD 5000-series review here and our Intel 12th Generation review here.

Video version can be found here.

Test Setup:

We did disable hyper-V and core integrity so our benchmarks don’t run in a sandbox. Nevertheless, we also ran into some oddities in some legacy benchmarks that have been replaced.

Windows 11 vs. Windows 10 Alder Lake vs. Zen 4 Processor

Performance Benchmarks:

Windows 11 results are promising in CINEBENCH R23. One of the bigger issues people were seeing earlier with Windows 11 was lower performance, and that doesn’t seem to be the case. Everything is pretty much the same with a few slight bumps up or down, all within what we would consider the margin of error. We want to mention that R23 is the latest version of this benchmark, and that R15 has issues with the Alder Lake CPUs; avoid using the CINEBENCH R15 benchmark for performance evaluations.

Once again we see almost no difference between Windows 10 and Windows 11.

Here we do see some odd results. wPrime utilizes all threads to calculate Prime numbers using Newton’s square root method. It is old and is not like Prime95, so AVX isn’t used in this case, which would force movement of threads to P-Cores in the Alder Lake CPUs. While this behavior isn’t what we expected at all, it proves there is a deeper change in Windows 11 when it comes to the Alder Lake P-core/E-core design and that Thread Director is monitoring and working in the background. The results show us how an older legacy benchmark is still useful in producing outliers that prove something different is going on, and in this case, that Thread Director is actually working. It might be working a bit too well, as everything goes well for the first 10% of the run and then plummets in speed after that, as if the threads are being de-prioritized. It is possible that background threads were deemed more intensive than any one of the 24 threads (12900K) wPrime executed. The benchmark is designed to utilize all threads, but its difficultly on each thread is not as intense as it could be. We will replace this benchmark in our future testing, but we included it to display how with the new architecture of the hybrid core CPU, older benchmarks don’t necessarily show us what we expect.

Everything is going normal in legacy Handbrake transcoding. However, something is up with the 12600K, and we don’t see the same behavior with the 12900K. We saw 4K to 1080P transcoding increase by a reasonable amount in both cases in Windows 11, but the 12600K does much better in 720P format rendering, and the 5950X and the 5800X do better as well.

Funny enough, for the most part, the new Handbrake benchmark is mostly showing more improvements in different tests and without outliers.

Results either stayed the same or improved in AIDA64 CPU tests.

Not much change in memory bandwidth.

Not much change in latency.

Another legacy benchmark here, but in this case we see that the Alder Lake CPUs show noticeable improvements with Windows 11, while the other CPUs don’t change much. The large memory bars in the charts are due to the fact that ScienceMark isn’t able to detect the memory structure in the CPU, so it falls back on a default.

SuperPI is strongly single threaded and memory latency dependent. All the Intel systems take a hit to some extent going from Windows 10 to Windows 11, but the AMD systems all improve by a small amount.

Gaming Benchmarks:

The only thing to take away from these results are that most of the results stayed the same going from Windows 10 to 11, with the exception of the 10th and 11th generation Intel 6-core parts that received a small boost.

No changes here.

Minor boosts for most processors in CloudGate.

Not many changes in FireStrike either.

Most CPUs in our charts went up slightly, except for the older CPUs that received a slight ding to their scores.

Not much of a change in Rise of the Tomb Raider either.

These results are a bit more telling. Before we go forward, we should mention that Visual C++ Distributable needed to be re-installed to get the game to work in this version of Windows 11. It seems the 6-core parts from Intel did fine (they are the bottleneck in this game anyways), but the rest of the CPUs took a hit.

Thread Director pulled through in GTA:V for Intel, with increased for both Alder Lake CPU (i5-12600K and i9-12900K).

Thoughts:

There isn’t too much to say here. Our original reviews for both the 5000-series AMD parts and the 12th generation Intel parts still stand and our ratings haven’t changed. Windows 11 is different than Windows 10, and while Microsoft has padded the OS and changed many things and needs to optimize the OS more, we do see improvements in some aspects and problems in others. Thread Director is working better in Windows 11 than Windows 10, and we see Windows 11 improving in performance from launch. There will obviously be some hurdles to get over with the relatively new OS, and it will take a while to smooth them out. We look forward to the September 2022 update, which is only a week away.