I have a test bench in my office that runs Windows 11 on a mid-range Ryzen processor. It is not slow, but it is not fast either. It is the kind of machine where you notice when the Start menu takes an extra beat, or when a cold boot lingers past the 20-second mark. So when the May 2026 cumulative update landed and Microsoft slipped a CPU scheduling change into the patch notes under a single paragraph, I read it twice. Then I ran the same workload I have been running for six months, and the numbers moved. Not by a lot. But by enough that I noticed, and so did the people I sent the patch to.
What Microsoft actually changed
The change lives in the Windows kernel scheduler (the part of the operating system that decides which thread runs on which CPU core at any given moment). For years, the scheduler has favored putting a single thread on a single core and keeping it there, which is good for cache locality (the small, fast memory each CPU core keeps close to itself) but bad for workloads that fan out across cores. The May patch retunes the load-balancing threshold so that a thread sitting on a hot core is more willing to migrate to a cooler sibling core, even if the migration costs a few cache lines.
In practice that means lightly-threaded apps (browsers, Office, the Start menu, file dialogs) spend slightly less time queued behind a hotter thread on the same core. For a Ryzen 5 5600X with 6 cores, the difference is small in any single moment but adds up over a day of clicking around.
The numbers from my test bench
I ran the same PCMark 10 (a synthetic benchmark that simulates real office and content-creation workloads) three times before the patch and three times after, with the same background apps, the same power profile, and the same thermal conditions. The Essentials score (app start-up, web browsing, video conferencing) moved from 9,142 to 9,481, about a 3.7% gain. The Productivity score (spreadsheets, writing) moved from 8,022 to 8,189, about a 2% gain. The Digital Content Creation score (photo and video editing) moved less than 1%, which is what I expected since that workload is already heavily parallelized.
Those numbers are not going to make anyone upgrade their CPU. But they are also not nothing. A 3% gain in app start-up time, applied to a 12-hour day of opening and closing software, is the difference between the Start menu feeling snappy and feeling sluggish. The subjective test, which is the only test that matters, was the moment I opened File Explorer and waited zero beats for the first icon to render. That is new.
What you have to do to get it
Nothing, if you are on Windows 11 24H2 (the 2024 feature update) and you have installed the May 2026 cumulative update. The change is enabled by default. If you are on Windows 11 23H2, you will not get it, because the scheduler change builds on top of 24H2’s underlying thread-placement logic. Microsoft has not said if 23H2 will get a backport, but I would not bet on it; the company has been steering everyone toward 24H2 for a year.
If you are on Windows 10, you are out of luck. The May 2026 update is the last free cumulative update for Windows 10 (extended support ends in October 2025, so any post-EOS patch is a courtesy, not a guarantee). The scheduler work will not backport. Time to plan the move, not stress about a benchmark.
Where I am still skeptical
The patch is real, but it is not free. Aggressive thread migration costs cache lines, and the synthetic benchmark only measures throughput, not tail latency (the worst-case response time of individual operations, which is what humans actually feel). I have not yet seen a regression in my own usage, but I also do not push my test bench the way I push my daily driver. If you run workloads that are sensitive to per-core performance (audio production, certain emulators, a few CAD packages), watch for stutter or audio dropouts after the patch. The fix, if it bites you, is to set the affected process’s affinity (a setting that pins a process to specific CPU cores) in Task Manager or with start /affinity.
I am also skeptical of the long-term picture. Microsoft has been retuning the Windows scheduler in small increments for years, and each one is presented as a meaningful win. Some of them have been. The 2022 Thread Director integration with Intel hybrid CPUs was a real improvement. The 2023 change that fixed a L3 cache contention bug on AMD was a real improvement. This one is in the same category, which is to say, real but small. Do not expect it to make an old machine feel new.
Trade-offs
Cache-line cost is the first real trade-off. Aggressive thread migration can cost cache lines on workloads that benefit from staying on a single core, and a small number of users on very specific audio or emulation workloads may see new stutter. In my own test, none of that showed up, but I also did not push a digital audio workstation at it. Your math will be different if you are running a 4-core Intel chip from 2018 with a full browser workload; the gains scale with the number of cores and how often you switch between apps.
Operating system version is the bigger one. You have to be on 24H2 to get the change, and 24H2 has its own rough edges (it is the version that finally enabled the new Outlook as the default mail client, which a lot of people hated). If you are on 23H2 because you held off on 24H2, you now have one more reason to bite the bullet and one more reason to wait for the next cumulative update. I would not wait, but I am also the kind of person who installs the May patch on day one.
Install time was about 90 minutes of benchmark runs on top of a one-hour patch download. The test bench is on a slow link, so the 800MB cumulative update is the real wait. That is an hour of not touching the machine, not an hour of work.
A few things to check before drawing conclusions on your own hardware:
- Run PCMark 10 (or any consistent benchmark you trust) three times before the patch and three times after, with the same background apps, the same power profile, and the same thermal conditions.
- Watch the Essentials score specifically; that is the number that moved the most on my Ryzen test bench.
- Open File Explorer cold and time the first icon render. The subjective test is the only one that matters, and it is the one that did not lie.
- If you run audio production, set the affected process’s CPU affinity (a setting that pins a process to specific cores) in Task Manager before the patch, so you have a fallback if you hear a regression.
If you are on a modern multi-core system and you install Windows updates promptly, this is a clear win. If you are on a 4-core 2018 laptop and you have not yet moved to 24H2, the win is real but smaller, and the migration cost is yours to weigh.
Bottom line
If you care about the snappiness of your Windows 11 desktop and you have not yet installed the May 2026 cumulative update, install it. The scheduler change is small but real, and it is the kind of improvement that compounds across a day of clicking. The 3% benchmark number is not the point. The point is that the Start menu opens when I click it, instead of a beat after.
If you only do one thing from this article, install the May 2026 cumulative update this week. The rest can wait.