Windows 10's GPU scheduling switch: what it changes and how to test it

Windows 10’s May 2020 Update added an optional switch called Hardware-accelerated GPU scheduling (HAGS). It moves high-frequency scheduling work from a CPU thread to a dedicated scheduling processor on a supported graphics card. Windows still decides which applications get priority.

That is a change to the graphics plumbing, not a promise of extra frames. Microsoft introduced HAGS as an opt-in feature, off by default, and said most people should notice little or no difference. The first round of game tests available by 8 July 2020 points in the same direction: results vary by game and hardware, and there is no reliable reason to turn it on just because the switch exists.

What the new scheduler changes

Windows coordinates work from games and other applications that share a GPU. Before HAGS, a high-priority CPU thread handled GPU scheduling. With HAGS enabled, Windows can move most of that scheduling work to a dedicated processor on the GPU. Windows continues to set application priorities; the graphics card does not take over every decision.

The change targets scheduling overhead. It does not make a GPU more powerful, guarantee lower input delay, or directly raise a game’s frame rate. Microsoft described the first release as a foundation for future graphics features and said it expected the transition to be mostly transparent. That is a better starting point than assuming the word “accelerated” means a faster game.

HAGS arrived in the same Windows update as new DirectX features, but it is a Windows and driver scheduling change. A game does not need a separate HAGS option in its own graphics menu.

Which PCs can use it in July 2020

The toggle appears only when both the graphics hardware and its driver support HAGS. Windows 10’s May 2020 Update (version 2004) provides the operating-system support, but the graphics driver must also expose the feature. Microsoft’s DirectX explanation gives the setting’s location and describes the hardware and driver requirement.

The first vendor releases were narrow:

• NVIDIA: GeForce Game Ready Driver 451.48, released on 24 June, added HAGS support with the Windows 10 May 2020 Update. See NVIDIA’s 451.48 release page .
• AMD: Radeon Software Adrenalin 20.5.1 Beta with Graphics Hardware Scheduling Support added beta support for Radeon RX 5600 and RX 5700 series cards. AMD’s release notes warned that Oculus VR headsets might show no content or suffer severe performance drops with hardware scheduling enabled. This was a beta driver, not a general recommendation to update a stable PC.

Those names describe the early July 2020 rollout. Check the release notes for the driver version you intend to install; support depends on both the GPU and its driver.

Find the setting—and know when to stop

First confirm that Windows is on the May 2020 Update, version 2004. Open Settings → System → About and check the Windows specifications. Microsoft began rolling out the update on 27 May 2020; availability could vary by device.

If your GPU vendor has a driver that supports the feature, open Settings → System → Display → Graphics Settings. When the hardware and driver are compatible, Windows shows Hardware-accelerated GPU scheduling there. The setting is off by default. Enabling it requires a restart.

If the switch is missing, treat that as a compatibility result: this Windows, GPU and driver combination does not expose the option. Don’t edit the registry or install a driver from an unofficial site to force it to appear. If you rely on Oculus VR with an AMD card, avoid the 20.5.1 beta while its release notes list that issue.

Microsoft’s May 2020 Update announcement covers the update itself. The name “version 2004” refers to that release; the month in “May 2020 Update” is not a statement that every PC received it at once.

What the first game tests found

The early independent results do not support a blanket FPS claim.

Tom’s Hardware compared five games with HAGS on and off using several CPU and GPU combinations. Its overall average changes were close to zero in the tested configurations, though individual games moved in both directions. The tests used an RTX 2080 Ti with a Core i9-9900K and a Ryzen 9 3900X, plus a GTX 1050 with the i9. The article’s test results are useful evidence about those systems—not a prediction for every PC.

GamersNexus tested with an intentionally CPU-limited Core i3-10100, then checked results on a Core i9-10900K. It reported no significant frame-rate or frame-time consistency gains in its game suite and found performance losses in at least one case. Its 4 July report also explains the limits of its one-off test setup.

These reviews measured game performance and frame timing, not a universal reduction in click-to-screen input latency. Their results cannot prove that HAGS never helps; they do show why a single before-and-after FPS number is weak evidence.

Test it on your own PC

Jitsumi has not benchmarked HAGS. If the switch is available and you want to try it, compare the same workload under the same conditions:

1. Pick a repeatable built-in game benchmark or a short route you can replay. Record the Windows version, graphics card, driver, game version, resolution and graphics settings.
2. Leave HAGS off. Run the benchmark once to settle the scene, then record three runs. Note average FPS and frame-time or low-percentile results if your tool provides them.
3. Turn HAGS on, restart, and repeat the same runs without changing other settings.
4. Compare the runs, not just the best score. If the difference changes direction between runs or is smaller than the normal variation, call it inconclusive. Check for stutter, visual glitches, crashes and any VR issue as well as the FPS number.
5. Keep HAGS on only if the improvement is repeatable in the game you care about and the system remains stable. If performance worsens or a new problem appears, turn it off and restart.

The practical answer in early July 2020 is simple: leave the default alone unless you can test a specific problem or benefit. HAGS is an optional change to how Windows schedules GPU work, and the first published gaming tests found no consistent reason to expect a noticeable gain.