Keep Hardware Accelerated GPU Scheduling On or Off?

With the Windows 10 May 2020 (2004) update, Microsoft introduced a new GPU scheduler to counteract these shortcomings: HAGS! It sets the foundation for many new technologies to come. For example, NVIDIA’s Frame Generation technology requires HAGS to function.

Hardware-accelerated GPU Scheduling offloads most of the GPU’s tasks and memory management to a dedicated processor on the GPU. This eliminates the need for the intermediary (CPU) and mitigates most of the input latency issues introduced by traditional methods.

All said and done, does HAGS actually increase performance? Theoretically and technically, it should. Microsoft claims that HAGS will not cause performance regressions, whereas both NVIDIA and AMD claim that this feature could improve performance and latency.

We’ve observed and will cover many instances where HAGS can lead to a significant performance boost at no cost. However, in some other cases, HAGS can cause instability and crashes. We’ll discuss these in a later section.

How to Turn On Hardware Accelerated GPU Scheduling

First things first, does your GPU support HAGS? Let’s see the supported GPUs and architectures from all three major GPU manufacturers.

• NVIDIA: GTX 10 or Higher with the GeForce 450 Driver
• AMD: RX 7700 and Above (On Windows 11 Only)
• Intel: Not Supported Yet

Here’s how you can turn on Hardware Accelerated GPU Scheduling if your system and driver support it:

1. Press the shortcut Win + I on your keyboard as shown in the image.
   
   
   
2. Once the Settings tab opens, type “Graphics” in the Search Bar and select the option that says “Graphics settings“.
   
   
   
3. When the Graphics page opens, click on the highlighted option saying “Change default graphics settings“.
   
   
   
4. Turn on the “Hardware-accelerated GPU scheduling” option as shown.
   
   
   
5. It’s recommended that you restart your PC to ensure everything works properly.

Should You Turn HAGS On?

While only NVIDIA realistically supports HAGS (and AMD to some extent), let’s discuss whether you should use this feature. We’ll go through various benchmarks, both synthetic and gaming, to see if HAGS makes a difference.

Generally, a small performance loss/gain should be negligible and will be within the margin of error. What we want to look for are outliers and check if the performance drops significantly in specific scenarios.

Fortunately for us, Puget Systems conducted a large number of synthetic tests a while back. We’ll use their numbers as a reference to see if HAGS is really worth all the hype.

1. Synthetic Testing

Again, credit goes to Puget Systems for investing valuable time conducting these tests. From the results, we find that HAGS has very little impact on synthetic benchmarks. So, if you’re looking to impress your friends with a higher score, HAGS won’t do the trick.

Starting with Photoshop, the largest difference was with the AMD Threadripper PRO 5995WX and the RTX 4080 system where a configuration without HAGS leads to 8% better performance.

In After Effects, there is a consistent lead when HAGS is turned on, with the largest performance gap at 10%. Moving over to Unreal Engine 5.2, most results are within the margin of error, so no conclusive decision can be made in this case.

Blender in CPU mode encounters much of the same fate, with little to no contribution in performance after toggling HAGS on. Moving over to Blender with Optix, here HAGS once again shows an almost invariable lead, extending up to 7% with systems having extra GPU headroom.

HAGS has both upsides and downsides, at least in synthetic tests: at one extreme, we gain 10% extra performance but lose 5% at the other. Let’s move over to gaming next, where one might expect some actual differences.

2. Gaming

Let’s rip off the band-aid: HAGS is a mixed bag, and we’ve seen cases where different architectures caused HAGS to result in either a +15% or a -15% performance delta in the same game. It is that variable. While most games won’t show a massive uplift in the average FPS, we want to keep an eye out for the 0.1% and 1% lows, as these numbers determine how stuttery or smooth your gaming experience will be.

↪ NVIDIA

The first set of benchmarks features NVIDIA’s RTX 4090 paired with a Ryzen 7 5800X (Credit: DannyzReviews). Users can review all the benchmarks, but we found most to be within the margin of error. However, the ones displayed in the images below showcase a substantial improvement in the experience and fluidity of the game, with HAGS turned on.

The RTX 4090 at stock in Far Cry 6 dipped as low as 62 FPS, almost entering the “unplayable” territory. Upon turning HAGS on, the 1% lows improved to 82 FPS. In Gears 5, the average FPS improved from 107 FPS to 118 FPS and the 1% lows also jumped from 62 FPS to 90 FPS. Combined with ReBar, HAGS offers an almost free-of-cost improvement in your performance.

↪ AMD

On the AMD side of things, Ancient Gameplays conducted various benchmarks a few months ago with the RX 7700 XT + Ryzen 5 7600X and the RX 7900 XTX + Ryzen 7 7700X. To our surprise, there is an endless tug of war between HAGS on and HAGS off. This can be attributed to the hardware differences between RDNA3 and Ada Lovelace. Anyhow, let’s go over the final results.

The slower Ryzen 5 7600X + RX 7700 XT setup managed to deliver worse performance with HAGS on. At 1080p and 4K, not only was the average FPS slightly lower, but the 1% lows didn’t see any increase, either. Would you consider a 1-2% decrement within the margin of error? The difference isn’t much, but the headline is that this setup didn’t see any notable improvements.

Moving over to the Ryzen 7 7700X + RX 7900 XTX, the opposite is seen here as this configuration delivers better 1% lows at 4K and 1440p, whereas 1080p experiences a marginal regression.

↪ Edge Case Testing

Interestingly, in CPU-bound scenarios, HAGS almost becomes a necessity. YouTuber 5700X BAE often conducts tests in CS:GO 2 and Valorant using different new technologies. With the June 2024 patch in CS:GO, the RTX 3060 12GB and Ryzen 7 5700X deliver consistently better averages and lows with HAGS on. Additionally, HAGS also flattens out the frametime graph as shown in the video.

Valorant experiences a similar impact from HAGS, where the 1% lows improve by almost 100 FPS. Likewise, the framerate also jumps from 500 FPS to 600 FPS in some scenarios. So yes, in some edge-case scenarios, turning HAGS on is a no-brainer. It is best to try your games with and without this setting, do a bit of testing, and see if you find any noticeable differences.

From what we’ve seen thus far, turning HAGS on does you more good than harm. Sure, you may lose 1-2% performance, but that’s within the margin of error. In the best-case scenario, your game will run much smoother, and the worst that could happen is that it may crash, especially if it is an older title.

Just like how you need to try on a new shirt to see if it fits your size, it’s best if you try HAGS yourself to see if it suits your needs.