NVIDIA Reflex & Boost vs. AMD Anti-Lag & Radeon Boost under DirectX12 – Image Quality, Performance and System Latency in Practice!

1 - Einführung und Testsystem 2 - Testmethodik und Benchmarkruns 3 - Die Systemlatenz bei niedrigen Einstellungen 4 - Die Systemlatenz bei ultra Einstellungen 5 - Die dynamische Latenzmessung 6 - Zusammenfassung und Fazit

Ultra settings at 1440p and DXR on

Since I haven’t changed anything about the in-game test conditions, we’re good to go. In the following I only changed the settings ZERO off/on, Reflex & Boost off/on as well as AMD Anti-Lag off/on and Radeon Boost off/on under static conditions. Here we go…

The RTX 3060 Ti in detail

Here the RX 6700 XT

Yes, I couldn’t believe my eyes either, but what you are seeing is constantly so measurable. Radeon Boost just didn’t do what I expected here. So I had to ask what was going on. Basically, if you enable Radeon Boost, then Anti-Lag (figuratively) goes off. This is how all AMD GPU owners see it in the driver settings. Now the but! According to AMD itself, Anti-Lag is a sub-function of Radeon Boost and is automatically triggered upon activation.

You could also see this in action very well in Overwatch under the static conditions. Only the conditions in Overwatch regarding the game settings were something completely different. In Overwatch, I tested in 1080p with 200% render scaling. In response, AMD told me that Radeon Boost was not created to up-scaling back down. That’s one of the reasons I switched to 1440p native. Although it worked great in 1080p with 200% render scaling, check out a slide on CoD MW in 1080p with 200% render scaling. Now it’s getting fun!

Same scene and there was no movement in the game. Radeon Boost on and Anti-Lag works statically like in Overwatch! I have discussed the results from the 1440p and the 1080p tests with AMD in several meetings. The statement from AMD currently sounds something like this, “Radeon Boost only triggers anti-lag when momentum kicks in.” That’s paraphrased here, but the substance is what was said to me. Currently AMD’s driver team is reviewing my measurements and other details I provided regarding this behavior of Radeon Boost or Anti-Lag. Since it was the exact same thing in Overwatch! So let’s move on to the 1440p data on slide one on this page. Here are the details of the measurements. Oh, I almost forgot… NULL can’t show anything here, of course, because it’s not supported under DX12. But NVIDIA Reflex & Boost bring a whole 11 ms. After all! So now let’s go and don’t forget to look at the Frametimes! 

And as you can see, you can’t see anything. Wait, let me differentiate a little bit.

Here with the RTX 3060 Ti, you can’t really make out a difference now. On average, this is probably on par, so NVIDIA Reflex and Boost here brings the desired effect from the implementation in the game and the framerate limiting I suspect. Let’s take a look at Anti-Lag off vs Radeon Boost on.

Here still anti-lag off vs on.

And Anti-Lag, as we saw on the first slide, brings a full 6 ms. But you can’t tell from the Frametimes! Here, the driver intervenes in the render pipeline to provide latency reduction. If you’re looking for the 4ms latency difference between the RTX 3060 Ti and RX 6700 XT in terms of frame time, you won’t find it here either. Since external interventions through the Radeon driver as well as the implementation of NVIDIA Reflex are the main culprits for the obfuscation! OK, let’s have a look at the dynamic tests and whether Radeon Boost now shows where grandpa gets the most.