AMD’s NEW Anti-Lag 2 TESTED in CS2 – Radeon Anti-Lag 2 Latency Benchmarks

AMD is back at it again after their last faux-pas with Anti-Lag+, this time with a whole new Anti-Lag 2! But before we go any further, let me explain what the hell anti-lag is and what’s new with V2. Anti-Lag is AMD’s input latency reduction tool that, until now, has been totally in-driver. That means no matter the game, so long as you have an AMD GPU, you can enable Anti-Lag and lower your latency. Games don’t need to support it, because it’s all in-driver. I’ve done videos testing it in the past, and in the right scenarios it can make a massive difference. Anti-Lag has been around for quite a while, a good four years at this point in fact, so when AMD announced Anti-Lag plus around 9 months ago, it was a welcomed upgrade. Anti-lag plus was basically an upgraded version of anti-lag, that did some still in-driver, but more active messing around with the render pipeline type tweaks. This turned out to be a not-so-great idea, because messing with the game engine’s backend functions is exactly what most cheat engines do, and so Valve made it clear that anyone using Anti-Lag Plus would be getting a VAC ban. 

And that nicely brings us to Anti-Lag 2. This is a pretty big step for AMD, as this is no longer just a driver-only tool, but a game specific implementation that requires the game developers to build it into their game. The good news is that AMD is pretty committed to being open, and making tools like this open source and freely available, so really any developer that wants an extra feature to shout about can add this in and everyone gets a win. With that said, the only game so far to take AMD up on that offer is CS2, so that’s what I’ll be testing here. Now AMD says in their blog post about Anti-Lag 2 that it’s for Radeon RX 7000 series cards and 8000 series APUs, although the driver page says the driver is suitable for all modern AMD GPUs, so before getting to the RX 7900 GRE that AMD sent me to do this testing, I want to test with an older card. I’ve got an RX 6900 XT which happens to be right around the same performance as the GRE, but from the previous generation, so we can see if that makes any difference. The setting in CS2 does show up on both cards – it’s just at the bottom of the advanced settings tab – although AMD would recommend you enable the original Anti-Lag in the driver too, so you’ll need to head to the driver, then the game tab, then CS2, and enable it there. Otherwise you’re pretty much set!

Testing wise, as with all my latency testing videos, I’ll be using my very own open source latency testing tool, or OSLTT, that I both designed and manufacture right here at home. If you want a latency tool to test your own stuff, I sell them over at OSRTT.com, which is linked in the description. As for the testing itself, I use the mouse move function where the tool will instantly snap the mouse cursor left or right, and record the light level change as soon as it’s initiated the move. I prefer the mouse move function over left clicking and trying to capture muzzle flashes, one because there’s no animation delay in moving the mouse, two it’s more consistent both for the tool’s light sensor and for results itself, and three because it means I can do 250 moves in one go to get beautifully accurate data without having to reload a gun every 20 rounds. So, that’s the test, now let’s look at the data!

The way I see it, we have four possible combinations. Both versions of Anti-Lag off, in-driver only on, in-game only on, and both on. So that’s what I’ve tested here, in that order, and at least on this 6900 XT on low settings at 1440p, it looks like none of these settings really do much. In fact, with both Anti-Lag and Anti-Lag 2 enabled, we actually got the worst average of 8.4 milliseconds – although that’s only 0.2 milliseconds slower than the next result, and 0.5 milliseconds slower than the fastest result, so not exactly a wide margin. Throwing in the FPS data too, you can see that with both features disabled we actually get MORE performance than any other result, and inversely with both features on we get less performance. Now we’re talking about 475 FPS versus 509 FPS, so not what you’d call an appreciable difference, but it’s interesting to see nonetheless. That difference in latency might well just come from the FPS difference, with the slight exception of the double-off result, showing that especially the in-driver version is doing at least a tiny bit of magic in the background. But realistically this isn’t doing much.

So, is that because Anti-Lag 2 just doesn’t work on older cards? Well, no. AMD sent over this Sapphire RX 7900 GRE for me to run some tests with, and testing on the same low settings at 1440p you’ll see the same pattern of functionally no difference between the four modes. In fact, the spread is actually even tighter with just 0.2 milliseconds separating the fastest from the slowest. The most interesting thing though is looking at the FPS data, where the double-off result had a significant performance lead – 10% more performance at 555 FPS versus around or just under 500 FPS – and yet it had the slowest latency, again showing that the features are doing something, but still not exactly much. 

So how is it that AMD is claiming up to an average of 37% lower latency in CS2 with Anti-Lag and Anti-Lag 2? I mean they show receipts on their blog post, showing pretty significant gains – the GRE goes from 19 milliseconds to just 11 milliseconds with both enabled – so what’s the deal? Well, AMD tested with basically maximum settings, the Very High preset specifically, and at 4K. Now I’m going to stick with 1440p so the gains are going to be a bit less impressive here, but I will stick it on Very High so you can see the difference. Finally, we can see some difference – a 17% decrease in latency in fact! Hooray! You can also see that just the in-driver does less than the in-game feature, and that they do work together to help lower the latency overall. Adding in the FPS data and you can see a few interesting things. First, there’s functionally no performance difference between the modes – unlike at lower settings – and that enabling Anti-Lag 2 in game you get a less stable gaming experience. The 1% lows drop considerably compared to with that mode off – and bear in mind that this is just from the player standing still in an empty lobby moving the mouse slightly, meaning this is as stable as it gets, so a change here is potentially more significant than usual. Also the fact it’s consistent between both in-game feature enabled runs indicates a pattern. 

Personally I think most of this discussion is kind of mute anyway, since if you cared about latency you wouldn’t be playing on the Very High preset, you’d play on low like everyone else, and if you did – regardless of these anti-lag settings – you’d get less latency and more FPS than even the best result I got with both Anti-Lag and Anti-Lag 2 enabled on Very High. In fact, I’ve actually found a way to get exactly 37% lower latency in CS2! Swap from Very High to Low and you’ll get the 37% lower latency AMD is claiming these features offer! Amazing! But no seriously, for CS2 in particular, but really any competitive game, the lower the settings, the lower the latency and the more performance you’ll get. It’s that simple.

For more story based games where graphics quality often takes precedence over outright performance and latency, having a feature like Anti-Lag 2 built in will be a useful tool and the fact that it’s freely available for developers to implement is fantastic. We’ve shown here that it CAN make a difference, and so for the right games I’m actually pretty excited to see this rolled out – something like Cyberpunk could really do with having this built in – but for competitive games like CS2 or Rainbow Six Siege, you’re better off focusing on getting the most performance with the least post-processing features on, and you’ll get better latency than these features could ever provide on their own. If you want to know what the best settings for Siege or CS2 are, I’ve done full latency guides that I’ll link in the cards above and on the end-cards for you to check out.