COD Warzone 3 Latency Guide – Best Settings for Performance and Input Lag

This isn’t the first latency guide I’ve done, so if you want to see the other games I’ve tested, or find out more about what I’m testing and the wider factors you might want to consider, I’ll leave the playlist in the cards above. Warzone 3 is a return to form, with the worst result being near-on three times slower than the best, so – unlike VALORANT – it’s a lot more important to run with the right settings. Luckily, I spent two whole days painstakingly collecting data on live servers for the most accurate data, and have a collection of results to run through. If you enjoy these videos by the way, let me know by hitting the like button!

So, starting with the high-end – if you happen to want visual quality out of a COD game for some insane reason – we’ve got the Extreme preset with no DLSS or NVIDIA Reflex, and the same again but with Path Tracing. Much to my surprise, adding Path Tracing does next to nothing to the results. It is ever so slightly slower on average, sure, but considering most standard ray tracing type features murder performance, seeing this feature – which is basically DLSS Ray Reconstruction – not hurt performance almost at all is mighty impressive. In fact, looking at the FPS results you can see a frankly miniscule difference – just 1 FPS on average between them. Naturally, you’d likely want a bit more performance, so let’s add in DLSS to see how that does. Unsurprisingly, that bumps the performance up considerably, going from 75 FPS average to 109 FPS average, and raising the 1% lows to well above even the average performance without DLSS. On the latency front, that bump in performance alone makes for a significant improvement, dropping from around 50 milliseconds, to just 31 milliseconds. That’s an incredibly significant jump, and clearly it’s well worth enabling DLSS here, both for outright performance, and for latency. Finally, at least for the extreme preset runs, let’s enable NVIDIA Reflex – a tool that is specifically designed to help lower the input latency. Somewhat unsurprisingly, it does help, but only by 0.7 milliseconds. It’s also a touch more unstable, with a slightly higher maximum, and slightly lower minimum. Still, an improvement is always good.

Let’s move down a preset, down to Ultra, with DLSS and Reflex enabled, but Path Tracing disabled. Surprisingly that’s about the same as Extreme with DLSS and Path Tracing! Clearly there isn’t much of a difference between those two presets, so let’s skip down again to Balanced. Finally we see a reasonable improvement – over four milliseconds faster no less! Performance wise we can see why, as Balanced offers 117.5 FPS average, up from the best so far at 109 FPS average. Interestingly, that’s actually only about 0.7 milliseconds faster per frame on average, so there’s clearly some pipeline optimisation between the two modes that allows for the other 3 milliseconds or so of latency improvement. 

I want to jump down to the Minimum preset as generally that’s where you’ll get the best performance, and so we can try out some of the other myriad of options available here. Minimum on its own delivered fantastic performance, dropping another two milliseconds off the total, bringing us down to 25.9 milliseconds on average. It also delivered the best performance at 128.6 FPS on average, although again that’s only 0.8 milliseconds faster than the balanced result, despite the two millisecond drop in latency. Anyway, I thought it’d be interesting to see how FSR handles things – in several games I’ve tested including Rainbow Six Siege, FSR offered significantly lower latency than DLSS. Warzone was updated overnight during my testing to swap FSR 2.1 to FSR 3.0, so I’ve got the FSR 3 data here. Surprisingly, FSR3.0 seems to slow things down, both in latency at 28.7 milliseconds, up from 25.9 with DLSS, AND in framerate, dropping to just 111 FPS average, down from 128.6 FPS with DLSS on. That’s quite surprising, considering the results I’ve seen in other games – on the same RTX 4060 Laptop based machine no less! 

Ok, so if FSR doesn’t help, what does DLSS Frame Generation do? In theory, frame generation adds in “fake” frames to help smooth out your experience. This should bump the framerate, although it’s likely to add a decent amount of latency as the “fake” frame has to look at the current frame AND the next frame to then generate the fake one, then display the current and fake before finally displaying the next “real” frame. And yeah, looking at the latency, that’s dreadful. The average shoots up to 48.4 milliseconds, or nearly as bad as running at 75 FPS on extreme! The thing that surprised me though was the framerate data… With FrameGen enabled, I expected considerably higher framerates, but what I got was 125 FPS average – SLOWER than without FrameGen enabled! I’m at a loss as to why this is the case, so if you have any ideas, please do leave them in the comments below! Also, just because COD lets you, here’s what FSR3.0 plus FrameGen looks like… Yeahhhhh it’s bad. Really bad. It’s actually the worst result by 11 milliseconds. It’s over 10 milliseconds slower than DLSS and FrameGen – possibly unsurprisingly. The FPS data shows FrameGen was only 1 FPS faster than not too, which again just leaves me very confused.

Anyway, the final result I want to show you is actually a higher preset than minimum, called Basic. Surprisingly, on my hardware, in the specific test pattern I was doing, it was the fastest result I got, and by a reasonable margin. It averaged 22.9 milliseconds, which is 3 milliseconds faster than minimum with DLSS and Reflex enabled. Interestingly, the FPS data I captured at the same time shows basic wasn’t even the fastest framerate. It’s a close match to minimum, but it’s 5 FPS slower than the minimum preset. 

So, if you want the best performance, at least on an NVIDIA system, it seems DLSS is your best bet. Both minimum and basic provide great results – both if framerates and latency, and for the love of god don’t use FrameGen. It’s interesting to see how COD is markedly slower than games like Siege – even the fastest results from COD only just about surpass the worst results in Siege. The best result I got in Siege was 9.7 milliseconds, which is well under HALF the best result I got in COD. Maybe on a different system you can eke out better results, but at least on my machine that doesn’t seem to be the case. Anyway, that’s a look at COD Modern Warfare 3 and Warzone 3. Let me know what other games you want to see me test in future videos in the comments below!