Outriders is one of the latest games to receive Nvidia’s DLSS technology. I’ve discussed DLSS previously in my reviews and analyses of Death Stranding, Watch Dogs Legion, Call of Duty Black Ops Cold War, and Cyberpunk 2077. I have long supported Nvidia’s machine learning technology as it continues to provide superior image quality and higher framerates – a rare win-win. For today’s analysis, I wanted to explore the performance savings incurred with DLSS in Outriders, in addition to understanding any potential quirks and impact to image quality.
The PC I used for this analysis includes the following specifications:
- Intel i7 8700k @ 4.7 GHz
- 16 GB DDR4 @ 3200 MHz
- Nvidia GeForce RTX 3080 FE (OC +100 core / +500 memory)
- Installed to SK Hynix 1 TB m.2 NVME
- Display: Acer Predator CG7 4K @ 120 Hz
Keep in mind, Outriders (our review here) is built on Unreal Engine 4. If you know anything about UE4, you understand its propensity for pumping out visually impressive titles. Look no further than Gears 5 on PC. Additionally, UE4 has quite possibly the best TAA (temporal antialiasing) solution I’ve seen. This TAA is what allows UE4 titles to look visually noise-free, yet maintain clarity.
Therefore, the key question I had going into this analysis was not whether I’d see an improvement to visuals and performance. Rather, I wanted to understand to what extent those improvements would manifest. To that end, I selected three DLSS values to test: DLSS Off (native 4K resolution), DLSS Performance (internal 1080p), and DLSS Quality (internal 1440p).
Furthermore, I maxed out the visual settings in Outriders. Unfortunately, the number of settings available to tweak in Outriders is frustratingly limited. For example, things like lighting and texture filtering are just bundled in with other settings. As a PC gamer, I hate this. Give me all the options and let me decide which ones to tweak. For players who don’t feel like adjusting settings, that’s fine. This is why a Quality Preset option exists. I cannot believe we’re still dealing with limited PC options in 2021. This lack of granularity is something that will come up later in this analysis.
I measured three distinct sections of the game to understand visuals and performance. These included the intro cutscene, the hub area, and combat. I selected these deliberately to test consistency, contained density, and open dynamic effects of the engine, respectively.
Let’s start with the intro cutscene. Any cutscene in Outriders is a good de facto “benchmark” of sorts because it’s repeatable. To that end, I observed the following performance numbers.
Right away, the obvious observations become immediate. DLSS Performance provides better average framerates than Quality and native 4K. But the scaling here is interesting. On average, DLSS Quality provides 31% higher performance than native 4K, with DLSS Performance nearly doubling that improvement and providing a whopping 61% increase to performance compared to native 4K. Keep this rough doubling in performance improvement scaling from DLSS Quality to DLSS Performance in mind moving forward. Of course, cutscenes are not playable, therefore any improvements seen here are of academic value at best. Still, the scaling here is impressive.
Next, let’s take a look at the hub area. For this test, I walked a circuit from the fast travel point, up to the observation deck, and then back to the fast travel point. Here are the numbers.
It’s here where the first inklings of some potential interruptions crop up. Once again, we observe a roughly doubling in performance improvement in average framerates when moving from DLSS Quality to DLSS Performance.
However, take a look at the last column, something I called 1% Low Percentage. This indicates how far below the 1% Low framerates are from the Average framerates. You’ll notice that the 1% Lows in DLSS Performance are 47% below the Average framerates. Comparatively low 1% Lows result in more erratic frametimes which manifest as stutter. This is why while average framerates are important, better frame consistency – or how long each frame persists on screen – is the more important measurement for perceived smoothness.
Finally, let’s take a look at combat numbers. Combat is by far the least consistent activity to measure in any game which makes accurate repeatable measurements almost impossible. However, try we must. Here are the numbers.
Once again, notice the roughly doubling in performance improvement in average framerates from DLSS Quality to DLSS Performance. But more worryingly, note the 1% Low Percentage measurements. At 35%, this difference is far greater than either DLSS Quality or native 4K. And during combat, this difference was felt as I experienced stutter. In a highly reactive activity like combat, stutter is detrimental to your survival. This is where the folks at People Can Fly can improve still.
Let’s shift gears and take a look at the visual impact. Overall, the image provided by both DLSS Quality and DLSS Performance is largely superior to native 4K, with a few exceptions. For example, look at this image from the battlefield.
You’ll note that both DLSS Quality and Performance resolve greater detail in the distant trees than native 4K. However, both DLSS options antialias the beams in the middle of the image better than UE4’s TAA in native 4K. This is yet more proof that DLSS is more than just a fancy image upscaler. It’s a superior antialiasing solution outright.
There were, however, a few oddities and quirks I observed in my analysis. For one, I noticed some distant shadows cast by this meshed structure presented more visual dithering in DLSS Quality and DLSS Performance compared to native 4K. Perhaps “dithering” isn’t the accurate term to describe what I’m seeing, but it is curious nonetheless. I’m not sure what the cause could be. But given that it’s relatively far in the distance, impact to gameplay is minimal.
What is more obvious is the volumetric lighting. Remember how I said earlier that developers need to provide more graphics options? Because Outriders’ graphics settings do not have dedicated settings for lighting or volumetrics, seeing something like this is frustrating.
You’ll notice here that the resolution of the volumetric light appears to be tied to internal render resolution. This is easily observable due to the obvious stair-stepping (or aliasing) of the froxel grid in this light shaft. Note how the aliasing of the light shaft becomes more pronounced as we move from native 4K to DLSS Quality to DLSS Performance. This is because the internal render resolution is moving from 4K to 1440p to 1080p.
This quirk is something which I feel could be potentially circumvented by the user if greater granularity of graphics options were exposed. Perhaps an option could be provided to decouple volumetric rendering from internal resolution. Perhaps it’s a limitation of UE4. Either way, to not have a lighting option at all only compounds my frustration with limited settings exposed to PC gamers.
Overall, I remain impressed with DLSS. The scaling I observed here is impressive to say the least, and should theoretically scale across more mainstream cards like the RTX 3060. It’s clear that Nvidia’s bet on machine learning and AI have paid and continue to pay big dividends. DLSS is not magic. It’s a genuinely innovative solution built on intelligent engineering and breakthrough R&D. DLSS can and should be implemented in every single PC game.
I have said several times that I believe this decade to be the decade of machine learning and AI. If we can achieve superior image quality and superior performance in just the first few years of this decade by leveraging this AI, what does this mean moving forward? This is why I remain excited about DLSS, not because of what it’s doing today, but what it represents for the breakthroughs of tomorrow.
Disclosure: Outriders PC code provided by Nvidia.