I’m trying to settle on settings for a game where the image gets noticeably softer with upscaling. I replayed the same 17-minute section using DLSS 5, then switched to native resolution and lowered shadows and reflections.
Which option do you actually use: DLSS 5 with higher settings, or lower settings at native resolution, and why?
Don’t judge this from screenshots alone. Upscaling artifacts and softness usually stand out more during camera movement, especially around foliage, particles, hair, fences, and thin geometry.
If the softness is already bothering you, I’d choose native resolution and lower shadows or reflections. Those settings are often easier to sacrifice because you notice them mainly in specific scenes, while resolution affects the entire image all the time. Slightly simpler shadows fade into the background once you start playing. A soft picture rarely does.
The exception is when higher settings meaningfully change the presentation, such as ray-traced lighting or reflections in a game built around them. Then DLSS may be the better trade, provided it keeps motion stable and doesn’t introduce obvious ghosting. Sharpening can improve perceived crispness, but it cannot restore detail that the reconstruction lost.
I’d compare both options in the busiest area you can find rather than replaying a controlled section. Keep whichever gives you steadier frame pacing there. Between two otherwise playable setups, though, my vote goes to native with reduced shadows and reflections.
Native isn’t automatically the sharper option. Some games have awful native TAA with more shimmer and crawling than DLSS, so compare both while moving and check the render scale is truly 100%. If native still looks cleaner, I’d lower reflections first and keep textures high, since texture quality usually costs VRAM more than raw frame rate.
The hidden downside of lowering settings is that you can end up with sharper pixels but a noticeably flatter-looking game. I’d still choose native if the cuts are limited to shadows and reflections and frame pacing stays smooth. @novapanda is right about judging it in motion, though. Stable detail matters more than a crisp screenshot.
Realistically, neither option will keep the game looking exactly as intended while gaining the same amount of performance. You are deciding which flaws annoy you less, and softness is a difficult flaw to ignore because it follows the camera everywhere.
The confusing bit for me is that “DLSS looks soft” can actually be several effects stacked together. Motion blur, depth of field, film grain, chromatic aberration, and an overly aggressive anti-aliasing pass can make reconstruction look much worse than it does by itself. I would disable those effects temporarily and compare again. Leave sharpening near its default at first, since cranking it up can turn a soft picture into a grainy, haloed picture rather than a detailed one.
I would pay attention to the output resolution too. Upscaling has far less information to work with at lower resolutions, so a mode that looks clean on a high-resolution display may look muddy on a lower-resolution monitor. That could explain why other people insist DLSS looks great while you immediately notice the softness. They may be judging a completely different input and output resolution.
If native only requires reducing shadows and reflections by a level or two, that seems like the less distracting compromise. I would avoid lowering textures, geometry detail, or draw distance unless necessary, because those changes can be much more visible during normal play. Reflection quality is usually where I would start, especially if the game uses expensive effects that only appear on certain surfaces.
My choice would be native in your situation, but I would stop testing the same section once the basic comparison is clear. Replaying a scene repeatedly can make tiny differences seem more important than they are. Pick one setup, play normally for a while, and see whether you keep noticing the reduced effects. If you forget about them but immediately notice the softer image after switching back to upscaling, native is the right answer for you.
The hidden downside of native is paying for pixels you may not actually see from your seat. On a large TV across the room, I’d take DLSS and keep the richer lighting. At a desk, softness is much harder to ignore.
Don’t compare with zoomed crops. Sit where you normally play and switch between the two. If native still looks clearly better without wrecking performance, lower the expensive effects and use it.
My cutoff is simple: occasional weaker reflections are acceptable. A permanently smeared image is not. If you keep noticing the softness during ordinary play, DLSS has already lost this particular argument.
Check the actual frame difference before you agonize over any of this. If native with lowered shadows and reflections already lands on your frame target and holds it, DLSS never enters the picture. The whole trade only exists when native can’t reach the number you want without dipping into stutter. So run both through your busiest scene with a frame time graph up, not just an average fps counter. Average fps hides the drops that actually ruin a session.
@git_dan83 is right that native isn’t automatically cleaner, and I’d go a step further. A lot of what people call ‘native’ in modern games is native resolution running through a heavy temporal pass that smears just as much as upscaling does during motion. So you might be comparing two blurry pictures and picking the less blurry one. Worth knowing which TAA implementation the game ships before you decide native is the sharp option.
Where I’d push back a little is the desk versus couch framing from @ghostpanda71. It’s true as far as it goes, but pixel density matters more than raw distance. A dense small panel up close can hide reconstruction softness better than a low density big screen at the same viewing angle. Distance is only half the equation.
My honest take: soft image is a constant tax, weaker reflections are an occasional one, and I’ll take the occasional tax almost every time. But that assumes your frame pacing is smooth either way. If native forces you down to settings where you start cutting textures or draw distance to hold the frame rate, that’s the point where I flip and take DLSS, because those cuts are far uglier in normal play than a bit of softness. Pick the version that keeps frame times flat, then let the visuals decide the tie.
If native only barely reaches your target, I would keep the upscaler. A setup that works in one repeatable section may fall apart later during weather effects, large fights, or heavier locations.
That said, I would try native with reduced reflections and shadows if it leaves some GPU headroom rather than sitting at full load constantly. The cleaner image is worth more to me than effects I might only notice occasionally, but not if every demanding scene causes sudden dips or extra input lag.
I’d set a modest performance buffer and judge from there. If native cannot maintain that buffer without cutting draw distance, textures, or lighting too far, switch back to upscaling. Sharper output is nice, but a settings profile with no room for the game’s worst moments can be misleading.