> ## Content Index > Fetch the complete content index at: https://www.siliconinsights.org/llms.txt > Use this file to discover other available public pages before exploring further. # Who is DLSS 5 even for? - URL: https://www.siliconinsights.org/who-is-dlss-5-even-for/ - Published: 2026-09-01T14:39:12.000Z - Updated: 2026-09-01T14:39:12.000Z - Description: DLSS 5 has more baggage than just visuals. - Author: Matthew Connatser - Tags: Analysis The first iteration of DLSS had a simple premise, using AI to essentially improve the resolution of a game without impacting performance at all. Then DLSS 2 refined this, DLSS 3 added frame generation, and DLSS 4.5 added in better denoising for ray tracing. DLSS was all about improving visual quality and framerates just by toggling an option or two. DLSS 5, the latest iteration, finally releases on Thursday the 3rd and also leaked last week, so we know much more about the final product now than we did in March when the technology was first detailed. Unfortunately, DLSS 5 seems to be almost the complete opposite of what DLSS used to be about: it doesn't provide a clear upgrade to visual fidelity, it reduces performance significantly, and it will probably be very tedious for gamers and developers to apply correctly. When we first saw DLSS 5 earlier this year, most of the focus was on how it altered the appearance of games, since that's all there really was to analyze. I don't wanna retread that whole thing, especially because my opinion hasn't really changed. In some instances, I think DLSS 5 does add some nice definition to textures and lighting. Some of the screenshots I've seen today look pretty good. In most instances however, DLSS 5 transforms graphics into something I'm not a huge fan of. To be clear, I haven't tried it hands-on. Alterations in color and lightning seem to go too far quite often, morphing the ambience and overall feel into something a game wasn't supposed to present. It would really take me out of the experience if I felt like I was playing from the perspective of an iPhone camera, and that's kind of the feeling I get from looking at screenshots and videos of DLSS 5 in action. ![](https://storage.ghost.io/c/1b/09/1b097c8c-01b6-4cc8-b402-d44b1cbfdc1b/content/images/2026/09/Screenshot-2026-09-01-094832-1.png) I think this screenshot is actually far more egregious and more clearly details what DLSS 5 does to Grace. DLSS 5 also makes certain assumptions about the way people should look. We've already seen how DLSS 5 "yassifies" Grace from Resident Evil and gives her a full face of makeup and even a (spray) tan for no apparent reason. Screenshots by The Verge appear to show something similar happening to Laura in Shadow of the Tomb Raider; in the original screenshot, her face does have kind of a makeup-like texture, but with DLSS 5 enabled that texture is really amplified, and also her lips take on a lipstick-like color. Much as early visuals haven't really changed for DLSS 5, performance doesn't seem to have substantially changed either. Initially, it required two RTX 5090s to work, which is insane. When DLSS 5 launches this week, it will work on all 50 series GPUs, but tests using the leaked version indicate that it roughly halves performance. This puts it about in league with enabling ray tracing on a 20 series GPU in 2018, which is not a good thing. Honestly, I think the performance alone could kill interest in DLSS 5\. Gamers really didn't tolerate ray tracing until enabling it only costed a modest chunk of performance, so I doubt DLSS 5 will receive a different treatment. Sure, it is possible to claw back some of that performance with plain old DLSS upscaling and frame generation, but the former can only do so much, and the latter isn't universally popular since it increases latency and introduces some visual errors. Nvidia claims that the version of DLSS 5 today runs five times faster than the version we saw in March; in the time that the original DLSS 5 could render a single frame, DLSS 5 right now can render five frames. Nvidia projects that some time later this year DLSS 5 will be seven times faster than it was in March, or about 40% faster than it is currently. That would be better, but the performance penalty from enabling DLSS 5 would still be about 30%, about on par ray tracing on modern GPUs. The final problem I foresee with DLSS 5 is all the tinkering. DLSS 5 has an intensity level that goes from 0 to 200%, tone and structure sliders that go from 0 to 100%, and three different models termed A, B, and C. Some parts of a game's render may be masked out, anything from faces to objects. That sounds like a good thing, but I suspect it's going to overcomplicate things for developers and gamers. While DLSS upscaling and frame generation aren't perfect, it kind of just worked. Sure, sometimes things looked smeary or introduced weird anti-aliasing artefacts, but nobody ever complained DLSS 2 or 3 or 4 just made a game look totally wrong. When game developers implement DLSS 5, they'll be able to configure it to their liking based on those levels, models, and maybe that masking tool I mentioned earlier. When DLSS 5 is enabled, gamers will use the settings the developer set; without mods, all anyone will see in the graphics settings is a toggle to turn DLSS 5 on or off. Here's what I'm worried about: devs already spend lots of time on art direction, getting the ambience right, and establishing the color palette. When they start integrating DLSS 5, it completely changes how everything looks, and now these developers have to configure several different settings in such a way that their application of DLSS 5 represents the game correctly everywhere. Maybe they overlook some things, maybe it only looks good on certain hardware, maybe they phone it in, or maybe the implementation just isn't done well. So, if gamers don't like the way a game has implemented DLSS 5, they can try fixing it themselves with their own settings, but that kind of completely goes against the original selling point of this brand of technology. It can already be really time consuming to get the colors right on a display, I can't imagine compounding that with what is essentially a very cutting-edge filter. Don't get me wrong, I do like to tinker with my technology, but there's a limit, right? I bought my first OLED TV this year, and by the time I was done getting the settings where I liked them, I was *really* done with it. On games with HDR, I have to fiddle with those because sometimes the brightness variation is really annoying. Personally, I don't want to add another list of settings into that mess, let alone one that's on a game-by-game basis. Of course, I don't own an Nvidia GPU at the moment, so it's not exactly a dilemma I need to ponder, but still. If something similar came to AMD or Intel GPUs, I'd probably not use it. And that's a shame, because DLSS 5 can absolutely look good. Like ray tracing, it's impressive this is able to be done in real time at all. But it can't both be a one-click remaster button and also look correct to everyone in every instance. I'm not dying for more photorealism in my games or "Hollywood VFX." I'm sure that sounds great to someone! I just don't know them. If I have to be optimistic for a moment, I hope that DLSS 5's performance can be improved and that game devs find it easy to implement this technology in a way that everyone likes. If I'm completely wrong and DLSS 5 actually turns out to be great, I'd be very happy with that! Personally I'd be happy if DLSS 5 could subtly but meaningfully improve visuals; I don't need a remaster.