By using this site, you agree to our Privacy Policy and our Terms of Use. Close
sc94597 on 04 February 2026
Edit Thread
bonzobanana said:

I would completely agree with that on other systems typically PC but on Switch 2 I think you have to factor in its a system with decent GPU resources but very weak CPU performance and we have seen many Switch 2 games with horrible input lag using DLSS so you can't just use the PC norm for Switch 2. PCs typically have very high CPU performance where DLSS was designed for. Games like Skyrim, Cyberpunk and Street fighter come to mind with DLSS and high input lag. This is why I think Nintendo have set CPU resources far too low on Switch 2 its not a comfortable or well aligned level for the GPU. There is a CPU cost for DLSS and on a PC that is insignificant but on Switch 2 it is fairly high. The display panel introduces a lot of lag in itself too so for portable mode its a bigger problem. There was a report before Switch 2 release that portable mode the CPU ran at 1.1Ghz and docked mode 1Ghz which seemed very strange but maybe that really is a thing to help compensate for the very slow panel and lack of overdrive. I mean Codemasters are experts at DLSS and have supported it on their games from the very beginning (F1 2020) but on Switch 2 they don't use it. You don't make decisions like that without a reason. At t he end of the day every bit of data I have read shows the Switch 2 has low CPU resources without question and I would say most of these Switch 2 issues are related to that. However don't get me wrong I'm sure the vast majority of Switch 2 games will have enough CPU resources its just when they try to push the system these sort of problems emerge and there will be sacrifices. Skyrim Anniversary Edition first release had input lag up to 240ms are you saying without DLSS that would be 300ms or more? Their fix was mainly turning off Vsync so the frame data was used by the GPU as soon as available. A passmark CPU score of about 2000 has been available to top end PCs as far back as 2006/2007. Even an old i5 from 2015 can achieve about 6000 passmark CPU score. Steam Deck is 9000, PS5 about 16,000. 

  • CPU Bottlenecks: If your system is heavily CPU-bound (weak CPU, strong-ish GPU), DLSS might not improve frame rates, and consequently, will not reduce input lag, and in some cases, might introduce stutter. 
  • CPU Bottleneck: If your CPU is too weak to handle the game's logic, DLSS will not help FPS and may slightly increase latency due to the overhead of the upscaling process.

No offense, but you have shown again and again in this thread (and apparently historically on the forum for nearly a decade), but every few months or so in this thread in particular, that you'll concoct a reality that isn't there to confirm your bias. For example, this hypothetical (but non-existent) "CPU cost for DLSS." 

The CPU has little specifically to do with DLSS upscaling/AA. DLSS upscaling/AA inference is very much a pure-GPU workload computed on the tensor cores (convolutions, fully connected layers, etc) and CUDA cores (activation functions, pooling, batch normalization, regularization) in the case of CNN models. 

Yes, if a platform is experiencing a CPU bottleneck, reducing GPU-loads (like what happens when using DLSS as a performance crutch) won't do much to increase frame-rate, but that goes in both directions. DLSS isn't going to suddenly cause input latency that wasn't already there because of the CPU bottleneck. And it is still useful to use DLSS to improve image quality, even if you are (or rather, especially when) you're CPU bottle-necked and want to utilize more GPU resources. We see this in many games on Switch 2 that have really good image quality, but are locked to 30fps, likely because of that CPU bottleneck you're pointing out. 

The SW2 does have a weakish CPU for 9th Generation console standards, but that CPU is still more powerful (albeit only slightly more) than the very weak Jaguar CPUs in the 8th Generation consoles. If you're forming a hypothesis about why a Switch 2 game is performing poorly that depends on the weak CPU being the cause then it will have to also be present on these 8th Generation platforms you're comparing it to that have even weaker CPUs. 

Finally,

  • Grid Legends doesn't support DLSS on PC, so it isn't surprising that it doesn't support it on Switch 2. 

  • I cannot find any analysis that has found Cyberpunk 2077 to have much higher input lag on Switch 2 than on other platforms. 

  • Street fighter 6 has a similar input latency on Switch 2 to PS4 Pro - 70-72ms versus. 55 - 58 ms on the other 9th Generation platforms. At the maximum this is about half a frame of extra latency for the 8th Gen platforms + Switch 2 versus. 9th Gen. That's not horrible. 

  • Skyrim is an especially poor example, given that it doesn't have these issues at all on systems with weaker CPUs, including the original Switch of course. This is a software optimization issue in a game originally developed by a company well-known for its software optimization issues. It's not a hardware limitation. 

Edit: Also let's not mix up input latency and pixel response time. Street Fighter 6, as an example, has the same input latency in handheld and docked mode. Its screen's poor pixel response causes ghosting and blur, but the input latency is the same as in docked mode. Both response time and input latency influence the feeling of "responsiveness", but they aren't the same thing. 

Last edited by sc94597 - on 04 February 2026