How will be Switch 2 performance wise?

[Chrkeller]

Soundwave said:

JRPGfan said:

FSR at 1440p->4k is almost as good as DLSS.
The area where DLSS shines, compaired to fsr, is at low res upscaleing.
Like 540p -> 1080p ect (thats where DLSS kicks FSR's arse).

No joke, but at higher resolutions, DLSS vs FSR, is near identical.... esp if your sitting on a bit away from the TV/Monitor.
You need to zoom way in, and take screenshots/recordings to spot minor differences between the two.

Well the Switch 2 should never be rendering at higher resolutions to begin with. It's a monumental waste. 

720p to 1440p DLSS looks terrific. Most normal gamers sitting on their couch several feet away from even 65-77 inch display are not going to know any better. 

540 to 1080p DLSS is a laugh, you can approximate this if you have DLSS titles like Alan Wake II or Cyberpunk 2077 ... set the game to "windowed" mode on your full size monitor, in that case I was getting a display image of roughly 13 inches on my 4K monitor PC monitor (way larger than a Switch 2's hypothetical 8 inch display but at least in the same ball park) and even at 13 inches display ... toggling back and forth between native 1080p and 540-to-1080p DLSS is extremely hard to tell a difference. 

A developer would be utterly stupid IMO to try and brute force native res on a machine like the Switch 2. You're asking the system to push so many more pixels for almost no benefit. If I'm a Switch 2 developer working on even a remotely challenging type of port for the system, I'm sticking with 540p (1080p DLSS) and 720p (1440p DLSS) as my main resolutions and not a single pixel more. 

A ROG Ally is basically a portable PS5/XSX ... like on no planet can you seriously state that's "just a PS4" when its running Starfield and Alan Wake II in playable states even at 15 watts only. Indoor areas it even climbs above 40 fps for Starfield and some areas in AW2 too as well. We have portables that can already run PS5/XSX tiers, we've had it for a while. The Steam Deck is almost 2 years old. 

The bolded part is where we disagree.  You seem to think running the same games makes hardware "basically" the same.  A rtx 2060 runs anything that a rtx 4090 does.  But saying the rtx 2060 is "basically" a rtx 4090 is nonsense.  I mean 1,920 cuda cores versus 16,384 is a slaughter.  

Likely we need to agree to disagree.  But I think "running the same games" is a terrible metric for hardware power.   We live in an age where game engines are scalable.  A wide variety of grossly different hardware can run the same game, just at massively different levels of fidelity.

Last edited by Chrkeller - on 13 January 2024