So could this way of doing it make it easier to have BC with Switch and Switch 2 on the Switch 3?
Did you already know? | |||
| Yes, I was aware of it. | 11 | 73.33% | |
| This is new information to me. | 4 | 26.67% | |
| Total: | 15 | ||
So could this way of doing it make it easier to have BC with Switch and Switch 2 on the Switch 3?
Different games are harder or easier to emulate depending on how they run code. Those that stick to the rules and run within the development drivers/APIs or whatever they are called which are a software layer between the code and the hardware like Direct X can be more easily emulated as you can rewrite the software layer to work on different underlying hardware however if you get exceptionally clever coders who low level program consoles directly accessing the hardware features become dreadfully slow when emulated like the Rogue Squadron games on Gamecube. Those games seem to need a lot more processing power for emulation than most Gamecube titles. I assume the Switch 1 will be the same in that the best titles may be sluggish but Nintendo can write optimised T239 code to replace the really demanding code of the original game. It's one thing for Nintendo to make the effort for their games that sold millions but for smaller Switch titles that didn't sell well anyway I'm sure there are developers that can't be bothered to update as not in their financial interest to do so. I suspect many Switch 1 games will get overlooked as not commercially viable to rewrite code and without those optimisations may not run that well. I suspect we will get the full spectrum of results. Some Switch 1 games that are massively superior on Switch 2 and others that are terrible compared to running on the original Switch 1 hardware.

| Pemalite said: Xbox's backwards compatibility is much more complicated. |
That's a good point regarding the difference between Xbox backward compatibility and the Switch's approach. You're right that the Xbox required recompilation for many titles, resulting in a more limited compatible library, whereas Nintendo seems to be targeting full library compatibility for the Switch.
The core idea of a translation layer remains relevant. It's similar to how Proton enables Windows games to run on Linux or how Rosetta allows x86-64 software to run on Apple Silicon ARM. These layers are crucial for bridging architectural differences.


| KLXVER said: So could this way of doing it make it easier to have BC with Switch and Switch 2 on the Switch 3? |
Absolutely.
--::{PC Gaming Master Race}::--

Digital Foundry has done a quick check with Arkham Knight and Witcher 3. This shows what was stated in the OP, the translation layer will improve performance and resolutions for DRS games.
https://www.youtube.com/watch?v=NoE41LJyXh0
I watched a video with compatibility analysed and generally it was very good and much better than expected on the release day of the console but I understand the chipset does date back to 2020 so I guess they had plenty of time to get the code in a decent shape but still much better than expected. It seems like the Switch 2 CPU's are mainly running code directly like some of the best emulators with systems with ARM processors but the graphics are more difficult. The reason I say that is games that had inconsistent frame rates on the original Switch are now running at much more consistent frame rates overall but surprisingly the resolutions stay the same even to the point where games that dynamically scale with the load were showing the same traits as the original Switch. I realise Nintendo want to sell upgrade packs so this could be a commercial choice but I was still expecting the dynamic scaling to be more forgiving on Switch 2. I suppose it could well be the Switch 2 has some sort of interpretative layer for games without upgrade packs and it shows the Nvidia chipset is fairly different this time around.
I'm also curious about the graphics hardware between the two systems because the original Switch supports high performance but lower precision FP16 graphics so its 388 gflops at fp32 is actually doubled to just under 800 Gflops for fp16 precision same as PS4 Pro gflops doubling at fp16 but the Switch 2 doesn't have that doubling for fp16 and some Switch games are written and optimised for fp16 calculations. I wonder how they will be effected. A modded Mariko Switch overclocked can do about 1.2 Teraflops at fpt16 i.e. a 50% GPU speed boost and then a doubling of gflops due to fp16. I don't have a modded Switch myself but I believe on those you get higher frame rates but also dynamic scaling is reduced with regard lower resolutions. I just wonder if there is a floor on Switch 1 GPU emulation on Switch 2 and to get better emulation with higher resolutions you have to rewrite the code to optimise it for the T239.
Overall I have to say I'm super impressed with the reports of the Switch 2 and its functionality out of the gate but I still think it will be a hard sell for a lot of people at current prices.
