DF: Switch 2 vs Steam Deck: Cyberpunk 2077 Benchmarked - Docked & Handheld Tested

So now the Switch obviously performs better, we have people starting about cheaper games and storage in a comparison thread between 2 versions.....
Talk about moving the goalpost.
moving-goalposts-goalposts.png
hBXi8SCpBrXnQdWG.jpeg
 
The image quality difference in favour of the Switch 2 is insane. Such is the power of Jensen and DLSS. I hope Valve would consider Nvidia for an eventual SteamDeck 2.
 
Last edited:

Completely accurate. Why would anyone be a hater of any platform in the first place? Keeps the shrinks in business I guess.

Nintendo hit a home run with this one. It is as good or better than what anyone knowledgeable could have hoped for at this power draw, in this form factor, in 2025 currency. Bizarre that people are shitting on it for not being something that it isn't aiming for.

And by the way, Cyberpunk running decidedly WORSE on Switch 2 than Steam Deck was a common prediction ever since the Switch 2 specs leaked way back when (the T239 specs and clocks). That would have been the default assumption, which is why this comparison video is triggering the Switch 2 haters I guess. Reality didn't conform to their expectations, and so reality must be wrong and reeeeeee.
 
Completely accurate. Why would anyone be a hater of any platform in the first place? Keeps the shrinks in business I guess.

Nintendo hit a home run with this one. It is as good or better than what anyone knowledgeable could have hoped for at this power draw, in this form factor, in 2025 currency. Bizarre that people are shitting on it for not being something that it isn't aiming for.

And by the way, Cyberpunk running decidedly WORSE on Switch 2 than Steam Deck was a common prediction ever since the Switch 2 specs leaked way back when (the T239 specs and clocks). That would have been the default assumption, which is why this comparison video is triggering the Switch 2 haters I guess. Reality didn't conform to their expectations, and so reality must be wrong and reeeeeee.
Not blindly loving =/= hating. Things are not only 8 or 80.
 
Not blindly loving =/= hating. Things are not only 8 or 80.
They probably think I hate the Switch 2 despite owning 103 Switch games, and 6 Switch 2 games, the console itself, two cases, the new Pro controller, and the new webcam. Being factual about hardware doesn't mean you have to blindly love or hate something.
 
Last edited:
It's native 720p. Absolutely nothing crap about that. DLSS 3 is not better than native at those resolutions, and no serious PC gamer has ever bothered to make that claim.


I already explained why, as XeSS allows you to get back more performance, especially in Dog Town. Unless you think playing well under 30fps is good? Because that is what happens when you drive anywhere in Dog Town, be it with the Switch 2 or if your play at native on the Steam Deck. Switch 2 can drop to a buttery 18fps in those scenarios, at least with the deck you can trade in a bit of visual fidelity for a closer lock to 30fps. Hence the obvious XeSS recommendations.

Yes, Richard said image quality is cleaner when comparing the Switch 2 to the Steam Deck when using FSR or XeSS. Notice they never said anything about it looking better than native at those low pixels counts, because only the most extreme PC fanboys would ever claim anything like this. Anybody who knows anything about PC gaming knows that at the more your resolution drops, the more DLSS 3 struggles to offer a comparable or better image as native. You have a PC, I'll reiterate, test it for yourself. You'll quickly notice that while DLSS has the upper hand in stability, in looses in clarity, especially in motion.

Hardware Unboxed did a good video about this:



That's at 1080p no less, dropping the resolution further does DLSS no favours.

Of course I'm serious. Don't try and gaslight me, or does Gigabowsers nonsense not count? I can pull up dozen posts on this forum which claimed the Switch 2 will outperform the Steam Deck in portable play, including yours.


PC DLSS CNN model is not equivalent to Switch 2's, switch 2 is much better at low resolution, it's been compared for same resolution on PC close to transformer model. There's a reason DF was blindsided for 3 weeks thinking the games were running native and no DLSS.

TAA implementation for CDPR is full of artifacts, one of the worst implementation since Halo Infinite.

But then, you say that it has to use XeSS to get back more performance.. so why are we even discussing TAA? You can't have both.
 
Last edited:
The large screen, kickstand and thin profile make it the perfect device for tabletop mode, where you can use the Pro controller. What is more comfortable than holding a portable device with an ergonomic grip? Not holding the portable device at all!
I don't really use table top mode tbf. The idea of carrying an additional pro controller with me as I'm out and about sounds very unappealing.
 
Last edited:
Rog ally xbox I doubt, I mean they are inefficient with so much compute and bandwidth starved, but they'll brute force their way. But one thing remains, they are stuck with non AI FSR. So...
Well not really - XeSS just gets faster on them (to the point of being useable).
The thing is though - Deck (and to a lesser degree 6800U/7840U), perform rather dreadful with TAA (all of them, DF missed this bit - but even regular TAA cuts the native performance rather dramatically on these handhelds), still on the account of the same bandwidth starvation - running desktop optimised shaders. Like I've seen games lose 15-20% performance by enabling TAA (no upscaling involved at all) on these chips - to the point where MSAA would perform better on the same game, even running at a higher native res!
Meanwhile Switch - even before Switch 2 - received actual handheld optimised variants (even for FSR) that performed considerably better under the same constraints.

Anyway - the 'AI 9 HX 370' / Z2e though at least in my testing so far do a lot better (whether it's just having that much more brute force or dealing with memory contention better).
And AI9 (not Z2e though) also has enough AI acceleration hw that it may actually run a version of FSR4 some day as well - just a question if AMD will bother with RDNA below 4 or not (or you know - if they open sourced it again this time...)
Of course the chips can also run 35W+ - there's a 'real' docked performance to leverage here.

I do think one area these tests undersell is that the CPU delta is significantly in favour of the Deck though. Even more dramatic on other APUs - that's where many of these devices outperform (quite easily I might add) even the PS5, heck there's a few games I can run at double the Series S framerate just because of the CPU.

Makes me wish either Sony or Microsoft would have went with Nvidia to spice things up.
Competition matters for sure - after Intel botched things, AMD had this entire market to themselves.

But I'd say even having a true dedicated handheld device would make a big difference for efficiency - note that Deck consumes over 10W for - non APU stuff when at 15W TDP, and many of these higher end devices are even worse. Now - many are built to dissipate in excess of 50W - so it's not really a 'problem'(cooling assemblies are really robust) - but the power optimisation is nowhere near of what console-platform makers have been known to produce, long before we talk about the chipset usage.
And then we get to things like having 12c/24t CPU inside latest handheld APUs which is just stupid overkill - these aren't elegant devices when it comes to power use - it's all about brute-force.
 
Last edited:
PC DLSS CNN model is not equivalent to Switch 2's, switch 2 is much better at low resolution, it's been compared for same resolution on PC close to transformer model. There's a reason DF was blindsided for 3 weeks thinking the games were running native and no DLSS.

TAA implementation for CDPR is full of artifacts, one of the worst implementation since Halo Infinite.

But then, you say that it has to use XeSS to get back more performance.. so why are we even discussing TAA? You can't have both.
I highly doubt it is using any sort of transformer model, the reason DF was blind sided as unlike PC DLSS that have most post-processing effects rendering at native resolution, the Switch 2 does not and renders the effects lower. Hence, making pixel count for them on trailer footage difficult. The difference in motion is apparent, the Switch 2 has none of that clarity you'd expect from DLSS 4 in motion. Just refer to the DF video review from the 04:05 mark:



The only custom DLSS model we have confirmation on is the mini/lite model that is used on Hogwarts Legacy and Fast Fusion. Apparently very fast to run, but the drawbacks in visual cutbacks are pretty clear. For what it is worth, the developer documentation only makes mention of DLSS 3 equivalent models and the lite model.

TAA is not great in Cyberpunk, but even DLSS is not going to do miracles at resolutions that low. As for the XeSS comment, I'm saying it is an option to get back more performance. You can run Cyberpunk on the Steam Deck at a native 720/800p pixel count while using the Switch 2 equivalent settings, but you will face some severe FPS drops in Dog Town. Just like the Switch 2 would. So instead of dealing with FPS drops around 20fps while driving around Dog Town you can sacrifice a bit of clarity to use XeSS and get a higher FPS. XeSS is not needed in the base game, or if you don't mind having a similar performance profile to the Switch 2. Most people would likely prefer a closer lock to 30fps though.

In the future, I actually agree the Switch 2 would likely start pulling away from the Steam Deck, as developers get use to the hardware and API, make bespoke changes to the engines running on the console, and the Switch DLSS models improve. But for now, the experience on Switch 2 and Steam Deck are extremely similar, technically a win to Steam Deck, but in regular play it will be hard to notice.
 
Last edited:
Well not really - XeSS just gets faster on them (to the point of being useable).
The thing is though - Deck (and to a lesser degree 6800U/7840U), perform rather dreadful with TAA (all of them, DF missed this bit - but even regular TAA cuts the native performance rather dramatically on these handhelds), still on the account of the same bandwidth starvation - running desktop optimised shaders. Like I've seen games lose 15-20% performance by enabling TAA (no upscaling involved at all) on these chips - to the point where even MSAA would sometimes perform better on the same game.
Meanwhile Switch - even before Switch 2 - received actual handheld optimised variants (even for FSR) that performed considerably better under the same constraints.

Anyway - the 'AI 9 HX 370' / Z2e though at least in my testing so far do a lot better (whether it's just having that much more brute force or dealing with memory contention better).
Quick question, since you mentioned testing on the Z2e, have you been able to test the Intel Core Ultra 7 258V at all? The one that is in the MSI Claw 8 Ai+? If so, I'd be curious to your thoughts as to how it stacks up against the Z2e?
 
While this is a Switch vs SteamDeck comparison, it begs the question what are we actually getting from those +250 Watt home consoles.
The return per watt is simply terrible.
 
I don't really use table top mode tbf. The idea of carrying an additional pro controller with me as I'm out and about sounds very unappealing.
Well the way I look at it, the Switch 2 is already too big to fit in your pocket, so you're going to need a bag or backpack. And the controller only takes up a little more space, so should be able to fit.
 
Well the way I look at it, the Switch 2 is already too big to fit in your pocket, so you're going to need a bag or backpack. And the controller only takes up a little more space, so should be able to fit.
It would've been great if the console had some way of attaching controllers to it
 
It plays and looks great in docked mode. I never thought I'd say this about the Switch line of products, but the ability to take it on the go might just be a perk.
 
It plays and looks great in docked mode. I never thought I'd say this about the Switch line of products, but the ability to take it on the go might just be a perk.
It's basically a portable PS4 with DLSS and Nintendo games, which is great as games didn't evolved much since last gen
 
Steam Deck is using 24.5w but Switch 2 is using 8.9w.
Damn, that's a difference. I've seen that Deck performs better at 24,5w, but the performance you get from the Switch 2 with 8,9w is real good.

Personally, I don't play anything on the deck that requires more than 15w for playable frame rate. Too loud, too warm, and battery drain is too fast.

Then again, I wouldn't want to play a game like CP2077 on any handheld. I want this on my 21:9 or TV screen with path tracing.
 
4 years into PS2 we had a portable console with 60% of its power running at 2-3W next to ps2 50, xbox 100.
No Amd involved, but Nvidia was there on the higher end.
Different times, also is not only Nvidia, is also ARM CPU architecture for the Switch 2, people were laughing at Nintendo when 8-10nm was spotted in the spec leaks, look at the results now, the 8-10nm chip blow away the 6nm in the Deck by a lot…
 
Yea seems to be europe so far

Yea the game looks like a fucking mess, any consoles, to be honest. Still somehow they managed to transfer the mess better than I thought they would. For sure there's gonna be cutbacks, it seems to be unlocked fps and very variable, seems to aim 40 fps, but it looks fucking clean ? Unexpected.. Switch 2 left, Series S right. I think it might have better parity in cutscenes, a bit less comparable in gameplay as it seems also the devs have worked on Switch 2 to tweak things around. It seems right now that on switch 2 they have fixed the global illumination bug that consoles were affected with that didn't have proper shadow. Look at the underside of the leave where the bug is in the screenshot. Its the same for when the character is underneath something overhead like a cave, the other versions are lit from everywhere while Switch 2 is dark as one would expect. Will be an interesting DF video for sure. I see they removed SSR water on switch 2. It was so bad on consoles too, not sure if a loss.

TzlcPSI9IZPsnyRb.jpg



M1CMAnwBtFXNNPhC.jpg
Finally got to play the demo... It's definitely not 30 fps, more like unstable 60 LOL. Also, IDK if this is gonna be the norm on Switch 2 but the games looks kinda pixelated, the fur in animals looks bad, like they forgot to add some kind of TAA or the TAA isn't resolving the fur properly since everything is so sharp. The game is ugly on its own anyway.
 
PC DLSS CNN model is not equivalent to Switch 2's, switch 2 is much better at low resolution, it's been compared for same resolution on PC close to transformer model. There's a reason DF was blindsided for 3 weeks thinking the games were running native and no DLSS.

TAA implementation for CDPR is full of artifacts, one of the worst implementation since Halo Infinite.

But then, you say that it has to use XeSS to get back more performance.. so why are we even discussing TAA? You can't have both.
The DLSS model of this thing is legitimately incredible just as well. I wonder what customizations they could have made, surely just a console environment isn't enough?
 
Top Bottom