Oh, I find Horizon to be absolutely stunning game, but I think this was more about how demanding open world games that look better than CP2077 are.

Horizon is tremendously efficient in its combo of tech and art style (and yeah, no RT at all according to devs), so maxed out on 5090 (which is some 4x PS5) @4K native it runs at 111fps.

CP2077, with everything turned on (RT Overdrive) @4K DLSS quality (so native 1440p) runs at 51fps.

Star Wars Outlaws, all maxed, RR+RTXDI @4K DLSS quality (so native 1440p) runs at 41fps.

Ah. How fast do they run without RT though? RT is a big resource hog, no idea how much CP2077 on Switch 2 uses it. Not RT Overdrive level I'm assuming :)

Horizon looks good on a base PS4, a Switch 2 would run that kind of game with ease. Cyberpunk 2077 with the Phantom Liberty DLC is likely more difficult to run, the PS4 would have problems with that if it could run it at all. 

I'd say HZFW is extremely cutting edge in bespoke tech, from the water to the clouds, its a perfect example that resources that are often poured into RT can sometimes be better be put elsewhere. But focusing on RT is easier spend of modern resources than designing bespoke tech for a single game world. And RT in Cyberpunk doesn't translate into much better visuals on console imo.

Part of that is it having a proper, dedicated port. Cyberpunk on PS4 can't relate. 

But looking better is a different question from harder to run. The notion of an impressive port lays somewhere inbetween. Witcher 3 for example is very impressive but actually quite ugly if we're being honest and there many more impressive Switch games if you wanted to showcase the hardware

I don't have data full data for 5090, but here are results for 4090 in Phantom Liberty from other source, all 4K Native:

Ultra, No RT - 69.3 fps
Ultra + RT - 36.3 fps
Ultra + PT - 19.5 fps

Those are averages.

Some areas have full breakdown:

PT: 23
RT Ultra: 39
RT Medium: 46
RT Low: 69
Ultra: 73
High: 86
Medium: 105
Low: 138

Here's a link for image comparisons, you just select on both sides what you're comparing, just be sure to look it on computer and click Maximize in top center

https://www.techpowerup.com/review/cyberpunk-2077-phantom-liberty-benchmark-test-performance-analysis/4.html

Burning Shores also skipped PS4, but doesn't mean it's not possible to scale it down to run on PS4.

A proper port of CP2077 with Phantom Liberty is possible as well, just not profitable anymore at this point in time.

Phantom Liberty runs on Steam Deck at 30 fps

Digital Foundry did a comparison a while back, and yes, the Steam Deck basically renders at the same fidelity and frame rate you could expect out of a PS4 title, but at 800p, not 1080p. It is underpowered when directly compared to a PS4.

PS4 is more capable so shouldn't have issues with Phantom Liberty.

Oh, no doubt about it, it's marvelous combo of smart tech and art style that produced something that really stands out, looks better than many (if not most) RT titles and doesn't require $2000 GPU to run in 4K.

CP2077 is mixed bag when it comes to RT and PT, some things look really good, some things not so much.

RT is no doubt inevitable future, that future is just not quite here yet, but with more and more games coming out that are requiring RT hardware to run at all, it is quite clear where we're heading. I'm expecting next gen console hardware to be much more oriented toward RT heavy workloads.

On the conversation of whether or not the Switch 2 Handheld mode can be that power-efficient. 

Here is a table I compiled of Ampere chips. I added different 5W, 6W, and 7W scenario's for the Switch 2 handheld and Switch 2 Docked (15W.) 

First of all, the TSMC 7nm Ampere chips are nowhere near the most power-efficient. Take the A30X for example. It is way down the list, despite being on that node. This illustrates how much the specific GPU/SOC configuration matters and how process node, while mattering, isn't the only factor. 

Secondly, notice how much more efficient the lower TGP chips are. In fact I added a plot showing how lower power consumption for Ampere chips correlates with higher efficiency (excluding the hypothetical Switch 2 efficiencies of course.) 

Conclusion: Given this correlation between power-consumption and power-efficiency, I don't think the Switch 2 Handheld mode being at the top of the list in efficiency for Ampere is unreasonable, and it isn't too far from other Ampere chips (like the A5500 Max-Q, for example.) 

Switch 2 vs RTX 5090

https://www.youtube.com/watch?v=1Qgpe6fl7os