For someone owning both devices and actually trying to decide which version to get, both are decent in portable mode with the Switch 2 taking the lead in docked mode (as the Deck doesn’t increase its power limits in docked mode whatsoever). So I’d probably get the Switch 2 version if I didn’t have a desktop PC to go with my Deck, but I do, so my “docked” experience (playing on my PC) is vastly superior anyway, with the Deck getting the portable part done.

For a technical comparison it’s kind of inaccurate I think. Yes, it’s certainly impressive that the Switch 2 can run this game in portable mode likely consuming less than 10 watts for the entire system while producing okay graphics. And it’s clear that DLSS does a lot of heavy lifting here, but:

• The 8.9 watts figure is likely somewhat inaccurate because it’s based on approximate battery life while playing the game. Even if the game is played from 100% to 0%, there’s still inaccuracies because the specific battery likely won’t have 19.3 Wh exactly. Instead it’ll likely be a bit higher than that when brand new, and a bit lower with 100s of cycles.
• The Switch 2 clearly consumes less power than the Deck needs to achieve “playable” framerates in Cyberpunk 2077, but that doesn’t tell us that much about the efficiency of just the SoC. I’d assume the Deck requires a little bit more juice for its OLED screen and also more for the rest of the system, for example the standard NVMe drive it uses. The “approximately 9 watts consumption” comparison they’re doing makes it look like the Switch 2 is around 3 times as efficient, but that’s not how efficiency curves work. You’re comparing the Deck at a power consumption level that’s probably the peak of Switch 2s efficiency curve.
• Game settings are (currently?) impossible to match. Some can be matched, others are either some in-between on Switch or even “lower than low”, for example some models/geometry. I assume these changes have a large enough performance impact that CDPR thought they were worth to implement just for the Switch 2.
• Scene-specific pixel counting wasn’t really done, so it’s not possible to say which device renders more “real” pixels (even though DLSS certainly seems to make the most out of these pixels).

I still think the Switch 2 is very impressive in terms of performance in portable mode, certainly more than I expected when hearing about the rumored Ampere architecture and the Samsung manufacturing process.

It also shows that something comparable to DLSS (likely FSR 4) would be hugely beneficial to PC handhelds so I hope that the Deck 2 will properly support that. Sad that AMDs Z2 series don’t, but I hope Valve is cooking another custom chip with AMD soon.