We've seen this sequence a hundred times before, right? The classic Cyberpunk 2077 benchmark scene from the PC version of the game. Only we've got it running on Nintendo Switch 2. It's part of a series of internal benchmark scenes buried deep inside the game. And once they're running on PC, it's possible to use Cyberpunk's cross-platform save system to port them over to any console version. Even got them running on the PS4 and the PS4 Professional. But we're going to be using these sequences to do something different to produce our first Switch 2 versus Steam Deck comparison. Um, it's one of our most eagerly anticipated videos according to backers of the DF supporter program. But this comparison comes with a gigantic array of caveats to consider. Steam Deck runs the PC version of the game through the Proton compatibility layer. The Switch 2 port is built for Switch 2 using a direct to the metal graphics API. Secondly, the Switch 2 version has an array of custom modifications, meaning that while the games may look basically the same to the untrained eye, they most certainly are not. Thirdly, well, Switch 2 has a 1080p screen while the deck is at 800p, 720p if you stick to a 16x9 aspect ratio. And finally, the Nintendo machine has a custom form of DLSS upscaling, which as we shall discover, gives it a huge advantage over Steam Deck. For its part, the Steam Deck OLED, the model tested here, may have a lower display resolution, but its HDR implementation blows Switch 2 out of the water. And it's got a larger battery, 50W hours, compared to the 19.75W hours in the new Nintendo machine. But let's just say it's going to need that extra battery capacity. Steam Deck efficiency is nowhere near as good as Switch 2. The contest is not even close, as you'll see. Finally, we need to consider the actual point of these devices. Steam Deck makes little sense as a quote unquote docked device connected to a display. Switch 2 is a hybrid console built from the ground up to support both handheld and TV gaming. We'll find lots of commonalities in the video ahead. Then some clear points of differentiation in game performance and especially power efficiency. But a matched like forlike comparison, this most certainly won't be. Before we crack on with a head-to-head, well, to help our understanding of the Cyberpunk port and any potential comparisons with Steam Deck and by extension the PC version, we need to compare the two directly and get some kind of ballpark settings lock. Not easy when the game has custom adjustments made for the Nintendo console. To try to create a level playing field of sorts in ascertaining settings equivalents, we captured PC at 1080p on DLSS quality mode unless otherwise stated. I'll try to keep this brief, but I do want to show we've done the leg work here, or rather our valuable contributor Muhammad Rayan has. So, let's kick off. Starting with crowd density, there's clearly a thinning of the crowds as we move from left to right. By concentrating a little more on the far end PCs, we'd say that medium quality is the closest approximate match to the PC version in both docked and handheld play. Next up, an easy one. Contact shadows where the Switch version in both modes seems to have them enabled. This one being a simple onoff toggle. Volutric fog resolution. Next, this one seems to be closest to PC's medium on both docked and handheld play, as is the case for the Volutric cloud setting as well. Ambient occlusion, though, here's where we start to see some divergence as the docked quality mode seems to look closest to PC's high, while handheld performance mode presents more like the low preset. There are also differences in mirror quality, as you can see here. Let's look at docked mode first, where Switch 2 gives a good showing. It's most similar to the high setting on the PC version. Looking at the handheld equivalent, we've tried to match the lower resolution on PC, which now seems to suggest that Switch 2 is closest to the medium settings. Beyond that, things get a little more complicated because it's clear there's a lot of custom work on the Switch 2 version of Cyberpunk 2077 when it concerns aspects such as shadow quality, screen space reflections, and definitely level of detail. All of these components have no definitive match with PC or other consoles. Screen space reflections, for example, check out this example. Scaling from ultra down to low, switch 2 specifically presents most similarly to the low preset on PC, but even then there are differences. Meanwhile, in this scene, the ground looks most similar to the PC's high preset, but at the same time, the door is taking on additional reflections that are not found on the PC version at all, which is certainly interesting evidence perhaps of extra SSR ray marches. Low quality is the broadest general match for our approximate PC settings, but clearly this is not a onetoone match. Moving on to shadows. Cyberpunk 2077 has a range of settings that we tried to match, but only with limited success. Local shadow mesh quality for starters. Well, this does seem to present in a way that looks like the PC's lowquality preset. Cascaded shadow resolution. Well, based on this comparison, medium is a good fit for the docked quality mode, while low is most similar to the handheld performance mode. Distant shadow resolution based on this Vista shot switch to in both docked and handheld modes is up there with PC's high. Beyond that, things get a little trickier. Cascade shadow range. We've opted for medium as the closest PC equivalent, but the truth is there are changes. This shot suggests that switch slots between medium and low. And the same thing here in what looks like definitive evidence, but then we get an example like this where the Switch version draws in a self shadow later than PC's low setting. However, in this final shot, switch 2 is equivalent to PC's high. It's pretty tricky then. So, we've gone straight down the middle with our approximate choice and medium it is. Geometry, detail, and level of detail is also custom very much so. So looking at this example where it's unlikely that the game is under too much stress, geometry on Switch 2 handheld and docked modes can be significantly simplified up against PC's low setting. Indeed, in some areas, the base PlayStation 4 exhibits some detail that is not present in the Switch 2 edition. There's the sense that CD Project Red has strategically nipped and tucked, usually in areas much less noticeable, like here from the Kabuki streaming test run, for example, where the tubes here are less rounded than the PC version. Interestingly, just as Alex discovered way back in the original review, changing the level of detail preset on PC doesn't seem to make many changes, if any at all. Rerunning Switch 2 against PS4 in the Kabuki run demonstrates a night and day difference in favor of the Nintendo console hybrid. The game presents with swavs of geometry, textures and shadows missing. A night and day difference. But we need to remember that Cyberpunk here is streaming in from a 5,400 RPM mechanical hard drive and it's using a 1.6 GHz Jaguar CPU. benchmark sequences like this existing at all may be down to CDPR tackling the streaming challenge on last gen consoles back in the day. It's just that those streaming test runs persist in all existing versions of the game. So, here's how our overall settings match up between Switch 2 in quality mode docked and performance mode undocked. And for added value, we've got the same settings locked in for the PlayStation 5 version of the game running in performance mode. Perhaps puzzlingly, you may notice that the PS5 version has some settings like ambient occlusion and volutric fog resolution that are lower than Switch 2. And there is an explanation for this. The game is running at a far higher resolution on the Sony console, so additional resolution may make up for the quality difference. Muhammad produced these settings back in the day, stacked up against the PC version running at 4K, by the way. Either way, this gives you some idea of the high ticket price items where Switch 2 needed adaptations beyond culling resolution, but it's kind of gratifying to see that CD Project Red didn't need a barebones radically cutback approach. Returning to the internal streaming tests here against the PlayStation 5 version of the game in terms of what's actually being rendered, Switch 2 is very much in the same ballpark. Going into the Deck versus Switch 2 comparisons, one thing becomes obvious. While not quite a magic bullet, whatever the custom form of DLSS is that Switch 2 gets, it delivers major advantages to the Nintendo console. Deck doesn't have it, of course. It's got an AMD APU, meaning that FSR3 and XESS need to be used for upscaling instead. I dialed in the same dynamic resolution range as used by Switch 2, but the fact that we're losing so much performance with the upscaling is telling versus native 720p. In this test at least, we're essentially locked for 30 fps for the vast majority of the sequence by staying at native resolution 720p while performance drops when upscaling. The fact we are losing frames demonstrates that we're upscaling from lower bounds 720p. So, this represents an interesting exercise in ascertaining the upscaling cost. It is significant. anything up to seven frames pers in my tests. Calling into question whether 1080p is viable for the deck at all. Personally, I think not. But as Switch 2 is upscaling to 1080p, well, I guess we should be doing it on Deck 2. Um, even though the image quality gulf between the two handhelds is vast, DLSS is so much better than anything Steam Deck can muster. Of course, we're talking about simulating a docked experience that I doubt many deck users will be interested in, but it is part of the experience, and we'll be talking about handheld play later on, where the deck fares much better. But let's kick off by trying to compare the 30 fps docked experience, and we immediately face a problem. There are two ways to lock the deck to 30 frames pers. use Gamecop's 30 FPS cap shown on the right here and then to use Cyberpunk's own 30fps cap that's on the left. Proton doesn't seem to work in the same way as Windows here where Cyberpunk at 30 fps delivers a laggy but perfectly frame-paced 30fps. Um, we are indeed capped to the right frame rate, but frame pacing is highly inconsistent. That's not good. On the right, we have the Dex system level cap. But this exhibits another problem. While 30 FPS presents correctly with even frame pacing, 33 milliseconds per frame for a smoother experience, when there are dropped frames, they persist for an extra frame longer than they should, 66 milliseconds rather than 50 milliseconds, meaning that when the game does drop frames, it looks super choppy. I'm sticking to the game cap then, but under protest really. Not a great start for the deck, but let's crack on. A very quick look at the actual official benchmark test from the PC version. On Switch 2, it is a locked 30fps. That lock meaning that leftover GPU power can go towards increasing resolution for a cleaner image up to 1080p. We're getting 95% of Switch 2 performance on deck with FSR3, dropping to just 89% using XCSS. Now, since we're dropping frames throughout on Steam Deck, chances are we're always upscaling to 1080p from lower bounds 720p. The gap widens if we compare a similarly setup 40fps performance mode, where the DRS window expands to 540p to 1080p. Switch 2 hits the target, but deck with FSR3 has 91% of deck performance, dropping to 84% with XSS. In short, demanding more in frame rate terms sees the Switch 2 advantage open up even further. Despite trying to match settings as best we could, the early part of the Kabuki CD Project Red run here suggests a missing shadow cascade on Switch 2. But more obvious is that the performance from Switch 2 is much more consistent in hitting 30 frames pers. There's so much headroom here that not only is Switch 2 running more smoothly, but you are getting better image quality. Not just from DLSS, but also from a wider dynamic resolution range. Switch 2 renders at anything from 720p to a,08p in this sequence based on our pixel counts. While Steam Deck with both upscalers is losing so many frames that the DRS window is just 720p to 765p here in this area where we do actually hit 30 frames pers. So a much higher resolution, a 29.5 fps average capped at 30 fps. remember um that's on the Switch 2 with 27.4 FPS average from FSR3 dropping to 26.7 FPS from XSS. In 40 fps performance mode, Switch 2 finds its limits in any area with a decent level of geometry. lowest frame rates are actually marginally worse than the quality mode here. And the opening of the sequence sees Deck take the battle to switch to in what may be a more CPU limited area, bearing in mind how XSS and FSR3 offer similar results. However, the center part of the sequence sees the Nintendo console hug the 40fps line in a way that the Steam Deck does not. Pixel counts in this sequence range from 540p to 756p on Steam Deck where corresponding shots on Switch 2 give us 630p to 900p. Uh now Tom Morgan, credit to him here for doing the pixel counts. He counted 900p at this area on switch 2 up against 540p on matching FSR3 deck content. And that's in an area where we're dropping frames. kind of suggests that the DRS system only drops when GPU limited, not CPU limited, as it should be. Some games don't, believe it or not. A quick look at this racing combat benchmark sequence from CDPR's internal tests taken from the Beast in Me City Center side mission. You'll note that the gap between FSR3 and XCSS is smaller than usual. And you'll also note that the gap between Switch 2 and both deck iterations varies considerably according to content. Switch 2 remains faster and better hugging the 30fps line, but drops consistently when more enemy cars in play. In the 40 FPS performance mode, well, this is interesting. In lighter areas, Switch 2 continues to have an advantage, but long stretches of play in more detailed areas. See, all three iterations here play out with similar frame rates. Perhaps we are looking at similar CPUbased limitations here. Switch 2 does seem to run at higher resolution in these areas though. One final note, power efficiency measured from the wall. Switch 2 pulls 18 to 90 W of power. Factoring in efficiency losses from the power supply. The machine the Switch 2 likely drawing 16 to 17 W of power. Now this is actually lower than measured in our Switch 2 review. But more interesting is that Steam Deck OLED is running pretty consistently at 24 to 25 watts here according to its performance meter and it's about 29 to 31 watts from the wall. A higher figure again likely via efficiency losses from the power supply. We aren't comparing like for like here. We've got to be clear about that. We're comparing what we've decided are the closest quality settings we could master. But you've really got to take your hat off to Nintendo and Nvidia here. We have resolution and performance improvements over Steam Deck using twothirds of the power. And at the same time, it's generally believed that the 8 or 10 nanometer Samsung process used by the Switch 2's T39 chip is not as performant as the 6nanmter process used by the Steam Deck OLED. Remarkable stuff to see that gulf. All of which brings us on to handheld performance testing, arguably more relevant, bearing in mind that Steam Deck has few pretensions in offering any kind of home console experience attached to a TV. And this in turn presents a challenge, right? It's easy to capture the Steamed Decks output at either 60 or 120 Hz, as I've just demonstrated. However, Switch 2 has no video output in handheld mode. So, how do you test performance? Kerry from the Forks did things the super hard way by manually counting frames from a high-speed camera feed. So, maximum respect for that. My approach different. I stuck to the 30fps quality handheld mode and used the Switch 2's internal capture system. This gives us a 1080p capture, but only at 30 frames pers, so half temporal resolution. However, the capped frame rate of the game matches the frame rate of the capture. Think of this as a temporary placeholder solution for measuring performance. It's not as precise as I'd like, but measuring this will highlight performance drops and a decent level of accuracy. Steam Deck well put it through the same frame rate decimation and we have the same temporal resolution sampling. Beyond that, what we need to remember from these comparisons is that once again, head-to-heads, exact head-to-heads aren't possible. settings can't be matched, only approximated. We don't have DLSS on deck, nor can we set an 810p output resolution to match what switch 2 is doing in quality mode. Best we can do is to use FSR3 with a 900p output resolution and a 50 to 90% dynamic resolution range to give us the same 450p to 810p DRS window. Kicking off with CDPR's streaming run in the Kabuki area of Night City, we kick off with like for-like results with just a small level of instability on the Switch 2 side. The climax of this run sees further instability from the target frame rate. But interestingly, Deck is more stable unless collision detection registers a hit. Then the deck seems to momentarily lose its 30fps lock. Suggests a CPU advantage for the Steam Deck. But the elephant in the room here is power draw. Steam Deck OLED is still maxing out at around 24.5 watts on average based on its own internal power meters. But Switch 2 is delivering remarkably similar results, just a little less stable. And it draws around 8.9 W on average based on overall battery life. And it has DLSS of course, so image quality is cleaner. Like for like Pixel Council, that's tricky owing to the compromised nature of Nintendo's internal capture system. Even so, I'd argue in this as close to like forlike scenario as we can get for Switch 2 to be doing what it's doing using just 36% of the power used by the Steam Deck. That's astonishing. Moving on to the Beast in Me City Center test. This is a much more demanding test with far more drops under 30 frames pers. Although we can match up the runs here thanks to the CDPR benchmarks, what actually happens within them is dynamic. The more interactions with other races, the higher the likelihood of a drop beneath 30 FPS. Switch 2 caps at 30 FPS, of course, but it can drop to as low as 23 FPS. While Deck is undoubtedly smoother with fewer frame rate drops and only dipping into the late 20s. Now, switch 2, of course, still has a huge efficiency advantage, plus the image quality benefits of superior upscaling via DLSS. So, let's address the next elephant that's in close proximity, that gigantic power efficiency advantage enjoyed by Switch 2. By using the internal power metering offered by Steam Deck, I found that dropping TDP down to um 5 watts versus the standard 15 watts gives us a much closer like forlike test um up against Switch 2 in terms of powered draw. The effect on performance is devastating on Steam Deck, however, dropping frame rates in this test down to 9.2 frames pers against Switch 2 with its 23 to 30 FPS window. The effect is so devastating on the deck that it's actually running in slow motion. With a similar powered VA, Switch 2 simply streets ahead. Going back to Kabuki, average frame rate on deck rises to 13 frames pers, but of course, Switch 2 is running this content mostly locked at 30 frames pers with a big resolution advantage. thing is it's okay for Steam Deco OLED to use a lot more power because its battery is much larger, 50W hours versus Switch 2's 19.75W hours. And as demonstrated earlier, we're sapping a lot of performance from the deck by using an upscaler we don't need because Steam Deck doesn't have a 1080p screen. For the purposes of our capture, I then reran these benchmarks at native 720p, a much better fit for the deck's screen. um 800p external capture is a little bit trickier and as the larger battery accommodates the much higher powered draw pretty much I ran at full TDP 15 watts going back to the Kabuki run performance is generally smoother than Switch 2 now though there are still some blips beneath target frame rate same situation in the beast in me city center side mission test still some fluctuations under 30 frames pers but the deck is a touch more stable overall Both runs are the same or better than they were when we were using the arguably unnecessary upscaler and it's a better fit for the Dex use case. It's not a 1080p machine, plain and simple. But the fact is that in these test sequences at least native 720p is actually fine in performance terms in this test content at least. So what have we learned from this experience? Empirically, not a great deal. A PC game running on a Linux handheld through the proton translation layer without custom retooling for mobile gaming and without an iteration of the best upscaler on the market means that Switch 2 goes into this one with a range of unassalable advantages. In docked mode though, the gulf in performance and rendering resolutions is such that it's difficult to come to any conclusion other than that Switch 2's GPU is far more performant. Even though the processor is on a supposedly suboptimal fabrication technology in theory, Steam Deck should have a CPU advantage, but it's hard to tell that this is the case when you're so heavily GPU limited. In handheld play though, there's a strong case that in realworld head-to-head comparisons based on actual use cases, Steam Deck holds its own. Native 720p is a better fit for the panel. There's no need to use DRS or upscaling to achieve similar performance to the Switch 2 operating in a 450p to 810p dynamic resolution window. In areas where we kind of think we may be CPU limited, the Steam Deck also performs more smoothly in adhering to the 30fps target we opted for. And yes, of course, Steam Deck OLED's HDR is actual HDR, not something we can say about the Switch 2's effort. A lot of subjective analysis there, but a hard fact you can take away from this is the efficiency of the Switch 2 being in a world of its own. way way ahead of Steam Deck, which in itself is highly efficient compared to a great many Windows handhelds. Switch 2 is delivering handheld cyberpunk at under 9 watts total. Handheld efficiency was one of the big question marks surrounding Switch 2 before we went hands-on. But well, this is a phenomenally good result. However, the realworld impact of this is blunted somewhat. Steam Deck OLED's battery is basically two and a half times larger. So ultimately, both Deck and Switch 2 will deliver broadly similar battery life. Anyway, I hope you enjoyed the video and I also hope you enjoyed the use of CD Project Red's own internal benchmarks. Interestingly, none exist for the Phantom Liberty content to the best of our knowledge, but the benchmarks we do have do seem to run on every single console version. I mean, here we are running one of them on Switch 2, PlayStation 4, and PS4 Professional. Interesting performance data there, but do check out that level of detail difference here. I don't have too much interest in doing a deeper dive comparison into the last gen versions up against Switch 2. Tom already did that in his review without having to use these benches by comparing gameplay, which is arguably more relevant. But we'll put that data up on our DF clips companion channel at some point. But that's all for me on this one. Like, subscribe, share if you enjoyed this content. Ring bells for notifications, but please do consider backing Digital Foundry via our supporter program, highquality video downloads of everything we do, early access to a bunch of content, including DF Direct Weekly, bonus material, news updates from the team, and more. And yes, another way to back us is to check out our merchandising wares at store.digital digitalfound.net. Still, that's all for me for now. Thanks for watching and supporting Digital Foundry.