I watched that video and it appears to be an inferior fabrication process to Samsung 8Nm with a higher level of 10Nm fabrication within the chip. Also that video seems to be simulating the Switch 2 in portable mode using far higher wattage than a real Switch 2 can use. Again it only has a 20Wh battery (3.7Vx5.2A) with a minimum runtime of 2 hours in portable mode so only gets about 5W on average for the SOC with 5W for the screen per hour. That video seems to show portable mode using far higher wattage than that. I'm sure the docked mode figures are fine but doesn't seem to have calculated for battery capacity and runtime in portable mode.

It does feel like docked mode is right though. Docked mode is about 2 Teraflops but they are claiming more like 1.2 Teraflops for portable mode which seems an impossibility for a 8/10Nm fabrication process and only 5W per hour. I still feel real performance in portable mode is between 600-800Gflops to get to that 2 hours minimum run time. I did claim docked performance before this video would be between 2 and 2.4 Teraflops and it's at the lower end of that expectation not the high end.

Surely using such a fabrication process means they will need to downclock severely to get 2 hours portable runtime?

No one here seems to be addressing the battery issue of the Switch 2 it just gets constantly ignored. If I've got it completely wrong tell me but I'm really curious about this. I would say with games like Cyberpunk dropping to 360p native rendering surely that is something required to get 2 hours out of the Switch 2. Also 5W is the absolute maximum it can use on average, the Switch 2 will need to use far less for games that have longer runtime than 2 hours.

The original Switch used about 4W on average in portable mode and it looks like the Switch 2 is very similar, hardly surprising but people are writing like the Switch 2 can use much more than this in portable mode but how is this powered?

It really comes down to whether or not the leaked clocks are correct at this point. 

Docked clocks seem to be what we'd expect given that it is indeed 8N (8LPP/10LPE hybrid.) I am leaning toward the TDP being somewhat higher than Switch 1, though. 

The handheld clocks seem ambitious, but Nvidia could've possibly improved low-power efficiency even without a die-shrink. If it were just a cut T234 (reduced core counts and down-clocked, without changing the actual chip), we'd expect a very low clock-rate given what we know about Orin T234, but more was changed than that and we have no idea what the power curve looks like now. 2 hours battery life @561Mhz doesn't seem unreasonable if they were to get more power-efficiency at lower clocks than T234.  

I was wrong about the 16WH for the Switch 2 that I posted in an earlier post. Apparently it is 19.3Wh.* 

That means the Switch 2 can average around 9.7W total power in handheld mode to last exactly 2 hours. Moderately more achievable than the original 8W I was thinking. 

*
https://www.eurogamer.net/digitalfoundry-2025-switch-2s-full-reveal-analysed-how-powerful-is-nintendos-new-hardware#:~:text=However%2C%20the%20new%20console's%20battery,handheld%20will%20consume%2010%20watts.

"However, the new console's battery capacity translates to around 19.3 Wh up against an equivalent 40 Wh in the original Steam Deck and 50 Wh in the OLED model. With Nintendo promising a minimum two hours of battery life for Switch 2, that means that all functions of the handheld will consume 10 watts."

Edit: 

By the way, 1007 Mhz in docked mode corresponds to 3.094 Tflops, and 561 Mhz in handheld mode to 1.72 Tflops. Not that it tells us much in itself, other than the relative performance of the docked mode vs. handheld mode and a rough comparison to other Ampere chips. Could use heuristics like estimated flops:raster performance ratios to compare to other architectures, but that gets a bit too rough. Nintendo/Nvidia could probably have gotten up to 4 Tflops docked/2.2 Tflops handheld if they went with a Lovelace chip. 

Yeah, 3050 6GB mobile and 4050 6GB mobile have identical configurations (2560:80:32:80:20), but former is 8N and latter is 4N. When limited to same wattage, 4050 is on average 30% faster, with around 30% higher clock, which would translate to that ~4TFlops.

They're just being Nintendo and going with cheaper version for higher profit and keeping 4N for OLED refresh, which will give them better battery life with the same battery.

I was using the performance figures from that video regarding Gflops and also don't forget the screen will take half the power. However we are both agreed about less than 10W in total per hour. I guess at this point it is who makes the best guess of how much power the larger screen with a 120Hz mode takes. I feel like half that 10W would be fair. Obviously there are other minor devices taking power like the sound amplifier, wifi etc. I'm still unsure if the board featured in the video was a development board or retail though or maybe they are the same. I don't believe this is the conclusive video on spec for sure and would prefer a known retail model to be examined. However I think we are moving much nearer to a realistic spec now. I certainly was not expecting the chip to end up having mainly 10Nm fabrication surely a severe cost cutting measure. As always Nintendo keeps their hardware interesting even if its for the wrong reasons. I still think Nintendo hardware and software pricing is unfair but have learnt from past experiences where it was me that was unrealistic about sales. I predicted the wii would not sell well just being a regurgitated Gamecube with a new controller system but how wrong I was. I've had more luck with predicting hardware performance than sales figures. However the world is a really different place at the moment with disposable income reducing in many countries. As ever if you are interested in the gaming industry its interesting to watch how Nintendo fares and reacts to the challenges of getting a new console to succeed.

https://www.videogameschronicle.com/news/hogwarts-legacy-switch-2-resolution-confirmed-along-with-dlss-support-and-other-features/#google_vignette

"According to a Japanese press release (via 4gamer ) by Sega , which is publishing the Switch 2 version of the game in Japan, the game will render at 1440p resolution while docked, and at the screen’s full 1080p resolution when in handheld mode. Both modes will also support HDR."

Almost certainly upscaled.

There is also a possible leak that Civilization VII will have a 4k 30 fps mode.

Edit: Lol I missed the thread on the front page.

Is the below tweet true? That in handheld mode it is 1/5 the power of a Series S. Not that I was expecting it to be close to the Series... just not that distance from it. 

Basically none of this is likely true, other than arguably the CPU comparison, which does roughly correspond to 1/3rd of the other 9th Gen systems. 

The handheld GPU mode would have to be clocked at 260Mhz to be a "0.8 TFLOPs system." That would be a lower handheld clock than Switch (not that this can't happen, but it is unlikely), and much lower than the minimum clock-rate of Ampere chips in an active power state. That's almost the idle clock rate of consumer Ampere chips (210Mhz.) 

With the leaked clocks (which multiple insiders seem to be quite confident are real) we're looking at 1.71 TFLOPs in handheld mode, and 3.1 TFLOPs in docked mode. 

"1/5th the GPU compute" of the Series S is laughably wrong. But would make sense if one is assuming it is a "0.8 TFLOPS system."

Handheld mode is more like 1/3rd of a Series S if we are generous and accept the 3DMark benchmark test. Generous, because the Series S doesn't have infinity cache nor the memory capacity of the RX 6600 (even if down-clocked), being used as its proxy, the RTX 2050 being used as the Switch 2's proxy has memory capacity bottleneck issues that can only be partially alleviated by system memory, and feature advantages like DLSS are not being considered. But yeah, 1/3rd of the Series S in raw GPU compute should be the rough estimate, with docked mode being something like 3/5ths of a Series S. DLSS can cut into some of that, as we are seeing with cross-generation games allowing handheld mode to achieve an upscaled 1080p-equivalent and docked mode 1080p (or higher) - equivalents. The games speak for themselves. 

The Switch 2 might struggle with GTA 6, but I don't think it is an entirely impossible port either. I anticipate that the game will run pretty decently (at low settings) on current performant PC APUs, and the Switch 2 handheld mode isn't too far off from those performance-wise. Probably more possible than the Witcher 3 was on Switch. 

I have no idea what you said but saying GTA6 will run, have you seen games on PS5 lately, we have good reason to doubt that GTA6 will run on base PS5/XSX never mind Series S or Switch 2. 

...unless it's true that the PS5 isn't being taken advantage of and devs are being lazy which I wouldn't know much about, I just know what I see and experience and this gen is cursed so far apart from Horizon Forbidden West and Demons Souls, both launch titles and one a cross gen title looking the best 5 years in. 

No we don't. The trailer we saw the other day was on a base PS5. It ran at a consistent looking 30fps at a resolution 80% of 1440p. The Switch 2 handheld is about 1/6th - 1/7th a PS5 in GPU-compute. 

If the game isn't CPU-bottlenecked, 540p (1/6th of the PS5's resolution) internal upscaled to something like 1080p using DLSS isn't impossible. If it is CPU-bound, then there would have to be some rework done to move some CPU-compute features to the GPU, where possible. 

Comparatively, a lot of work had to be done to get The Witcher 3 on Switch that I don't think would need to be done with GTA 6. 

I also didn't say it "will run" just that it isn't impossible. 

You seem to be focusing on what the chip can potentially do rather than what the battery system can provide for 2 hours. Yes we can all understand the potential for the chip to be more powerful but the reality is for portable use it has to be be powered by a battery. 0.8 Teraflops is surely all it is going to be capable of with a 10Nm/8Nm fabricated chipset. I'm surprised that video had docked performance so low but I guess that means they can have it cooler running, more reliable and cheaper power components. My original guess was between 600 and 800 Gflops in portable mode so my guess for this was at the higher end where as for docked mode by guess was 2-2.4 Teraflops and this is at the lower end. Ultimately my guess was based on how the original Switch was downclocked compared to the underlying chipset in the Nvidia Shield. There is nothing new here in how Nintendo downclocks hardware and chooses the cheapest manufacturing options.

800 Gflops is still a huge upgrade on the original Switch which was more like 120-170 Gflops in portable mode. That is definitely a generational leap and with DLSS on top we are looking at that pixel count boost that Nvidia has claimed of 10x.

Who would even want a portable system that lets say can only be powered for an hour before recharging which is what higher gflops figures would have meant. Surely as a practical gaming piece of hardware we want at least 2 hours which over time will shrink a bit anyway. Lets not forget higher current drain shorten's battery life too. A laptop with a 30W current draw will have a much shorter lifespan for the battery unit in general than a Celeron laptop that only draws 12W max. This could be part of Nintendo's plans to restrict current for reliability and lithium battery lifespan benefits. My ryzen laptop needs a replacement battery but my Celeron laptop gives me up to 18hrs runtime and despite being much older the battery pack is still in perfect health with pretty much full capacity because the chipset only sips energy from the battery pack. I can use that Celeron laptop all day without issue. The Ryzen laptop at best was under 2 hours for gaming and went under 1 hour before the battery pack finally failed. The Ryzen laptop has a GPU of about 1.6 Teraflops (3500U), the Celeron laptop about 120 Gflops (Celeron N4120). The Ryzen laptop sounds like a jumbo jet taking off when gaming, the Celeron doesn't even have a fan just a heatsink. There are great benefits to being lower powered as well as disadvantages. Nintendo's sweetspot in performance vs battery life maybe a better choice overall than just focusing on performance for a short period of time. Yes of  course they could have had more performance or more battery runtime if they had coughed up for a 5Nm fabrication process but this is Nintendo and they went as cheap as possible.