Switch 2 Leaked CPU/GPU Clock Speeds - Are They Accurate?

[@skieth9999]

The CPU being lower frequency in docked mode makes me think that they literally just turn off some cores in portable mode but clock the remaining ones higher to save power overall while maintaining a certain performance expectation. Maybe running all 6 cores takes more power but to maintain the thermal profile within tolerance, they under clock the cores collectively

[@TakisEnjoyer]

I think leaker just accidentally swapped them

[@redcrimson1028]

That's a very good hypothesis my friend.

[@AndrewTSq]

on my nvidia jetson, you can set the wattage and depending on wattage, it will turn off cores... so highest watts = all cores

[@Eepy-Rose]

Either this or it's just an error.

[@err4071]

Makes sense.

[@jptellsyou]

In docked mode I think the CPU clock was meant to be 1998.4 Mhz.

[@itsuadman]

that'd be my guess I cant see less cpu speed when docked to power

[@cheerydownylox56]

It wouldn’t make sense to have a large change between handheld and docked cpu performance however

[@PolitiClank]

There are at least two other probable typos (1534 vs 1536 CUDA cores and 2 RT cores vs 12) so I feel that is probably accurate. Still weird that they are different from each, but I feel that is the most logical explanation.

[@kjellbeats]

​@@cheerydownylox56well there's no point in holding back performance when docked

[@mimimimeow]

@@cheerydownylox56 entirely possible that Docked performance is forced to 30fps. In this case a CPU freq reduction makes sense.

[@NotoriousRKB]

When the switch 1 came out not all the clocks where available. They later on added higher clocks for the devs to use if they wanted. So maybe they will have higher clocks available in the future for docked mode

[@whoever_81]

Good hopium!

[@Ghostsonplanets]

They didn't. The only profile added was the loading speed profile

[@ibrohiem]

They down clocked the original switch to save battery. They never clocked it up to spec clocks it's still way below what it can do

[@darickfoxo7986]

Cope 🤣🤣🤣🤣

[@xtr.7662]

They literally didnt the leaked clocks were exactly the ones available for devs

[@TheBoltMaster456]

Because Switch 2 games aren't out yet, I suspect that this is probably clock speeds from running a Switch 1 game upscaled to a higher resolution and running at a higher framerate, hence the higher memory clock speeds and higher GPU clock speeds but not as big of a gap between handheld and docked in terms of GPU, and the much lower CPU clock speeds.

[@epicon6]

Except Nintendo’s business model is making the weakest possible hardware for large profit margins. Performance is literally the last priority for them.

[@Gaea1]

​@@epicon6, GameCube was the last time they cared about performance

[@SturmAH]

I wonder if there will be a CPU “boost mode” for loading times, like Switch 1.

[@StiffAftermath]

Lower cpu clock speeds while docked could make sense if the load is spread around some chips in the dock itself. A co-processor in the dock, like the og Switch patent talked about.

[@smidlee7747]

I saw on gamepost the cpu specs were reversed. the Cpu was slightly higher in Dock mode.

[@StiffAftermath]

@smidlee7747 well, could be so. These specs are not official anyways. We will just have to wait..

[@jesusmckraken]

He's the most knowledgeable tortoise on the Internet

[@cyrus8886]

The oogway of technology

[@ravekiller.]

That made my day 🤣💀

[@JaimeIsJaime]

nah, that's vedal987 🐢

[@anmolagrawal5358]

Ahaha

[@bes12000]

Docked mode using a way higher GPU clock speed and so is the ram, more power is going to the GPU in docked mode ..seems like a power limitation to me, lower the speed on the CPU so it uses less power while putting it all into the GPU and ram. Maybe their test show that it's still getting higher FPS in games with that setup. Also thermal issue too I think, they all probably share the same heat sink.

[@JackFrost-1856]

This is just senseless, its a mere 100 MHz difference. The adapter is rated for 60W so it makes no sense for the clock speed to be lower and it won't be in the end retail unit.

[@bes12000]

@@JackFrost-1856 How is it senseless?, ram speed is a lot higher in docked mode also as well as a GPU speed boost, lowering the CPU clock by 100mhz for power reasons and possibly dropping some cores to compensate makes perfect sense.. as it wont make the FPS drop, since your giving more power to ram and GPU. We will find out later anyway when official number release.

[@JackFrost-1856]

@ I already stated the adapter (when docked) is rated for 60W. That is MORE than enough headroom for a modest speed such as 1 GHz CPU clock to be feasible among boosts to other parts of the SoC. The system already clocked so low (as per leaked specs) so why would it need to be further underclocked when docked? And 100 MHz is a very small difference so the power excuse makes no sense. You're telling me the R&D engineers did not catch this flaw? Lastly, the Switch 1 CPU was the same clock speed in both handheld and dock. Why would there be a difference here?

[@EversonBernardes]

@@JackFrost-1856 Adapter is rated for 60w from the wall, but is only passing 42w to the Switch. And that maximum power draw i.e. for the whole system, including charging battery, joycons, radios and a safety margin. Docked Switch 2 is likely drawing less than 35w total. 2050 Mobile (same process and arch, bit wider) is a 30-45w part depending on how conservatively clocked it is. That's just the GPU.

[@JackFrost-1856]

@@EversonBernardes Correct, and a 1 GHz CPU clock when docked should more than feasible at a 35W profile, no? The Nvidia Orin NX has a max range of 2 GHz (though that is the A78AE and not A78C processor), and it is also rated for a max of 40W

[@DevMeloy]

My guess is they're disabling a core to and bump in the clock speed for battery life and heat dissipation reason.

[@jamesfranko5098]

When ive messed around with clock speeds on my laptop it uses less energy if you lower the clock speed overall than if you disable cores.

[@DevMeloy]

@@jamesfranko5098 Is that under load? I'm sure there is some pretty complex math involved and would be very application specific.. I'm sure the engineers have a good reason for having a higher clock speed while in handheld mode and this is just my guess as to why the Switch is doing this. Really this leak could just be wrong :) Tnx for sharing!

[@carl8790]

Since it's a ARM mobile chipset, i wouldn't be surprised if it had a mixture of high to low performance cores, where as in portable mode it prioritize the low performance CPU cores, but boost it at a higher clocked speed. I'm speculating here ofc, but I agree with you that not all of the CPU cores will be on all time on portable mode.

[@MrJojofoto]

720 P for the size is still great, and agree with John Linermann. More frames and performance for 720 P as well. People don't know, but 720 P at 8 inches is way sharper than a 4K 55 inch TV...

[@easyygo3008]

Is 540p native. Will reescalate to 1080p portable

[@MrJojofoto]

@@easyygo3008 540 P on a 1080 screen P no, not native at all, but doubled through upscale could be decent though if done right... We are talking no more than an 8 inches screen!

[@skycloud4802]

Having really enjoyed the Steam Deck and bearing in pc games these days are less optimised for small lower res screens, I'd say 720p is more than enough for Nintendo Switch 2. I had a thoroughly good time with the Deck, so I know Switch 2 will feel awesome if it manages to hit a similar feel in graphics fidelity.

[@LegacyRap7777]

Keep telling yourself that, y'all keep excusing mediocrity on the hardware end from Nintendo

[@MrJojofoto]

@@skycloud4802 Can understand that point, and hope some Zelda games can hit 60 fps!

[@AmazonicoIndio]

I love how the crew is getting more confident in these values because of how underwhelming they sound.😂😂😂 Never change, Nintendo.

[@ekinteko]

I wouldn't mind this performance in 2018, but in 2025 it does sting quite a bit. And at USD $400 and up, really doesn't feel too good.

[@raymax_1071]

Did you genuinely expect the new Switch to have the performance of a PS5/XSX? Nintendo has been a "generation" behind for a while already.

[@ekinteko]

@raymax_1071 Nah, I expected Nintendo to be consumer-hostile by releasing a console at a higher price with lower capabilities. I merely held out hope that they would surprise us, the consumers, with something that was more powerful and value for money. If they bumped the resolution to 1080p, maybe upped to 1TB storage, a usable 16GB storage, and a modern MediaTek Dimensity 9400, an off-the-shelf chipset from ARM. I think it could have been cheaper and a good upgrade for the users. Instead, it looks like we're very limited by 256GB storage, stuck with inadequate 12GB RAM, sticking with 720p for another generation. Most of the money is going into Nvidias pocket. That Tegra Orion chipset inside is a 2018-esque technology using RTX-2000 technology on Samsung-8nm silicon. So we should be looking at XB1 level of performance, perhaps with many optimisations punching above the PS4 level. But it's a far cry from the XSS, let alone the XB1X or PS4 Pro. Again, I don't think performance is everything. But it is important. And I think the paying customer, for their money, should get something a lot nicer than this. I hope that I am wrong, and will get pleasantly surprised..... but all the leaks turned out to be accurate. And all the biased, shills, and fanboys that exclaimed PS4 Pro level of gaming seemed to be wrong. For those that like the system, more power to you. I just wished Nintendo was more consumer-friendly they are a big corporation with lots of money, after all.

[@JackFrost-1856]

​@@ekinteko??? 2018 what. The Cortex A78C processor was launched in 2020. RTX 2xxx series is Turing, not Ampere which is 3xxx and launched in 2020 - and that is what the Switch 2 is based on. 12 GB RAM is more than what the Series S has, which is more than enough since this is not a PC (due to a unified memory architecture) We don't know what the screen resolution is but it is rumored to be 1080p, not 720p. Overall, this is a 10x increase in power over the previous Switch.

[@ekinteko]

@@JackFrost-1856 That gives me more hope and promise. Someone else said they won't be using Samsung-8nm node and instead opt for a better (TSMC-7nm level) Samsung-5nm silicon. I hope it is better. Based on the previous comments and leaks, it put the performance around XB1 level, maybe slightly below. If what the other comments say are correct, it would bring that higher, perhaps slightly above PS4 level. Nintendo is probably going for the easiest build (game ready drivers from Nvidia) and maximum profit margins, so we're probably not going to get anything ambitious like XSS level, let alone PS4 Pro level which is considered really slow by tomorrow's handheld standard. But I will reserve my final judgement until I can get my hands on the hardware.

[@JackFrost-1856]

@@ekinteko I'm expecting 7 nm (based on die pic leaks) I don't think the node matters that much at this point.

[@karma9471]

Awesome stuff but I really wish you would have put the Switch 1 speeds on the screen next to the Switch 2 just so we can see a comparison...

[@cyrus8886]

307 portable 768 docked

[@DragonOfTheMortalKombat]

Doesn't matter, different architecture, different IPC.

[@VidzForDayz]

I would just say that it's thermal based, they're pushing the GPU for a higher pixel density display (obvious) but due to that they need to save energy (heat) by dropping the CPU clocks. Seems pretty sensible to me 🤷‍♂️

[@aceofspades001]

Sensible my ass 😂

[@theredhead42]

I think the lower cpu clock is for cooling. With the GPU maxed out, there only isn't enough cooling to cool both chios at full throttle

[@OllieOtterOxen]

You won't save in the way of heat much from 100mhz

[@DrJones20]

Why is the cpu slower in docked mode? That doesn't seem to make any sense

[@bearwynn]

they gave you a potential answer. This is standard behaviour that you see with multi core PC cpus like ryzen. If you are using more cores, the clock speeds tend to drop. It is likely that the portable mode disables some cores for power savings, and is able to boost the clocks up sligthtly to gain some performance back while still using less power than if it had extra cores enabled.

[@JackFrost-1856]

These speeds were datamined from an SDK so it could be that the CPU clock speed when docked wasn't updated to match the handheld. There is zero reason for it to run a mere 100 MHz slower especially when CPU is supposed to maintain parity between both modes

[@alperen1383]

Maybe because its an apu and gpu using way more power limits the cpu a little? Just guessing

[@easyygo3008]

​@@JackFrost-1856😂😂😂😂😂 The mental gymnastics by fanboys are real.

[@wodensthrone5215]

​@bearwynn this is a custom Nvidia set, we don't know what it can do, its not something available on the market

[@currywurst2434]

In docked the GPU is the exact downclock they gave the Switch 1 from Max Tegra X1 gpu clocks (1ghz to 768 or so). Nvidia says Max Clock of Orin is 1,3ghz for the bigger Orin models, so around 1ghz for the GPU is exactly what i expected. I did expect a lower clock from mobile mode though as Orin in 8nm is quite energy hungry. 561mhz isn't bad for 8nm.

[@ehrensto]

If it can run multiplatform titles at 720p 60fps handheld, i'm sold already. But we shall see.

[@EversonBernardes]

I'd temper any expectations of it running current gen games at 60.

[@DragonOfTheMortalKombat]

Lol with less performance than a 2050, it ain't happening lmfao

[@xtr.7662]

The only games that will are doom 2016 and eternal thats about it

[@MrBl3ki]

Depends on the multiplatform title, obviously. Demanding current gen (XS/PS5) games will probably run at 540p (upscaled to 1080p) at 30FPS (at least target it).

[@LuminairPrime]

I don't see why it doesn't make sense. They fit a given cooling capacity into the case and set the power usage accordingly. The clock speeds are just the output of the equation.

[@zcizzorhandz5567]

I think I can explain (Assumption) it's about balance: - They get the desired performance by almost doubling the GPU, but now overall heat is increased - The heat of the CPU at base is too hot for the increased GPU perf - So in reality they probably wanted to clock the GPU to 920Mhz but they figured out by reducing the CPU and increasing the GPU to 1007.25 they have the perfect balance of performance and heat dispersion - Doubling the CPU has no benefit to graphical performance, the CPU is mostly running game logic, the PS5 Pro and PS5 clock the same speed CPUs. It could also imply that upping 1100.8Mhz could be more efficient for reducing the GPU power consumption. Assume that all games are designed to run at 998.4Mhz cpu.

[@Nobody-sp7ug]

Switch 1 never had any "stock clocks" for the gpu. It has 3 performance profiles. I dont think a single clock speed on a DSU era arm cpu means much, as it can have seperate voltage domains within a cluster, as opposed to a75? And before era where you needed a seperate cluster to do this. If I had to guess, there would be 3 clocks in the cpu. 1 core with the highest clock for the main workload, 1 core with the lowest clock for os, and the rest for supplemental cores that should never really get stressed and should remain pretty low.

[@SaccoBelmonte]

The CPU clocks looks like they simply don't change and that was margin of error. The CPU usage should probably be the same on both modes since the game logic doesn't change between modes (same number of players, NPCs and other logic). Unless the numbers are simply wrong.

[@MarcoBayod_MB]

As an owner of a GB, GBA Sp, DSi, 3DS, Wii and a Wii U, I don't have too much faith in Nintendo hardware, I didn't buy the Switch because I didn't believe it was going to triumph, and when it was a success I lost my good job and wasn't able to ride the train with the rest of the world... I really hope they are cooking with the mouse support, I don't play LOL but imagine having that game there and plenty of other PC mouse exclusive games... It could be big

[@gewdvibes]

I just don’t see how they could realistically put gaming quality mouse sensors into BOTH joycons without the joycons being like $100+

[@pothitoskourtis]

@@gewdvibes explain gaming quality mouse sensors...let me guess you buy razer products? lol

[@marverickbin]

John states that 1080p is too much because he is assuming DLSS will be used only on dock mode, right? Why not use DLSS in handheld? Target 540p rendering when handheld, DLSS it to 1080p While docked, target 720p and DLSS to 1440p Or even 1080p to 4k in less demanding games. Seems that rendering at low resolution + DLSS just spend less energy than rendering in a higher res, and could be the choice for handheld. If so, 1080p makes sense.

[@tomianz-1242]

I wonder what's best ... 720 native or 1080p through dlss... 🤔

[@cyrus8886]

I'll take 720p native over 1080dlss

[@MrBl3ki]

Depends if the Tensor cores are gonna be enabled in the handheld mode.

[@Kniffel101]

I'd 100% take native res on a 720p OLED, maybe with some AA applied, over upscaled 540p to 1080p.

[@gewdvibes]

720p native would still be less demanding and look better than 540p to 1080 upscaling, DLSS isn’t free

[@thet0ast3r]

Interresting that noone comes up with the suggestion that the lower clocks are at the cpu can match a certain memory latency in cycles? Like that the cpu clock is as similar as possible to the handheld mode, but still matching a multiplier of the memory clock? (which is radically different)

[@peebow1000]

The lower docked CPU clock is probably because yes, they're running more cores. Taking a fun wild stab here, but specifically the GPU's DLSS tensor cores. Maybe they're only bothering with the AI upscaler to get things onto 4k screens. I'm also going to suggest about the tensor cores specifically, because, disabling full CPU cores between dock modes would create an awkwardly jarred processing environment for a game to run on I'm also making an assumption, that higher frame counts via machine learning, is less taxing on the CPU cores, as they don't have to cell draw each AI frame, so they're just saving heat on unuseable gained CPU resource.

[@singleplayer2280]

The cpu speeds are probably a typo, no way it's slower on the dock mode

[@arch1107]

it could be a typo, but since the gpu uses double the speed i bet it uses double the power and double the heat, docked and in 1080p or 4k the cpu matters less so it could be just that

[@EversonBernardes]

@@arch1107 Double the GPU frequency is probably more than double the power dissipation, the relationship is rarely linear.

[@arch1107]

@@EversonBernardes it might or might not be true here, it is clear they did some tricks to keep power consumption down so something else has to be affected, they still depend on a heatsink and a fan it is still guaranteed to generate more heat, so it still falls in what i imagine

[@brettsuddreth5171]

Haven’t been more excited for a console in a long time

[@bruceinmyheart]

The original is 768 MHz in GPU speeds in docked, isn’t it?

[@DrJones20]

Yes

[@mimimimeow]

The chip itself was rated for 921 MHz, it was limited by board power.

[@cyrus8886]

768 Nintendo allows You can overclock it

[@bruceinmyheart]

Very incremental upgrade; the original must be 7 years old.

[@JoonKimDMD]

Dang not much upgrade

[@MiChael92617]

I think they should have an option in the settings were you can change the performance. For some people they don't care about battery life and want the max performance and for other people they do care, I feel like we should be given that option.

[@xxnike0629xx]

I think these rumored specs seem odd but I could see it being the actual case.

[@Dorfdad]

Is it possible that in portable mode half the cpus are used at a higher clock to save battery, and in docked all cpus are used at a lower clock which would result in more performance overall?

[@arch1107]

i think it is more related to power and heat, all focus on the gpu when docked, so more heat and more power there, higher resolution, cpu matters less

[@JackFrost-1856]

Nope. Games need to run at same clocks in both modes for logic parity. That is the reason why the Switch 1 CPU was the same speed in both modes, so this won't be any different

[@arch1107]

@@JackFrost-1856 iirc when people midified switch 1 they overclocked cpu, gpu and ram and it ran fine, so no idea what you are referring to there

[@JackFrost-1856]

@@arch1107 those overclocks apply to both handheld and dock profiles as they remain the same across both modes. It doesn't invalidate what I said.

[@arch1107]

@ i think you lost focus of what i originaly commented so

[@endame90]

Simple math made by people who know about density and stuff have concluded it's 5nm because the original chip is much bigger, 5nm is 70% more dense, and the reduction of size of the chip within the Switch2 matches with that, only these people keep sayin 8nm

[@Loundsify]

I hope you're right.

[@Niko25467]

Lol

[@johndavid5424]

The Switch sells not based on its hardware, but on the games it has.

[@cyrus8886]

These specs addicts can't read

[@michaelmayers3622]

that is simply not true , maybe in the past but not today

[@LegacyRap7777]

That's getting old bud, in this time & age of powerful hardware and incredible PC handhelds that can emulate Nintendo games, the switch 2 can have a repeat failure of a launch like the Wii U

[@cyrus8886]

@@michaelmayers3622 it does People like the concept and the Nintendo games You're average consumer couldn't care less about specs

[@Michel-r6m]

The hardware may pull in third party developers, so it does matter.

[@One-ct3xe]

It's almost like someone leaked the power levels of a Switch 2 demonstrating performance sliders, lowered stats for more battery and more stable frame rates instead of its maximum ability

[@Pfizenmaier]

John looking so worried and handsome at the same time

[@syferz]

The portable clock was upticked just before firmware 20.0.0 which leads me to believe the docked cpu mode would increase to 1.1GHz

[@MrTomemac]

Wouldn't this just indicate they are pushing the GPU to do more of the heavy lifting in docked? Then you bump the CPU up a bit in handheld to offset the cut to the GPU performance?

[@JackFrost-1856]

There is not reason to "offset" anything. 1 GHz is a very modest speed and there needs to be CPU parity between handheld and dock for games to keep things smooth. A simple GPU boost won't solve anything if the CPU is dropping.

[@urazoktay7940]

Amazing video, thank you Digital Foundry.

[@gerowen]

The CPU running "slower" when docked is a bit odd. Could it be a power delivery/budget issue, or maybe thermals, where they're willing to knock 100 MHz off the CPU in order to nearly double the GPU clock?

[@JackMeyHoffer562]

What is the network hardware?

[@mattvmani]

My assumption is what’s missing here is the tensor/AI cores to make up the difference and need the power/cooling that has not been divulged yet.

[@Sad7Piano]

Assuming the leaked frequency is accurate, could we also assume that this represents the performance limit of the NS2 with a single fan? The NS1 could enter TV mode using a 15V/2.6~3A PD charging protocol, which led to the emergence of many third-party portable docks. Similarly, the NS2 will likely encounter similar conditions, such as being placed in fanless docks while running in TV mode. Even if third-party docks include fans, they may not guarantee the same performance as the official ones. In the end, the cooling would rely solely on the console's single fan. A bold assumption would be that the NS2 fan-equipped dock may feature special interactions to unlock higher frequencies. Does this reasoning make sense?

[@LambdaWave84]

What drives Nintendo technical choices : minimum heat dissipation in a such small size factor, low power comsumption for maximum battery life and low production cost. You can't fit a power hungry AMD Ryzen SOC like the Steam deck inside this.

[@bobbysworldofficial]

lmfao only reason you posted this was to hate on AMD

[@LawlessNate]

"a power hungry AMD Ryzen SOC like the Steam deck" The Steam Deck's APU is pretty efficient when it comes to power. It has surprisingly good performance at 11 watts, and still gets impressive performance per watt below that.

[@DragonOfTheMortalKombat]

Power hungry ? Steam ain't power kiddo, it works great at 15w and so do almost all other handhelds. Curb your nintendog fanboyism

[@LambdaWave84]

@@DragonOfTheMortalKombat 15W is enormous for such a small form factor. On a small 22Wh battery, it's less than one hour of gaming (without accounting display, etc)

[@LambdaWave84]

@@bobbysworldofficial I've a AMD Ryzen in my PC, so not really dude. I try to understand Nintendo choices as an industrial choice for producing 50 millions consoles at a contained pricepoint.

[@radiod78]

Do we think there’s any extra tech in the actual dock this time ?

[@OllieOtterOxen]

No chance, usb-c even with its latest revision is too slow without causing a massive bottleneck for there to be an extra GPU in the dock

[@Quake2815]

I hope so.

[@bigdeagle1331]

Nintendo is going to face much more fierce competition this go around!

[@indexclone8130]

What if its just a power restriciton? More on gpu, less cpu budget for TDP?

[@reggiefurlow1]

" since we didn't leak it, we gonna talk about it" y'all funny lol

[@AndrewTSq]

The gpu clock is close to Nviida Jetson Orin Nano (1020MHz) Cpu is lower.. its 1.7GHz max for the Nano Orin

[@Nowpinion]

So.. are there any CPU Power in the dock?

[@ocosmosinfinito]

Let's calculate which is faster A78c Switch 2 vs Jaguar Xbox One X --- Jaguar (8x at 2.3 GHz) 1. Jaguar's average IPC is 0.9. 2. The Jaguar clock is 2300 MHz. 3. We multiply the IPC by the clock: 0.9 by 2300 equivalent to 2070 units of performance per core. 4. The total number of cores is 8. 5. We multiply the performance per core by the number of cores: 2,070 by 8 is equivalent to 16,560 units of total performance. --- Cortex-A78C (8x @ 997 MHz) 1. The average IPC of the Cortex-A78C is 2.8. 2. The Cortex-A78C clock is 997 MHz. 3. We multiply the IPC by the clock: 2.8 by 997 equivalent to 2,791 units of performance per core. 4. The total number of cores is 8. 5. We multiply the performance per core by the number of cores: 2,791 by 8 is equivalent to 22,328 units of total performance. --- Comparison 1. The total performance of the Jaguar is 16,560 units. 2. The total yield of Cortex-A78C is 22,328 units. 3. We divide the performance of the Cortex-A78C by that of the Jaguar: 22,328 divided by 16,560 is equal to 1.35. 4. This means that the Cortex-A78C is 35% faster than the Jaguar. Here are the calculations in the requested format: Jaguar (8x @ 2.3 GHz) 1. The average IPC of the Jaguar is 0.9 . 2. The Jaguar's clock speed is 2,300 MHz. 3. We multiply the IPC by the clock speed: 0.9 times 2,300 equals 2,070 performance units per core. 4. The total number of cores is 8. 5. We multiply the performance by core by number of cores: 2,070 times 8 equals 16,560 total performance units. -- Cortex-A78C (8x at 997 MHz) 1. The average IPC of the Cortex-A78C is 2.8. 2. The Cortex-A78C clock is 997 MHz. 3. Multiply the IPC by the clock: 2.8 times 997 equals 2,791 performance units per core. 4. The total number of cores is 8. 5. Multiply the performance per core by the number of cores: 2,791 times 8 equals 22,328 units of total performance. 1. Jaguar's total performance is 16,560 units. 2. Cortex-A78C's total performance is 22,328 units. 3. We divide the performance of the Cortex-A78C over that of the Jaguar: 22,328 divided by 16,560 equals 1.35. 4. This means that the Cortex-A78C is 35% faster than the Jaguar. This calculation does not take into account the instruction path. What happens is that in addition to ARM CPUs having smaller instructions, they have significantly higher LDS transfer rates and cache than Jaguar. The A78c has an L3 cache and a system cache that serves as a last-level cache. GDDR5 vs LPDDR5x latency is much lower in favor of the T-239. As the data exceeds the cache size, it will fall into GDDR5 and will have a long way to go, increasing the bottleneck. Jaguar was weaker than a tablet Atom in its time and its virtue was HSA, taking some of the GPU's performance to speed up CPU operations. This dependence comes with costs. Synchronizing the CPU and GPU was a serious problem, forcing most games to 30fps. If not well optimized, even Tetris could suffer performance drops. So it's not unrealistic to think that a certain PS4 or XO game that has the slower Jaguar versions, runs on the A78c at 60 or 120 fps. It has the GPU for that.

[@kishaloyb.7937]

I personally think that the more development the Switch 2 OS, kernel and drivers goes through, it's stability will increase bit by bit, allowing Nintendo to unlock higher clock profiles for developers to use for their games. For now, at launch, everything will be down clocked for better stability. This is without considering the silicon binning process and QA that the Switch 2 SoCs have to go through, so the earlier manufactured batches of Switch 2 will have less stability at higher clocks compared to the batches that will be manufactured years down the line.

[@ascissordollynamedgwen9409]

Honestly I don't see the interest in clock speeds these days. Ever since the 90's to the mid 2000's clock speeds don't indicate performance anymore in CPU or GPU. At all. Back then when CPUs in like one year went from 600Mhz to 1.2Ghz, yeah that was big. Now that isn't the case. And hasn't been the case for the last 18 years now. Why is this still not registering to people? Remember when 10th gen intel cpus were hitting almost 6ghz yet Equivalent ryzen cpus were like 2 Ghz slower but were still just as effective? Or faster? Guys c'mon. And mobile part clock speeds are even more irrelevant. Because they work on a different level man.

[@nathanalbritton3766]

Based take

[@papasears1982]

They should have just made a eGPU dock, it's time Nintendo, give us the NX

[@Gshekela]

This underpowered Nintendo switch 2 can hold up 3-4 years.

[@eded9157]

That cpu clockspeed has to be a typo

[@djtomoy]

hopefully it will have a 24 hour clock, i certainly find those much more useful than a 12 hour clock, though it should have one of those too for the folks that prefer them.

[@Boomtendo4tw]

Its just a hdmi out and usb splitter. I guess call it a dock but it's like when your laptop goes into power save mode when unplugged vs plugged

[@mjjferreira]

Even if the screen it 1080p it doesn't mean all games will rum at 1080p on portable mode. Light games will reach 1080p while intense graphic games will not. Even for the system UI, browsing, etc...1080p will look much better!

[@OllieOtterOxen]

Anything not running native resolution looks worse. 1080p might barely be possible now at native resolution. Now imagine 8 years from now, it's going to be a struggle to run any ps6 ports above 240p

[@mjjferreira]

@ still a lot better than when switch1 was released and s1 had/has amazing games and sells like hot pancakes. Switch2 will be even better!

[@Twitch_Moderator]

*I can 100% verify that the docked mode is NOT 998 MHz. I've had hands-on experience with the ARM Cortex A78 chip and it goes up to 3.3 GHz (3,300 MHz).* *Nintendo is known for downclocking the CPU, but no further than 50% of max speed. EVER!! The Arm Cortex A57 was a 2.0 GHz CPU (in the Switch 1) and Nintendo downclocked it to 1.02 GHz.* *I PROMISE that is a typo!*

[@AnthonyJames1983]

I fully agree. I also have a few devices with the Cortex A78C and it is a powerhouse. Nintendo wouldn't go with an A78 seties to run it at 1/3 of its power potential. They would've just picked a cheaper, more power-economical processor from Arm. Plus, the previous leak by a reputable leaker stated at least 1.8GHz.

[@EversonBernardes]

What's the power draw of an A78 core at 3.3ghz on Samsung's 8nm? An A77 core on TSMC's N7 @ 2.6ghz draws 1W. Switch 2 supposedly has 8 of those cores, and the _total_ power budget for the system in handheld mode is likely under 15W, probably closer to 10W - that includes CPU, GPU, RAM, radios and an LCD screen. You want to be using 2, maybe 3W tops for the CPU. That puts the power budget for each core at a quarter to a third of a watt. As for why going for A78s and downclocking them? They've got the best perf per watt and are smaller than an A77 or an A76 (A5x and A3x series are very underpowered for the task).

[@jonzie24]

For reference, the OG Switch had the following specs: GPU clock: 307.2 MHz-460 MHz (handheld) and 768 MHz (docked). CPU clock: 1020MHz (on average), can get up to 1785 MHz when needed (when loading games, for example). Memory bandwidth: 21.3 GB/s (1331.2MHz handheld) and 25.6 GB/s (1600 MHz docked). In general, it offered roughly 236 GFLOPS in handheld mode and 393 GFLOPS in docked mode. It seems like Switch 2 will not be as strong as PS4, as some suggested. At least these specks don't suggest it.

[@Niko25467]

If the gpu clocks are true then yes it is 1.7 against 1.8 for ps4 and switch 2 using newer architecture and can use dlss to upscale from lower resolutions.on docked it's even stronger according to this leak reaching 3tf on gpu

[@blakenaftel3637]

And among the people actually clicking this video, there's a tiny little audience who legitimately cares how good Mario or Zelda is going to look when the only option available to them is to play them in a portable capacity. The other 99% are a curious mix. A lot of them ones actually interested in new Mario/Zelda are willing to wait until they can emulate them at 4K120. 4K30 is just a little too last-decade. And the rest? Are they legitimately squabbling over how powerful the latest smart-device-plus-Nintendo-properties is?

[@Bulletproof_Van]

The Bandwith of the Ram will slow down the Switch again.

[@hzafary]

I want 1080p in handheld mode. We can have nice things

[@Joscraft_05]

I don't want native 1080p in handheld lmao for a logic reason valve choose to use 1280x800 in the steam deck. DLSS quality to go from 720p to 1080p is the best. Even in performance from 540p to 1080p in that screen will look good tbh.

[@NogaBlitz]

If there's anything I want to see on my video preview it's these guys faces.

[@McDaniel77]

If this 1 GHz joke is true, Nintendo could easily double the performance by just raising the clock speeds to 2 GHz in the future.

[@rallyscoot]

I think that DLSS will be standard on this device.. So it can render lower resolution = lower gpu clock. But output wil be the same. And un-docked probably without.

[@type2red]

looks like something Nintendo would do smh.

[@8BitRetroMonkey]

Not interested in doing leaks since they got the PS5 Specs wrong hey 😂

[@pac0113]

I’ll take 60fps over 1080p any day

[@NivraNabru]

stupid question. would it be feasible if the dock had like a hardware that could increase bandwith and memory to gpu and cpu? or is the power draw from.the dock be enough to boost performance on switch2,? noob to technical terms here. thanks! 😅

[@playerroku4412]

Please stop. The answer is NO. Nintendo is not using thunderbolt. There is no boost from the dock. The handheld simply realizes it's docked and uses it's full power. Not any additional performance from the dock itself. This was said from the original switch and it's the same here

[@JuanDiaz-jo1rw]

The Switch 2's CPU clock speed is 1100.8 MHz in handheld mode and 998.4 MHz in docked mode The Switch 2's GPU clock speed is 561 MHz in handheld mode and 1007.25 MHz in docked mode The Switch 2 is expected to have an NVIDIA Tegra SoC, 12 GB of LPDDR5 system memory, and an Ampere-based GPU The Switch 2 is expected to have a custom-made version of Nvidia's T239 processor The Switch 2's performance is expected to be similar to a PlayStation 4 or Xbox One. However, Nintendo's focus is on their ecosystem, not raw performance.

[@bobertoroberto953]

It’s not a typo on the CPU speed. Leaker said 1.8GHz was, “Hopium.”

[@arch1107]

no, the cpu can go up to that i bet, a78 the core design can go up to 3ghz or more but heat and power consumption matters alot here, so it is a wiser choice, focus more on battery and gpu

[@bobertoroberto953]

@@arch1107It’s factual he said that. Total system power consumption on this thing is going to be like 5-7W.

[@arch1107]

@@bobertoroberto953 i am not sure about those values when docked still it has a fan and a radiator, so it is still a concern

[@Pyovali]

Nintendo is finally bringing their machines to 2013

[@bits2646]

Probably they're not pushing tegra to the limits cause it's probably on tsmc 7nm or even Samsung 8nm, I'd suspect Samsung has capacities left now that Orin is almost gone or maybe tsmc could squeeze out on its 7nm line, but probably it's decided by Nvidia anyway so who's producing their automotive is also producing t239, that's why they changed the intake system, remember we still don't know what's the actual cooling inside, but they probably were looking at up to 8w whole device, 5.5-6w soc non docked and 10w docked pure SOC power cause screen is off. Also, the bandwidth increase had big costs too, they got wider bus for start, and docked also up the frequency... But that shows me it should be very overclockable, at least on CPU side and depending on ability to cool it.

[@robreich6881]

Maybe the increased GPU clocks make it thermally more difficult to support the CPU clocks?

[@arch1107]

most likely, when docked it will push for 1080p or higher, so gpu eats more power and generates more heat, cpu the higher the resolution goes, matters less so downclock, less heat all works well, makes sense

[@CynHicks]

I'm not buying that this leak is accurate. The reason it seems so off is because it probably is.

[@Liminal_Aero]

fr

[@rudiger86]

3:11 I think Olie is spot on about some CPU cores being disabled in handheld mode and therefore they can run the CPU at a slightly higher clock speed.

[@buggy946]

Nintendo’s plan is obvious. I can definitely see them streaming higher quality games on the Switch 2. If your concerned about certain games running properly, I can assure you they will give us the option to stream them, prompt us to have a raw connection to the internet for select titles that require more power and performance. Maybe NVIDIA GFN technology.

[@ForTheViolence]

Nintendo can’t even get multiplayer quality right. I highly doubt they are going to pull out game streaming at a decent enough quality out of no where.

[@Spanluver]

I dunno their streaming quality is pretty bad

[@midwestpotato]

Cloud games on swift did horrible sales I doubt they dig into that route at all

[@NicolasSilvaVasault]

that was the absolute worse idea nintendo had, for such an underpower console, it was painful

[@danfay6201]

F Nintendo. Antiquated hardware for modern prices yet again.

[@zeniktorres4320]

i totally agree.

[@Kniffel101]

Remember that Nintendo makes a profit with their hardware, Sony and Microsoft didn't when they launched theirs.

[@xtr.7662]

The switch is the same price it was at launch what did you expect

[@Ixiah27]

And yet they make billions, unlike sony or xbox, so......

[@aviatedviewssound4798]

What will power the AI core to enable DLSS for docked mode if they don't downclock other components?

[@desfile21]

I wonder if the CPU # is correct, and maybe a seperate dock CPU is used, so the Switch 2 doesnt need to run as fast.

[@MrWhis-fe5hs]

So this thing won't even be as powerful as the steam deck, which is already considered outdated

[@darickfoxo7986]

Them fan boys on Reddit were like, “guys it’s really happening it’s going to be as powerful as a Xbox one x pro and it has dlss so it will be as powerful as a ps5 with upscaling. It’s really happening 🤣🤣🤣🤣🤣

[@bettybunbun9664]

Steam Deck Oled is still selling at 570 EU. If the (presumably) 400 EU Switch 2 was more powerful that would make the Steam Deck a rip off.

[@Liminal_Aero]

tf are you talkling about

[@xtr.7662]

Not on the cpu side but everything else especially with optimization its gonna be substantially better

[@shifu9276]

We’re not interested in doing leaks anymore… but here’s a leak from someone else. 👍

[@Al3x-c4s]

Lpddr5 102gb/s speed would have been the barrier anyway, the maximum speed from current handhelds, all stronger handheld systems on paper, are bottlenecked, thats why they all play in the same league, where steamdeck maybe has a bit more raw power, it lacks the dedicated OS and the possible more modern DLSS. It was the right descision from nvidia and nintendos standpoint, batterylife isnt that much important, powerbanks and powerplugs are everywhere today. They can ship it for 399 and sucess will coming

[@ChristopherZarate-g1s]

Maybe typo error from the leaker or maybe in handheld mode it shuts down some cores, maybe.

[@cyrus8886]

Or he's just full of shit

[@r.g.8339]

What if NS2 has VR model also, that would be next revolution gaming. Imagine having a 3 in 1 console. Could it be, having near ps4 pro raw power with DSSL?

[@redder3076]

What if they release LABO 2 and use that for the NS2's VR function instead?

[@r.g.8339]

@@redder3076 Yes, I would think they have to make the pheriperal for the VR mode. It would need a camera and sensors, and the screen and power comes from the NS2. And the joycons would work perfectly. But Im not expert, thats why I asked if is it possible with a 3 teraflops and DLSS to do VR?

[@r.g.8339]

And if the screen is the console also. It would be the first true inalambric VR gaming console

[@pedrobaptista9213]

You could do some tests with an RTX 3050 GA107 Ampere, similar to what you did with the RTX 2050, I believe it would be closer to what the Switch 2 will be.

[@gemini-mg6sc]

It is closer to the RTX 3050. Switch 2 reportedly has a 128 bit memory bus width, which is the same as the RTX 3050. The RTX 2050 has a 64 bit memory bus width.

[@DragonOfTheMortalKombat]

​@gemini-mg6sc rtx 3050 has gddr6 not lpddr5. Rtx 2050 is closer.

[@DragonOfTheMortalKombat]

They've already tested 2050

[@EversonBernardes]

Switch 2 is closer to a 2050 Mobile than to a 3050. They've already tested an underclocked 2050.

[@JackFrost-1856]

I don't buy this. The CPU being just 100 MHz slower makes no sense especially when it has access to more power when docked. That is without even factoring in that the Switch CPU maintained the same 1 GHz parity between handheld and dock - so why would that change here??? The likeliest reason is that the datamined info is incomplete/was not updated to match the handheld CPU clocks.

[@Joscraft_05]

The info is probably too incorrect still. We need to wait until know more about it.

[@brendangannon9286]

Now these guys look like they play video games

[@gian9568]

I don't agree with John about the screen. Imagine playing your Switch 2 in 6 or 7 years, you would be playing on a 720 P screen in 2032...That would be weird. With an 8 inch screen, 720 P is not enough. 1080 P, like the Portal, is perfect I think.

[@cyrus8886]

720p on an 8inch screen Is Sharper than 4k on a 42 one

[@evanskan6326]

720p on an 8 inch screen is more than enough. Don't look the resolution but the "density" of pixels.

[@valhallaoutcast]

you are overthinking that, my SteamDeck Oled is 800p and looks amazing the Switch not being Oled would be a bigger miss thand 720p vs 1080p

[@MisterPikol]

nah bro 720p would look good on 8 inch, aliasing would be removed by dls, and it would give the devs sooo much more power lol

[@Slickman2006]

Maybe a few CPU cores are disabled in handheld mode?

[@WigWoo1]

I hope unlike the switch one developers can choose to have the switch 2 run at full dock power even in handheld mode. I wish we could choose to sacrifice battery life if we would rather have better performance I would rather play at 60 FPS for two hours instead of playing at 30 FPS for five hours

[@JonnyLin7]

What I found strange with the Switch 1's is the GPU was downclocked, even in docked mode. I know a few select games later in the Switch's lifecycle (MK11 was one of them) were able to boost the clock speeds a bit, but even those games didn't have access to just the full base clock speed. I understand downclocking for handheld mode to preserve battery life and at a lower resolution in handheld the GPU clock speed is less impactful, but for docked mode it doesn't make sense why they won't let it run at the maximum base clock, given it would probanly help a lot to increase resolution on the big screen. Curious if the Switch 2 is also downclocking in docked mode or if the ~1ghz GPU clock speed is the maximum? I feel like it is downclocked again though given the Steam Deck and ROG Ally's processor both having significantly high GPU clock speeds. Maybe for Switch 1 the cooling could be the concern, but we've now had multiple similarly powerful gaming handhelds that have good enough cooling to take advantage of the max clock speeds, so I would have hoped wih Switch 2 they could figure out how to have good enough cooling to take advantage of the full base clock speeds.

[@EversonBernardes]

There's nothing "strange" about them running low clocks even docked - you still have to cool a handheld system, and the cooling is limited by an internal fan, small mass of heatsinks and very small case volume. Switch form factor is inherently thermally limited.

[@krgaming3738]

could be enabling more cores in docked

[@anthonyjohnson8168]

Yeah I never bought a Nintendo console for power. They make great games and that’s all that matters. In todays age there are other options for more power. Nintendo is doing everyone else a favor being underpowered. If they had the most powerful handheld its game over for Sony and Xbox.

[@Aggrofool]

Switch 2 gonna be underpowered

[@andraszoltan2]

Almost certainly engaging more performance cores on top of any efficient cores, so thermally, the clock rate has to come down. The huge bump to memory bandwidth will likely eliminate any loss in performance caused by raw clock speed. Especially in any title which actually takes advantage of multiple cores (not necessarily great for UE-based titles, but then, nothing ever is).

[@JackFrost-1856]

The Cortex A78C consists of 8 cores in one cluster. There are no separate performance/efficiency cores. I do not for a second believe that CPU clock speed will differ in the final product, this just seems to be a case of missing information

[@easyygo3008]

​@@JackFrost-1856 😂😂😂😂😂😂😂😂 Ao fuckn fanboy 😂😂😂😂😂

[@Niko25467]

They are no efficient cores on Switch 2 it's using a78c in single cluster it's not a fanless 200 euro phone

[@itchyisvegeta]

As long as we get the equivelent of Series S performance, i will be happy. I will absolutely take 1080p 60hz over 4K 120Hz as a primarily console, for the benefit of portability and docking, assuming the games I want to play come out on it.

[@HALFSQUASHED]

What games do you want to play on it?

[@davidmuller6011]

It won’t be getting series S performance because it’s still a handheld

[@iamtofuboy]

I'm almost 100% positive this will not get Series S performance. It's likely closer to something between a steam deck and an ROG Ally. so more likely between PS4-level and PS4 Pro performance. What it does have going for it is that it is rumored to have 12GB of RAM which is 2GB more than what the Series S has, so texture quality would likely be better, but performance, perhaps not as much.

[@cyrus8886]

You're not getting series s performance When even 700$+ pc handhelds aren't reaching those heights

[@jonzie24]

It will not even reach PS4 performance, come on now.

[@D.u.d.e.r]

Nintendo for several console generations cares more about profit than exciting/amazing HW specs or top notch visuals. Even 2 Ghz CPU clock for rumored 8 nm chip is quite low especially with active fan cooling while docked/plugged in power! Switch 2 is definitely NOT a Gamecube-like release which had incredible HW at the time of its release. At least memory bandwidth has been multiplied 3+ times. I hope M$ will eventually steal the show in this space and deliver truly revolutionary Steam Deck like handheld console with the new Xbox which will become even more powerful when docked and connected to the stationary HW either by doubling of the GPU power through proprietary multi GPU tech or classic docked eGPU feature. Competition is great, imagine market without the PS, Xbox and PC and just Nintendo with its childish games and low tier HW specs to maximize profits. On the other hand Nintendo is surely going to make lots of money and become successful with the Switch 2, however it's definitely not cup of coffee for the enthusiast/adult gaming community.