-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Switch dataminer from Famiboards suggests the Switch 2's portable GPU clocks will be 561MHz. He also said 1.8GHz for the CPU is "hopium"
Little Mac
Gold Member
SweetTooth
Gold Member
Zathalus
Member
There was a handheld with that capability in 2013?Welcome to 2013 then, I can't wait to see your reaction to 3D audio in 2042
LordOcidax
Member
Is that supposed to be a joke?Welcome to 2013 then, I can't wait to see your reaction to 3D audio in 2042
Zannegan
Member
You mean doomers acting like everyone else is a hyper in denial, and hypers acting like everyone else is a seething doomer, with most of us just sitting on the sidelines with popcorn watching the warz get ugly? New hardware reveals really are the best.Can't not wait to revisit this thread after they finally announce it and reveal the specs.
We go through this every Nintendo hardware cycle.
LordOcidax
Member
I didn’t say 4Kr, that’s depends on the game. 1440Pr is going to be the most common resolution when docked IMO, We are going to see some games at 4Kr, especially from Nintendo and some less demanding PS4 Era games, but i don’t think they are going to go above 1080p input resolution due the memory bandwidth limitations… Maybe some games from ShinEn that are super optimized, but those are going to be exceptions.
Last edited:
BlackTron
Member
Nintendo hasn't revealed the price but they have revealed many times over their strategy of making mass market consoles to play Mario. It's supposed to be as easy to get under trees as possible so they can sell games. The ROG Ally which currently goes for $4-500 discounted can't surpass Series S.
I actually expect games to be easier to put on Switch 2 running fine though, thanks to nvidia vs AMD and 12GB unified ram instead of 8GB good/2GB crippled lol.
One has to wonder, just how thick the helmet that comes with this post is.Welcome to 2013 then, I can't wait to see your reaction to 3D audio in 2042
Zannegan
Member
Ain't it fun though?
Generic
Member
Fortnite on the PS4 runs on UE5. The performance was variable but recently Epic released an uptade that made the game smoother.
Buggy Loop
Gold Member
Without nanite or Lumen. That's what I've said, it was backported for continuing compatibility across all platforms, but they disabled the 2 main features in-bricked in the engine...
James Sawyer Ford
Gold Member
Perhaps, it seems like Switch 2 could be a fairly well balanced system overall
I’m more curious just how far Nintendo will push the Switch 2
There’s some really great looking switch 1 games, like MK9, MO, BotW, TotK, and LM3, although they are rare.
Also how much will be cross gen….i suspect a lot will be
bundylove
Member
Yes
This is next level. Ps5 performance lol
LordOcidax
Member
Specs speculation is one of the most fun things before the announcement of new console in gaming forums, the problem comes when some console warriors feels that their favorite plastic machine is menaced by the new plastic machine. Still easy to spot who they are, the low knowledge about the topic is always obvious.
Last edited:
nordique
Member
Those gpu numbers are actually very nice. Handheld performance should be fantastic. Docked should be fine.
CPU rumoured clocks are low but still a next gen leap over the switch and should outperform the jaguar cpu in ps4 easily presuming it is 8c a78c
ram - so is it not lpddr5x?
Let’s see the price
CPU rumoured clocks are low but still a next gen leap over the switch and should outperform the jaguar cpu in ps4 easily presuming it is 8c a78c
ram - so is it not lpddr5x?
Let’s see the price
bundylove
Member
I talked to some people and the price is set at 449 Us
SweetTooth
Gold Member
??
Is that supposed to be a joke?
What? All of you know that Nintendo has always been cheap and a generation or more behind others.. don't get riled up when I state some fact, accept it and move on.
(Ninty has always been behind tech wise) Only exception was Gamecube:
NDS vs PSP
3DS vs Vita
Wii vs PS3/360
WiiU vs PS4
Switch vs PS4 (LOL)
Switch 2 vs PS4 (yet again Lol)
And now we have some poor souls trying to overhype Switch 2? Just enjoy the one dimensional Ninty 1P output and just pray that Nintendo might be able to add some proper online functionality and move out of early 2000s.
SonGoku
Member
So its docked mode is roughly on par (being generous) with a HD 7870 which is around 20% better than base PS4 GPU before DLSS kicks in...
Still nowhere near the PS4 Pro let alone the Series S GPU which is over 2x faster on raster and light RT...
Coincidentally nearly a match for the GTX 570 which is the go to GPU Digital Foundry used to compare PC with the PS4 early on the gen before the PS4 pulled ahead
That famiboards thread is like an alternative reality where posters think the Switch 2 leaked specs are anywhere close to the Series S, some even went as far as to claim better image quality than Series S because of DLSS but DLSS cant make up for a 2x Gap in power specially when the Series S already going as low as 540p for some games, i seriously doubt 240p games reconstructed with DLSS will produce better results.
Overall the Switch 2 will be in a similar situation to the Switch one when compared with PS5 and SX
Also considering how low powered the GPU is, it wont have significant Tensor Core grunt (1/4 RTX 3060 is ~25 AI TOPS) to do any significant image reconstruction to make up for the performance delta with Series S
Last edited:
Zathalus
Member
You’re laughing about a handheld in 2025 having the GPU power of a 2013 console. Obviously the two are not the same.
Kataploom
Gold Member
Lol there are numbers already, we don't have wish anything.
Also, Switch 2 chip is from the same generation as PS5.
Buggy Loop
Gold Member
SonGoku
3.1/1.84
?
20%
Tegra Orin AGX 32GB with 14SMs (2 more than T239) is 200 TOPS. It loses 8 Tensor cores and lower clock than the full fledge Orin AGX @ 275 TOPS. At worst the 8 tensor core less than the 32GB version but with higher clock will lower it down to ~140 TOPS range in dock mode. More than the best of Turing's 89 TOPS
Where the fuck are you even fetching your numbers. What a clusterfuck. We have data for T234. T239 is based on that.
Last edited:
ReyBrujo
Member
It worked out well 4 out of 5 times for them so why change now?
SweetTooth
Gold Member
Tell that to Nintendo fans
LordOcidax
Member
Remember that the TFs doesn’t matter anymore is all about IA upscalers and the final results… Oh wait… Did the PS4 had ML hardware?
Zathalus
Member
I think most are aware of that fact, it’s a handheld that just has a boost mode when you plug into the TV. It’s still designed around being portable and needing to last around 3 hours on battery. Having the GPU power of a Xbox One (in portable mode) plus 50% more RAM, fast storage and dedicated compression, faster CPU, and much newer feature set is perfectly fine for a $400-$450 handheld in 2025.
Last edited:
asdasdasdbb
Member
There were people on GAF who claimed it had to be at TSMC, you know. And yes it would be meaningfully better (even with no spec changes) if it was fabbed there... but Nintendo is cheap.
Last edited:
HogIsland
Member
It's not even doomer to say switch 2 will be underpowered and underwhelming. It would be less of a mass audience device if it tried to match the Steam Deck. And it would be more power than Nintendo even wants/needs for 1st party development. Nobody thinks the next Metroid is going to look like Cyberpunk... do they?
Last edited:
SonGoku
Member
Different architectures cant compare TFLOPs in a straight forward way, for 3000 series and up nvidia started something similar to RDNA3 dual issue shaders with the multiple operations per core to come up with the number to the comparable RDNA2 GPU has less flops.
Im going off the claim 1/4 RTX 3060 GPU, i used relative performance graph from TPU to avoid TF miss matches
Im going off the 1/4 RTX 3060 comparison, the RTX 3060 has ~100 TOPs
Last edited:
yurinka
Member
Basically a raw TF numbers similar to PS4 or Steam Deck with extra stuff like DLSS that would help it give the feel specially in portable mode of maybe something like a PS4 Pro / almost Xbox Series S depending on the case.
But take it with a grain of salt, because AMD and Nvidia TFlops aren't comparable, and also in each brand the TFs of each generations aren't comparable. So who knows, we'll have to wait and see DF benchmarks.
Last edited:
onnextflix5
Member
Development Sdk, let’s hope it’s an old sdk and not final, even for 8nm, this is poor.
Buggy Loop
Gold Member
Not even going to bother
AMD's dual issue on RDNA 3 doesn't even work for gaming or anywhere close to Ampere. Its a solution that relies on compiler solution and it failed big time.
One quote from Chips and cheese microbenching of RDNA 3
So yea, dual issue is never happening in games for RDNA 3. It's not even anywhere near a comparable solution. It's not even physically close to the same approach.
Ampere's 2nd cuda core is exact same one as Pascal and works concurrently with the FP32 dedicated unit.
Like I wouldn't even know where to start to explain here. My posts before already cover all this anyway. T234 Orin AGX, go look it up. You're 2 years late to the party of this discussion since T239 was rumored.
Last edited:
HogIsland
Member
Lol no. Maybe at 50W. At 7w it's not going to come close to "theoretically PS4".
yurinka
Member
It will depend of many things, we don't have the full story. Maybe they have some secret sauce as Sony did in PS5 where in theory they had worse specs than Series X but later with how they managed the clock and the I/O ended having the same or better results in most games.
Last edited:
thicc_girls_are_teh_best
Member
Yes
I meant anyone here; any random on Twitter would of course say that.
Yep; I'm confident it's a smartly designed system and will provide adequate performance for multiplats that aren't lazy, unoptimized ports. And to counter
HogIsland
's point, ...we don't know what wattage Nintendo want for Switch 2 when portable. We don't know how efficient they & Nvidia have made the architecture and chip design for performance relative watts.And, we have no idea what the wattage profile will be like when docked. For all we know they could let it get up to 40-50 watts when docked. That should afford lots of room for game performance, and that's not even getting into DLSS boosts when docked.
It won't "smoke" PS4 Pro or Series S when docked, but it'll likely perform very favorably with them all things considered.
Last edited:
nordique
Member
Interesting
If true
That’s about $649 CAD which seems to be $200 more than mrsp OLED
From where you guys taken that the cpu will be faster, can you show some benchmarks or data proving this?
Last edited:
Soapbox Killer
Grand Nagus
By what metric are you comparing the SWITCH 2 with PlayStation 4, while also comparing Wii U with PlayStation 4?
Fake
Gold Member
bundylove
Member
Just remember when switch released thr candain dollar was on par with US dollar.
And ever since there is nothing but inflation.
I just paid a 1000 canadian bucks for the pro. I paid 400 candian for ps4 pro.
Is that not using the DLA? The 2050 with 64 Tensor cores is rated for up to 48 INT8 TOPS. So scaling that down to 1007 MHz and 48 Tensor cores should yield the 25 TOPS figure. That's not the end of the world though.
Rich from DF tested a 2050 running at 750 MHz, which should also have ~25 TOPs, and found that DLSS Quality at 1080p had a cost of 3.35 ms and DLSS Performance at 1440p had a cost of 7.7ms. So at least upscaling to 1440p should be viable for 30 FPS games.
Last edited:
Buggy Loop
Gold Member
Info - TOP 20 of the World's Most Powerful CPU Cores - IPC/PPC comparison
Added cores: A53 - little core used in some low-end smartphones in 8-core config (Snapdragon 450) A55 - used as little core in every modern Android SoC A72 - "high" end Cortex core used in Snapdragon 625 or Raspberry Pi 4 A73 - "high" end Cortex core A75 - "high" end Cortex core Bulldozer -...
forums.anandtech.com
A78 is more performant per clock than Zen 2.
But #1 ranking in PPA/GHz (not up to date with new CPUs, but Zen 2 is still there conveniently), which is Power Performance Area metrics compared to GHz. Meaning how much performances you get out for the area the core takes on Silicon and how much power they take.
Not even sure its even feasible to find scraps of a Jaguar equivalent chromebook IPCs. Closest is that Jaguar was pretty much IPC of Bulldozer and AMD's own figures
DF here explaining how terrible Jaguars were
So even if we assume that Jaguar somehow at the upper part in IPC on that AMD slide (its not trust me), Zen 1 was 40% higher IPC over excavator, and Zen 2 was 29% higher IPC than Zen 1. And A78 has higher IPC (but not as big difference) than Zen 2. And Excavator has definitely higher IPC than Jaguar. So it stacks up quite a lot.
Last edited:
Zannegan
Member
I guess that depends on your definitions of underpowered and underwhelming, which depend on your expectations. Where you expect a $400ish Nintendo/NVidia portable console (here defined as something designed to function as a home console, but fit into the power restrictions and chassis of a portable) to be able to pull off is kind of what distinguishes the doomer from the hyper from the rest of us.
HogIsland
Member
The sweet spot will ~2017 type games: Titanfall 2, Horizon Zero Dawn, Prey. Some cross-gen era games like Elden Ring, Cyberpunk and Red Dead 2 will be possible with noticeable compromises.
Zannegan
Member
I think that's a fair assumption, not that I'm conversant enough in silicone to really know what I'm talking about. At the same time, given what we're hearing, you could reasonably hope for a little more without being deluded. Personally, I like to speculate to the moon but keep my actual expectations pretty firmly tamped down. We'll have to wait and see exactly what it can do, but hooefully that's not too long a wait.
Regardless, it should certainly be a full generational leap over the Switch, and given the miracles some of Nintendo's studios (NOT Gamefreak, lol) pulled off there, it'll be exciting to see what they can do with significantly more to work with. I also want to see what, if anytjing, this "mousecon" can do.
Last edited:
SonGoku
Member
This has nothing to do with my post never did i make such a claim about RDNA3 inflated flops making an impact in games, i merely pointed out that you cant compare TF numbers across different architectures and gave a possible explanation as to why comparable cards with less flops have similar or greater performance.
Point in case the RX 6800 a 16.1TF card is 10% faster than the RTX 3070 a 20.3TF card
The equivalent Series S GPU the 6500XT at 5.5TF is 57% the performance of RTX 3060 at 12.7TF
The Switch 2 is rumored at a quarter the 3060 making it half the performance of the equivalent Series S GPU (5.57TF/2)
Which again puts it on on a similar ballpark of raster perfomance compared to the base PS4
Interesting will take a look at the video by any chance do you happen to know if the cost of reconstruction is higher the lower you go or is the opposite?
It comes to mind because if the Switch is half the raster performance of a Series S it will have to render games at 720P for the less demanding games and 540p to 240P for the more demanding games or drastically scale back settings to render at 720p for a "performance" DLSS 1440p reconstruction
Based on these specs I wouldn't be surprised if multiplatform devs target 540p with scaled back settings for a 1080p performance DLSS reconstruction on docked mode
Last edited:
As you can see in the video, it scales based on the difference between input and output resolutions. So dropping down to 540p for the base resolution for the 1440p upscale would incur a further cost.
Based on the titles tested in the video it seems that Series S level settings are only going to be viable at 1080p DLSS Balanced/Performance. So it's a question as to whether developers stick to a 1080p output or go with 1440p at reduced (eg. Steam Deck) settings.
Info - TOP 20 of the World's Most Powerful CPU Cores - IPC/PPC comparison
Added cores: A53 - little core used in some low-end smartphones in 8-core config (Snapdragon 450) A55 - used as little core in every modern Android SoC A72 - "high" end Cortex core used in Snapdragon 625 or Raspberry Pi 4 A73 - "high" end Cortex core A75 - "high" end Cortex core Bulldozer -...
A78 is more performant per clock than Zen 2.
But #1 ranking in PPA/GHz (not up to date with new CPUs, but Zen 2 is still there conveniently), which is Power Performance Area metrics compared to GHz. Meaning how much performances you get out for the area the core takes on Silicon and how much power they take.
Not even sure its even feasible to find scraps of a Jaguar equivalent chromebook IPCs. Closest is that Jaguar was pretty much IPC of Bulldozer and AMD's own figures
DF here explaining how terrible Jaguars were
So even if we assume that Jaguar somehow at the upper part in IPC on that AMD slide (its not trust me), Zen 1 was 40% higher IPC over excavator, and Zen 2 was 29% higher IPC than Zen 1. And A78 has higher IPC (but not as big difference) than Zen 2. And Excavator has definitely higher IPC than Jaguar. So it stacks up quite a lot.
Correct me if i am wrong but this is taking the best of all scenarios isn't? I mean with all the possible cache and everything
If that's the case I am not sure Nintendo will go this way, i wouldn't be surprised if actually trade blows with the jaguar
Last edited: