-
Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.
StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (Read the staff posts before commenting!)
- Thread starter Dakhil
- Start date
navierStokes
Octorok
I was looking at some benchmark recently and saw that laptopmedia had covered quite a few benchmark/gameplay footages with various laptop GPUs (e.g. RTX 2050 @ 35W, RTX 3050 @ 65W-75W, RTX 4050 @ 45W).
Now, of course, there's some variance with the results due to the CPU being used and perhaps even memory clock speed having an impact. I checked most of the benchmarks and the laptops or spec's they use and I'd say they're relatively close in capabilities. Although the amount of metrics on display isn't giving much validity, I know they're a legit source. One note to make is that they may not have been made with the same drivers....
Nonetheless, perhaps for a layman's showcase of how node process/power efficiency/clock speed all play a role.
I put a side-by-side of Death Stranding, and it's quite impressive how the RTX 4050 performs, practically doubles compared to the RTX 2050, with only a 10W increase of TGP, while also leading the RTX 3050 that has a 20W higher power consumption. Now there could be dynamic boost being at play (and other variables), although I suspect these laptops are hitting their max-rated TGP; 35W, 65W, 45W for the RTX 2050, 3050, 4050 respectively.
viewsync.net
I guess ideally, you’d want to compare at the same clock frequency and then observe the power consumption - frametime, but for such an analysis you’d need to have each of these devices (and a vbios that allows to have similar ranges of frequencies). Also the RTX 4050 has more CUDA cores (2056), maybe the RTX 500 (2048 CUDA cores) can give some interesting metrics. but given the recent announcement of that chip, it’ll take a while
.
Now, of course, there's some variance with the results due to the CPU being used and perhaps even memory clock speed having an impact. I checked most of the benchmarks and the laptops or spec's they use and I'd say they're relatively close in capabilities. Although the amount of metrics on display isn't giving much validity, I know they're a legit source. One note to make is that they may not have been made with the same drivers....
Nonetheless, perhaps for a layman's showcase of how node process/power efficiency/clock speed all play a role.
I put a side-by-side of Death Stranding, and it's quite impressive how the RTX 4050 performs, practically doubles compared to the RTX 2050, with only a 10W increase of TGP, while also leading the RTX 3050 that has a 20W higher power consumption. Now there could be dynamic boost being at play (and other variables), although I suspect these laptops are hitting their max-rated TGP; 35W, 65W, 45W for the RTX 2050, 3050, 4050 respectively.
ViewSync - Multiple YouTube Viewer
Streamlined multiple YouTube viewer with built-in support for content creators.
viewsync.net
I guess ideally, you’d want to compare at the same clock frequency and then observe the power consumption - frametime, but for such an analysis you’d need to have each of these devices (and a vbios that allows to have similar ranges of frequencies). Also the RTX 4050 has more CUDA cores (2056), maybe the RTX 500 (2048 CUDA cores) can give some interesting metrics. but given the recent announcement of that chip, it’ll take a while
.RTX 2050 [35W]
As per TPU; GeForce RTX 2050 35W = 982-1372 MHz
RTX 3050 [65W]
RTX 4050 [45W]
As per TPU; GeForce RTX 2050 35W = 982-1372 MHz
RTX 3050 [65W]
RTX 4050 [45W]
Amanieu Systems - Technology
Overview The Tango binary translator consists of several components: Dynamic translator - The dynamic translator is the core component of Tango. It wraps a 32-bit ARM process and runs it as 64-bit process by translating all of the instructions into AArch64 code. Pre-translator - The pre-translator
www.amanieusystems.com
Amanieu Systems - Performance
Tango uses state of the art optimizations to minimize the overhead of binary translation. The Technology page describes these optimizations in more detail. In this page we present a few benchmark results which show the performance of Tango v2.0.0. We run the benchmark on CPUs supporting both
www.amanieusystems.com
Here are at least the performance claims of this Tango emulator. It wouldn't work out of the box on a hypothetical Switch 3, since it seems fairly Linux specific, but a customized version could by used, should Nintendo choose to license it.
This is definitely not the only option that exists, though. Just off the top of my head, there's also dynarmic (which I believe was notably used by Yuzu) and probably qemu.
TLZ
Like Like
I'm counting on it!Relax, we'll have ubisoft to the annual fuck up and leak the specs.
Have faith in me brother.
For regularly coming in to stoke tensions in the thread you have been threadbanned for 1 month. - ngpdrew, xghost777, BassForever, mariodk18
I’m not gonna get into this argument, but this thread is years intelligent, well-meaning hopium-fueled fan-fiction math and theory-crafting:
From
“Team 2020”
“No wait 2021
“Definitely 2022”
“Based on history… 2023”
“Can’t fail 2024”
To
“It can’t be 8nm”
I don’t post here often anymore and this will be my last response for a while.
It’s been mostly fun!
ItWasMeantToBe19
Manakete
These PS5 Pro specs just suck, lol. It’s like barely more powerful than the Series X other than their machine learning stuff. At least they’ll be able to stay under $700?
The only good thing about this system is that Sony will have years of testing data on their DLSS competitor.
The only good thing about this system is that Sony will have years of testing data on their DLSS competitor.
lattjeful
Like Like
- Pronouns
- he/him
I mean, doing math to figure out a range of target specs and predicting release timing are two different things?
I know there's been some napkin math to try and figure out production and release timing but I don't think we should handwave the other stuff as hopium just because this one specific thing we've guessed on has been wrong. I'm not saying that you're wrong or that what we say in the thread is gospel btw, I just think it's a lil disingenuous to conflate both our shoddy release timing guesses and the technical makeup of the system as hopium or fanfiction when there's years of both speculation and hard data supporting the latter.
if you can't see the difference in these or basic "learn new info, adjust accordingly", I question if you were even here for honest discussion. you leaving just might be the best thing then
ItWasMeantToBe19
Manakete
The PS6 being able to do 3 node shrinks over the current gen while adding in better neural network and RT support will at least allow it to be a generational leap even though the node shrinks suck really bad.
If we can get 1080p>4K, path tracing on high detail virtualized geometry next gen, that would be good enough to stop on. No need for PS7 etc.
If we can get 1080p>4K, path tracing on high detail virtualized geometry next gen, that would be good enough to stop on. No need for PS7 etc.
Armoured_Bear
Like Like
- Pronouns
- he
As someone who built their own living room gaming PC (in an HTPC case) and who games a lot on the PS5, I can assure you that my PS5 does not at all feel like a PC, it's a console, and a good one at that.
If anything, the very fast SSDs and loading of games (in the Series X also) gives very much an old school console feel, rather than waiting during long loading times.
Steve
Koopa
- Pronouns
- he/him
Well... 2020-2022 were mostly switch pro leaks, which Nintendo sadly cancelled because of chip shortage and as well of the switch still selling massively.
2022-2023 were mostly silent from what i've gathered. Until september when they showed case the switch 2 to developers.
2024 sadly got delayed because Nintendo are freakish when it comes to having their main line games as polished as they want. Also because the 3DS and the Wii U launched with mostly mid games. So in their mindset is that they want to have ground breaking exclusives in their first year of the switch 2.
2025 is the most solid timestamp we have, mostly because of credible sources and Bloomberg and the japanese firm whose called nikkel something.
Lastly 8NM doesn't makes sense since Nintendo will want to maintain good batterly life and performance on handheld.
Lastly the thing we don't discuss is that the switch was one of the most high level machine when it came out in 2017 for handheld gaming. Most phones weren't capable of handling Botw and Mario odyssey, when i think about the switch I have the mindset of it being the successor of the 3DS, because if we compare it to the Wii U then it's slightly better, but if we compare the switch to every Nintendo Handheld, then it's a behemoth of power and capability.
But in the end that's my opinion and it's important to note that Nintendo are recently known to holding finished games in their cabinets when it comes to release, like metroid prime remastered was ready for a year, until Nintendo decided to launch it.
TomNookYankees
Koopa
- Pronouns
- He/Him
This has been my thinking lately too. A lot of things that people assumed couldn't be true because it would be bad ending up being true. Could be the same for 8nm.
I remember last year when people in this thread kept doubting Nate for thinking that late 2024 was much more likely than late 2023.
ItWasMeantToBe19
Manakete
8nm would just be extremely funny though as we know this system has to be pretty powerful based on the tech demos so this would have to be a just massive device.
OP
OP
And Kepler's saying that PlayStation 5 Pro's APU inherits RDNA 4's ray tracing technology based on additional PlayStation 5 Pro devkit information.
So I do wonder how close can Nintendo's new hardware can reach the PlayStation 5 Pro in terms of ray tracing with the help of ray reconstruction. (I'm not saying Nintendo's new hardware's going to be on par with the PlayStation 5 Pro in terms of ray tracing.)
Steve
Koopa
- Pronouns
- he/him
Was wondering, i've been hearing a lot of negative of 8NM, what would be the positive? just for a hypothetical if for some reason Nintendo decided to go for it.
Because we know the switch 2 is a 8 inch screen and that's pretty much it.
ItWasMeantToBe19
Manakete
It is cheaper per mm than TSMC 4N.
My guess is that the Switch 2 uses TSMC 7nm+ (called 6nm and the node that the PS5 Slim and probably Pro use)
- Pronouns
- He/Him
trying to discredit everything that's been found from the leaks as "fan fiction" by equating it to release timing theories is such a false equivalence that I'm sincerely wondering if you're doing a bit
necrolipe
Universo Nintendo Owner
- Pronouns
- he/him
And you'll be leading to error as anything that he said regarding Switch 2 up until now is specluation, or guess or was taken from this exactly thread finds over the last years
He absolutely knows nothing from his sources regarding Switch 2
MadMarvainy
Moblin
- Pronouns
- He/Him
Edit: typo fixesOk. You don't consider it such.
Others do. That was my entire point.
Your viewport of the world does not see the entire thing all at once, and nobody's can, but you can't make definitive statements about the markets, living environments and tastes of others purely based on your own window.
Size is a factor to LOTS of people, and Nintendo has intentionally marketed the Switch not only as a hybrid, but a home console and a handheld. Because it is one. It is a home console. Nintendo considers it one.
People MAY find a way to make a big console work. But then what's the point, aren't consoles meant to be the easy, don't have to think about it option? That doesn't track. It's just part of how Switch is viewed, used and marketed; its small size is appealing.
This reminds me of one of the biggest praises I've ever seen anyone give a console, it was Yahtzee Croshaw of Zero Punctuation ~ Fully Ramblomatic fame, when he said Nintendo Switch is the only console thst lives in his LIVING ROOM. Because it's small, so it's easy to set up, and has a wide variety of games! He's a notoriously HARSH critic, that is his brand, and one of his biggest compliments he's ever given a console is "easy to set up". A full time video game journalist, an office full of every variation of every relevant bit of hardware he could want. And what wins, at the end of the day? Convenient, easy to transport, easy to set
What are you going about? You've lost me... I'm not seeing anything just through my lens, I'm looking at numbers. Neither one of us speaks for Nintendo but I have a lot of doubts that most of the Switch's success has to do with its offerings as a home console unit based on their history of home-only consoles, it's a hybrid. No matter how someone decides to use it, it's a hybrid, that's why it's successful and it's the only one of its kind. If the Switch 2 wasn't a hybrid and only a home console that is small, they'd lose me at launch and might lose a lot of people.
How many people have become full-on devoted PC gamers because the PS5 is too big? Do we have anything supporting this? The Switch 2 is also probably going to be bigger, does this mean a ton of people will jump ship to the Deck or ROG? I don't think so. Some might but you might be taking "no one" far too literally. Even if it did occur, it wouldn't happen to a degree where Nintendo would rush back to formula to ship a smaller Switch2.
You name YouTube content creators like this means anything in regards to my original post on the matter. I'm talking about hard stats that tell me the PS5 is selling well in the West regardless of its size. While in Japan it isn't for obvious reasons. Some people wish it was smaller but it's not enough of a deal breaker to the mass market. If it was, it would've been a lot smaller out of the gate.
I don't get how your anecdotes are supposed to have reverence here when I can give you just as many if not more of people not caring. When someone says the PS5 Pro is probably gonna be so big that those who might want it might as well get a PC, it misses the point of why some people prefer to buy consoles in the first place.
Last edited:
Wait is this true? We know from information in the leak that the GPU consumes 4.2 watts and runs at 660MHz undocked?
ExoticSpice
Shriekbat
RDNA4 will be really interesting. We see AMD take RT seriously because of Sony and Microsoft. RDNA5/6 will further focus on RT and AI because the Next Gen needs even to even more powerful to get people to upgrade.
darthdiablo
+5 Death Stare
- Pronouns
- he/him
Okay? Even if that’s true (I’m skeptical), doesn’t mean he’s a reliable Nintendo leaker.
Steve
Koopa
- Pronouns
- he/him
wafers are cheaper, but you're also getting less product per chip
Ghostsonplanets
Ghost in the Sales
- Pronouns
- He/Him
It would be very cheap and thus we would be able to have a cheaper Switch 2 (Or Nintendo would have better margins at launch) or Nintendo could use the savings from the SoC manufacturing and spend them elsewhere (new gimmick, better screen, bigger battery, more RAM, etc).
darthdiablo
+5 Death Stare
- Pronouns
- he/him
Except 8nm isn't cheaper. Per wafer is cheaper, but you get lesser (poorer yields and lower density).
ItWasMeantToBe19
Manakete
We have no yield information since 2021 on Samsung 8nm.
Ghostsonplanets
Ghost in the Sales
- Pronouns
- He/Him
I was watching a DaveDoesStuff video about the RTX 2050 and one of the things that catched my eye was this:I was looking at some benchmark recently and saw that laptopmedia had covered quite a few benchmark/gameplay footages with various laptop GPUs (e.g. RTX 2050 @ 35W, RTX 3050 @ 65W-75W, RTX 4050 @ 45W).
Now, of course, there's some variance with the results due to the CPU being used and perhaps even memory clock speed having an impact. I checked most of the benchmarks and the laptops or spec's they use and I'd say they're relatively close in capabilities. Although the amount of metrics on display isn't giving much validity, I know they're a legit source. One note to make is that they may not have been made with the same drivers....
Nonetheless, perhaps for a layman's showcase of how node process/power efficiency/clock speed all play a role.
I put a side-by-side of Death Stranding, and it's quite impressive how the RTX 4050 performs, practically doubles compared to the RTX 2050, with only a 10W increase of TGP, while also leading the RTX 3050 that has a 20W higher power consumption. Now there could be dynamic boost being at play (and other variables), although I suspect these laptops are hitting their max-rated TGP; 35W, 65W, 45W for the RTX 2050, 3050, 4050 respectively.
ViewSync - Multiple YouTube Viewer
Streamlined multiple YouTube viewer with built-in support for content creators.
I guess ideally, you’d want to compare at the same clock frequency and then observe the power consumption - frametime, but for such an analysis you’d need to have each of these devices (and a vbios that allows to have similar ranges of frequencies). Also the RTX 4050 has more CUDA cores (2056), maybe the RTX 500 (2048 CUDA cores) can give some interesting metrics. but given the recent announcement of that chip, it’ll take a while
RTX 2050 [35W]
As per TPU; GeForce RTX 2050 35W = 982-1372 MHz
RTX 3050 [65W]
RTX 4050 [45W]
Which aren't nearly as bad for a 16 SM 8nm GPU at low clocks. Granted, we can't take the software reporting of the TDP at face value as we don't know what the hardware is actually informing, we don't know which parts of the hardware are active or not, etc. But such TDP and clock numbers on 8nm would be quite applicable for an even smaller GPU like GA10F/T239.
If you only pay for the working chips, yields barely matter. About yields, Samsung 8nm is a mature node with high yields at this point. They struggled with initial Ampere GPUs per rumors, but over time everything was resolved. The only point of comparison between SEC 8N and TSMC 4N that we had was SemiAnalysis claim that 4N had slightly better parametric yields while having the same catastrophic yields as 8N.
Density, yes. But T239 would still be a small chip due to having a small GPU. GA107 (20 SMs) is around 19x mm² die size on 8N.
Granted, I personally expect TSMC 4N. But I believe 8N might be possible on a 10W budget for portable.
darthdiablo
+5 Death Stare
- Pronouns
- he/him
Yeah - my view is that both 4N and SEC8N are the most likely candidates, with 4N being more likely if 12SMs claim is still true.
Just sharing that I'm not sure SEC8N is cheaper compared to TSMC 4N when considering how many chips one can get from those node processes.
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
SiG
Chain Chomp
So you're telling me MLID making claims about Nintendo switching back to AMD is credible?
omniryu
Like Like
- Pronouns
- He/him
Not if Sony makes it a time exclusive.
Funny thing, I was telling someone that earlier today.
redmutineer75
Like Like
- Pronouns
- He/Him
The problem is that 12 SM isn't a "claim", it was datamined directly from Nvidia's servers, and given that it was from a graphics API, there were no redundant SMs like with the PS5 and XSX GPUs. A 12 SM T239 is just the indisputable reality, and any speculation on node has to be made with that information in mind, otherwise it's completely meaningless.
Ghostsonplanets
Ghost in the Sales
- Pronouns
- He/Him
Pretty much. To be honest, I'm kinda past the node talk. I know Nintendo will deliver a device with reasonable battery life and weight. And we also know it can run the Matrix Awakens demo and BoTW at 4K60 (DLSS). So the node is basically academic at this point. Whichever it is, Nintendo will be able to get their performance, area and energy objectives from it.
omniryu
Like Like
- Pronouns
- He/him
MLID pretty much said that he believes that the Switch in portable mode will be around stream deck and below the Series S in dock mode.... the same thing everyone concluded around here.
Also, when you read the post his source actually says. They think Nintendo will go with Samsung 8 nm. Not that they will. So MLID went "Because Nintendo" and say they will have 8 nm.
Last edited:
omniryu
Like Like
- Pronouns
- He/him
I think it is ok for anyone here to say they don't understand the technical discussion here. While I understand stuff like ray tracing, the BVH, and other software related stuff sure I can hang. When it comes to power draw and the lithographic process, I don't understand.
@Vash_the_Stampede just because you do know doesn't mean it is hopium.
Steve
Koopa
- Pronouns
- he/him
Hey man, if we're able to boost most of Nintendo first party games at 4k 60fps (DLSS), that would be hella nice and i can personality see it being part of the marketing. Like the switch library is probably one of Nintendo best and people are still paying full price for these games.
let it be backward compatibility and we're golden.
Like the game sells speak to themself and reintroducing Nintendo select for Switch 1 titles (Eshop exclusive and not phusical sadly) would be the cherry on top (a man can cope)
so having DLSS, RT, good battery life, Good backward compatibility and lastly good performance then we have a success story in the making.
OP
OP
I think Nvidia has much more incentive to use TSMC's 4N process node to fabricate T239 than Nintendo does, because considering that I imagine Nintendo has exclusivity towards the highest binned T239s to begin with, I think Nvidia wants to avoid a situation of having a surplus of lower binned T239s. And I think Nvidia's going to have a much higher surplus of lower binned T239s with Samsung's 8N process node vs TSMC's 4N process node.
SiG
Chain Chomp
There's also the fact that 4N would still produce higher yields (due to its production also maturing) and Nintendo wouldn't want to encounter a shortage of units to sell again after what happened with the Switch wen it first launch. 4N fabrication would at least have long term support.
Z0m3le
Bob-omb
MLID is saying that Switch 2 is 3 to 4 TFLOPs, just like us. being hung up on 8nm, is a fools erran, it literally doesn't matter what size the transistors are, they only tell us the possible performance, which we've pegged at 3 to 4 TFLOPs for years. MLID didn't disprove anything in this thread, he is saying it can be done on 8nm, which I personally think is stupid, but it doesn't conflict with my performance expectations at all.
Yeah, the DLSS test was done with that clock and the clock's test was named "4.2w" Drake didn't exist for another ~9 months though, so these are just targets.
Last edited:
- Pronouns
- he / him
For years my thought on Dragon Quest XII is Switch 2 is no question, the only question is whether they're confident enough to skip Switch 1. With the exception of V being a relatively early SNES game, Dragon Quest has really only arrived on mature platforms with high userbase (plus things like PS4 or MSX adding on), even if that machine's successor is already out. Something like a PS5+Switch 2 release in 2025 or 2026 might have limited sales potential (in Japan) versus VIII launching to a userbase of 15 million, let alone compared to IX launching to a userbase of 30 million, or XI launching to userbases of 23+5 million.
Positive: More room for improvement in the future when they decide to go Not 8nm.
I think that's referring to some data about DLSS tests that were in the hack. They're without much context, but it's easy to speculate that the specific things tested weren't totally arbitrary.
Kise Ryota
Bob-omb
I prefer to expect SEC 8nm with Switch 1 clocks, that way I can't be disappointed (I think lol)
bloopland33
Like Like
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
Z0m3le
Bob-omb
That is against both MLID and this thread. MLID is saying the same performance we are. 3TFLOPs to 4TFLOPs. He just thinks Nvidia broke physics.
robertman2
Chain Chomp
- Pronouns
- He/him
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
Aufhebung RPG
History is tragedy becoming comedy
So what other information besides Nvidia's leaked data is there to prove it's not 8nm
Z0m3le
Bob-omb
No one cares what the node is. You just aren't understanding. The point of discussing the node is to try and figure out the performance of Switch 2, it would limit possible clocks and thus TFLOPs. MLID and this thread agree that the TFLOPs performance range is 3TFLOPs to 4TFLOPs (The DLSS test has a 4.24TFLOPs max clock that is in the leak). There is no conflict here. The debate about process node is to be had by people who do not understand that it is meaningless.
NineTailSage
Bob-omb
I was watching a DaveDoesStuff video about the RTX 2050 and one of the things that catched my eye was this:
Which aren't nearly as bad for a 16 SM 8nm GPU at low clocks. Granted, we can't take the software reporting of the TDP at face value as we don't know what the hardware is actually informing, we don't know which parts of the hardware are active or not, etc. But such TDP and clock numbers on 8nm would be quite applicable for an even smaller GPU like GA10F/T239.
If you only pay for the working chips, yields barely matter. About yields, Samsung 8nm is a mature node with high yields at this point. They struggled with initial Ampere GPUs per rumors, but over time everything was resolved. The only point of comparison between SEC 8N and TSMC 4N that we had was SemiAnalysis claim that 4N had slightly better parametric yields while having the same catastrophic yields as 8N.
Density, yes. But T239 would still be a small chip due to having a small GPU. GA107 (20 SMs) is around 19x mm² die size on 8N.
Granted, I personally expect TSMC 4N. But I believe 8N might be possible on a 10W budget for portable.
We might not be able to take the software at face value for reporting TDP but 12SM for T239 isn't far off from the RTX 2050 16SM in scope.
We can't attribute 6-7 watts for the GPU, this device would be 15W total system draw in handheld (it would have to be as large as SD)
I think that's the problem a lot of us have, it's not that 8nm can't be a possibility. It's that leakers are saying this device is 2-4Tflops between handheld and docked modes, while being slightly bigger than the current Switch and better battery life than the Steamdeck.
This sounds like wizardry to any of us that know better, because 8nm is 25-30% less efficient than TSMC 7nm and a 12SM GPU on that specific node is fairly large in comparison to the SD 8cu GPU.
Nvidia has pre-paid for so much fabrication at TSMC 4nm capacity that rumors are they have halted production to try and clear stock in order to have demand for the 50 series cards that are expected to come out later this year. My belief for one of the better reasons to put T239 on 4nm, is Nvidia are killing their premium value in the market by over saturating production. I believe on the new 50 series cards they might roll out everything much slower then they did with Lovelace.
At this point I'm more excited to hear more details about the new gamecard medium and internal storage.
Last edited:
