• 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| (New Staff Post, Please read)

Reader mode
LoneRanger said:
Capcom wants RE to be an annual IP, one entry per CY.
2019: 2 Remake
2020: 3 Remake
2021: Village
2022: nothing
2023: 4 Remake

Its strange they didn’t release anything in 2022. As you said, maybe their plan was to release a Switch-bound (temporal) exclusive RE title very near the launch of “Pro” model, “Pro” was cancelled so probably that title got delayed I guess?

What about next year? 2024, we don’t know, maybe RE9? Or it can it be the long rumoured Switch RE title? Either as a cross-gen game (with how well Rise looks, I think Capcom can make a good job with RE and don’t lose those +130 million users) or as a next HW exclusive…
Click to expand...
RE9 Its Not So close, RE5 Remake Probably also, so maybe Switch title
 
 
Last edited:
 


Thought this was an interesting look at a Switch like device, with a camera and 4 AR sensors on the tablet body, and these 'magnetic' joy-cons with analog triggers.
 
Simba1 said:
And yet, Switch gets quite hot.
Click to expand...
You mean the system that regardless of what happens when playing BOTW, Splatoon 3, Mario Odyssey, Pokémon Scarlet and Violet, Witcher 3, Dark Souls Remastered, etc., never throttles? Even when it’s over clocked it doesn’t throttle, that system? The system that has those 256 shaders overclocked very high for its process node, and that same system that does not throttle? I’m sorry but like, I don’t think the switch is actually a very good comparison anyways, considering that it’s fan isn’t even operating anywhere close to full tilt like I mentioned before.


and on a significantly worse node that gets really hot than what Drake would be on anyway! whether 8, 7 or 5nm, it isn’t going to be a 20nm.
Simba1 said:
I am not so familiar, can you give some examples of phones that regularly run GPU at around 2TF?
Thing with phones is that hardware is usually made that has smaller speed bursts than having hole hardware running all the time on max clock frequencies,
while in case of Switch we talking about about clocks that will run Switch same time when you gaming.
Click to expand...
Blomqvist said:
Snapdragon 8 gen 2 peaks at 3.5 TFlops FP32 and that is probably at 8 watts. It can't sustain it for nothing more than short bursts but even throttled to 1/3, it would be Xbox one level of performance. With active cooling and 5 nm process nVidia is able to deliver 3+ TFlops in docked Switch form factor.
Click to expand...
I’m going to reply to these two actually at once here, I’m not sure if you guys know this, but the phones have actually been changed a lot over the past couple of years where the curve has actually shifted in how much performance they can give and at much how much power draw and heat.

There’s a theoretical peak performance, and then there is a sustained performance. Previously, phones actually dramatically lowered their sustained performance to maintain a perfectly “acceptable” temperature and battery life without, you know, dying fast.

However, over the course of years, companies such as Apple and QUALCOMM have actually shifted into being able to deliver better sustained performance than just better Peak performance. For example, the QUALCOMM snapdragon’s as of late don’t throttle to 1/3 performance, they lose up to 1/3 performance, and they can sustain 2/3 of the theoretical paper performance.

Apple has also been able to deliver better sustained loads at a lower power draw, as they’ve shifted the curve. It’s not 2017 anymore, it’s 2023 :p.


Now, take into consideration that these are passively cooled devices, and they aren’t tethered to a wall. One aspect of the switch that really benefits it is that it can be tethered to a wall relinquishing the battery of having to ask bell that much energy to charge it, and it’s going to actually be cooled in a competent fashion with a fan(maybe some shade at apple 🤭). If phones that clock higher are able to sustain higher performance numbers without any of this, an actively cooled system that is tethered to a wall should be able to actually sustain that.


Mind you, in my opinion the peak performance of this device can actually hit in portable mode is 1.6 TF. And the peak in docked mode is 3.2 TF.

As in, that’s the highest they can reach it to me, not that they’ll actually hit it


Actually, let me just change the perspective of this, no one is asking for a device that can clock to 1.5 to 2 GHz for the GPU like the XBox Seires X, S, PS5 or desktop GPUs. This is about a device that clocks from 500 MHz in portable mode to 1 GHz or 1.1 GHz in docked mode.


This can be a wider, but slower GPU that reaches a higher theoretical peak, just like how the Xbox series X has a slower GPU clock than a PlayStation five, but it still faster in the theoretical sense of TF.


We focus too much on TF anyway….
 
Kazuyamishima said:
Yeah I know that's why I thought most people were saying that the clocks wouldn't be so much different from the first switch because of this reason
Click to expand...

Its my opinion of course, but its based on looking at the specs for Drake and my assumption that the Switch form factor will remain the same. Drake on 8nm is going to be much larger than the Tegra X1 and draw more power at the same clock speeds. 4N is nearly three times as dense as 8nm and far more power efficient. So when people talk about it being a 3Tflop machine, that likely doesn't work with 8nm since it would require a clock speed of 1Ghz. When you look at RTX30 laptop graphics cards with similar specs, they are pulling 35-45 watts at similar clock speeds. Could it be 8nm? Yes, but clock speeds would have to be low. Again, this is under the context that the Switch form factor remains the same size or at least very close and power draw needs to be similar. Between the die size, clock speeds and power consumption/thermals, 4N basically makes the problems go away. The die size gets smaller, clock speeds can go up and still pull less power compared to 8nm and thus heat is less of an issue. If you try and find a mobile based SOC on 8nm with specs anywhere in the ballpark of Drake, you wont find it.
 
Goodtwin said:
Its my opinion of course, but its based on looking at the specs for Drake and my assumption that the Switch form factor will remain the same. Drake on 8nm is going to be much larger than the Tegra X1 and draw more power at the same clock speeds. 4N is nearly three times as dense as 8nm and far more power efficient. So when people talk about it being a 3Tflop machine, that likely doesn't work with 8nm since it would require a clock speed of 1Ghz. When you look at RTX30 laptop graphics cards with similar specs, they are pulling 35-45 watts at similar clock speeds. Could it be 8nm? Yes, but clock speeds would have to be low. Again, this is under the context that the Switch form factor remains the same size or at least very close and power draw needs to be similar. Between the die size, clock speeds and power consumption/thermals, 4N basically makes the problems go away. The die size gets smaller, clock speeds can go up and still pull less power compared to 8nm and thus heat is less of an issue. If you try and find a mobile based SOC on 8nm with specs anywhere in the ballpark of Drake, you wont find it.
Click to expand...
I thought when people were mentioning 3Tflop that was just a theoretical max and not what people actually thought that it was going to run at similar to how the TX1 is severely down locked in the switch
 
Gerald said:
RE9 Its Not So close, RE5 Remake Probably also, so maybe Switch title
Click to expand...
I wouldn’t be surprise (and feet a little natural) they decided to remake 1 and 0 before 5, 2026 its series 30th anniversary and a remake of 1its something pausible i think, then next year a remake of 0. Both with a third person perspective like 2/3 remake.

I think RE9 is going to be a 2025 title with a 4 years development time from Village like VII to Village.

That leaves 2024 as a possible empty year, unless….

Regarding actual Switch HW discussion, I will wait until early may.
 
Last edited:
Apologies if this is rehashing info that’s been discussed for years but I thought there was speculation that Drake could have been redesigned from 8nm to a 5nm node, thereby solving some of the supposed power draw/clock concerns. Based on the leaked info and the speculated testing being done on the chip, is that possible or is consensus now that a node change is unlikely?
 
Klax0nKlax0ff said:
Apologies if this is rehashing info that’s been discussed for years but I thought there was speculation that Drake could have been redesigned from 8nm to a 5nm node, thereby solving some of the supposed power draw/clock concerns. Based on the leaked info and the speculated testing being done on the chip, is that possible or is consensus now that a node change is unlikely?
Click to expand...
if there were power concerns, this would have been known before they started testing chips. that's what simulations are for
 
Kazuyamishima said:
I thought when people were mentioning 3Tflop that was just a theoretical max and not what people actually thought that it was going to run at similar to how the TX1 is severely down locked in the switch
Click to expand...

No, there is an expectation by many that docked mode will offer 3Tflop of performance, but again, this requires a clock speed north of 1Ghz. I do not believe Drake at 8nm achieve this and fit within the power/thermal constraints of the Switch form factor.
 
Kazuyamishima said:
I thought when people were mentioning 3Tflop that was just a theoretical max and not what people actually thought that it was going to run at similar to how the TX1 is severely down locked in the switch
Click to expand...
The theoretical max for Drake would be 4tflop, since the Orin chip apparently maxes out at 1.3ghz. A similar downclock to the Switch(~25% off the max GPU clock in docked mode) would put you at around 1ghz, which would be 3tflops for Drake in docked mode. Definitely viable on a good node. If it's 8nm though then probably not.
 
Bosintang said:
I do not think that Doom is a great port. It's a serviceable one.
Click to expand...

Doom is a very good early port for Switch. Doom uses a lot of post processing effects that are really memory bandwidth intensive. Not all games hit the hardware the same way and honestly ID Tech is generally a well optimized games engine, so beyond dialing things way back, there probably wasn't a ton of performance to find in further optimizations. Doom and Doom Eternal are a couple of the better looking games on PS4/X1 and they still manage a solid 60fps. PS4 does a good job of maintaining 1080p most of the time, but Xbox One regularly drops to 828p. PS4 has much better memory performance compared to Xbox One, but Xbox One has much better memory performance compared to Switch. When you compare the Switch port of Doom to the Xbox One build, its not hard to see why they had to make the compromises they did for Switch.
 
Modded Mariko would crush DOOM. Unfortunately could only find an Erista overclock video.



Edit: Found Eternal 60 FPS mod
 
Dakhil said:
None of the information in the video is new.
Click to expand...

Maybe its not new to this forum, but its new to most gamers outside of this forum. The narrative that Switch Redacted isn't coming this year has been pushed and mostly accepted on YouTube, but when more and more of these talking heads see information like this come out, they start to question if the narrative is correct, and suddenly when you realize the processor was finished up in 2022, Nintendo spent a ton of money acquiring raw materials, and the second half lineup for Switch looks strikingly sparse, its easy to warm up to the idea that may new hardware is coming late this year.
 
Goodtwin said:
Its my opinion of course, but its based on looking at the specs for Drake and my assumption that the Switch form factor will remain the same. Drake on 8nm is going to be much larger than the Tegra X1 and draw more power at the same clock speeds. 4N is nearly three times as dense as 8nm and far more power efficient. So when people talk about it being a 3Tflop machine, that likely doesn't work with 8nm since it would require a clock speed of 1Ghz. When you look at RTX30 laptop graphics cards with similar specs, they are pulling 35-45 watts at similar clock speeds. Could it be 8nm? Yes, but clock speeds would have to be low. Again, this is under the context that the Switch form factor remains the same size or at least very close and power draw needs to be similar. Between the die size, clock speeds and power consumption/thermals, 4N basically makes the problems go away. The die size gets smaller, clock speeds can go up and still pull less power compared to 8nm and thus heat is less of an issue. If you try and find a mobile based SOC on 8nm with specs anywhere in the ballpark of Drake, you wont find it.
Click to expand...
I just want to mention that I don't think die size is much of a relevant consideration when we talk about which process node nintendo will choose for the drake. They could literally double the die size from the Erista and that'd only translate to an additional 5mm increase (or 1/5th of an inch for the americans) in chasis size in each direction. This might be a consideration for smaller form factors such as phones, but for the switch that'd only correspond to a 3% increase in length and 5% increase in height. A larger die would also benefit from increases in passive cooling so there also is much more to the story with regards to heat. I imagine the only major consideration for process node selection is going to be power consumption and the associated battery life implications.
 
Kazuyamishima said:
Doesn't switch run doom eternal at like 360p handheld?
Click to expand...
600p 30fps for docked mode I mean.

Bosintang said:
I do not think that Doom is a great port. It's a serviceable one.
Dying Light, Witcher 3, Nier Automata... Those are really great ports.
Click to expand...
I do think Doom is kind of a technical marvel on its own and I think it's fairly optimized. Not an impossible port though. Just one of the most extreme ports in which we see Switch's bottlenecks widen the gap further than we might normally expect when we look and compare on paper specs.

Serif said:
Modded Mariko would crush DOOM. Unfortunately could only find an Erista overclock video.



Edit: Found Eternal 60 FPS mod
Click to expand...

What's the resolution, and clock speeds they used?
 
Last edited:
space-coyote said:
I just want to mention that I don't think die size is much of a relevant consideration when we talk about which process node nintendo will choose for the drake. They could literally double the die size from the Erista and that'd only translate to an additional 5mm increase (or 1/5th of an inch for the americans) in chasis size in each direction. This might be a consideration for smaller form factors such as phones, but for the switch that'd only correspond to a 3% increase in length and 5% increase in height. A larger die would also benefit from increases in passive cooling so there also is much more to the story with regards to heat. I imagine the only major consideration for process node selection is going to be power consumption and the associated battery life implications.
Click to expand...

Honest question, how is that calculated? When I see numbers that show 5nm being 3 times as dense as 8nm, I expect a bigger reduction in die size.

Edit: Or are you saying the chassis that holds the SOC could only need to be 5mm bigger than what they used for Erista?
 
0
I will say that it would be hilarious if the one year new hardware isn't firmly ingrained in the zeitgeist it finally happens
 
Just a minor point to the Doctre video. The claim is that the person working on T239 worked for a firm that specializies in taping out SoCs. Their work history with said firm is to March 2022. There's still wiggle room he left the company before completing work but Doctre is assuming the task was completed before they left.

This doesn't necessarily mean tape out was done in Mar 2022, but if T239 was in the process of being taped out around MARCH 2022, even if this person left the project, The SoC would have finished sometime in 2022, safe to say.

Putting on the doomer hat:
The question of course is how this meshes in with the cancelled revision talk. As far as I can gather, @NateDrake 's info on the cancelled revision was around this time last year when he suddendly had new info he needed to verify which was also around the same time as the nvidia hack last year (hack was Feb 25,2022) So is it possible T239 was cancelled during tapeout?
 
Last edited:
Raccoon said:
I will say that it would be hilarious if the one year new hardware isn't firmly ingrained in the zeitgeist it finally happens
Click to expand...

funeral-prank.gif
 
Question regarding backwards compatibility, if Nintendo went the route of including a TX1 in the Switch 2 so when you play a Switch game the system reverts into ‘OG Switch mode’ would that then mean any type of boost mode or upgrade patches for Switch games would be impossible?
 
Bosintang said:
So was the first version of Ark I guess.
Click to expand...
Funny you bring that up, because Ark got ported again and it turned out to be a fantastic port this time.

The quality of the port matters a lot on Switch. Panic Button's efforts are okay, but meager when compared to some other developers who work much closer to the metal, such as Feral Interactive and Iron Galaxy.

If Doom 2016 were to get reported again, they have Metroid Prime Remastered as their benchmarking standard. I don't mind if some of the lighting has to be baked, but clearly bespoke porting does wonders in closing in the gap of percieved performance benefits.

tl;dr - Raw numbers isn't as important as percieved power.
 
ITSMILNER said:
Question regarding backwards compatibility, if Nintendo went the route of including a TX1 in the Switch 2 so when you play a Switch game the system reverts into ‘OG Switch mode’ would that then mean any type of boost mode or upgrade patches for Switch games would be impossible?
Click to expand...
When the OS starts a game, if its metadata says it's a regular Switch game, then yes, it runs in "OG Switch Mode".
An upgrade patch would upgrade the metadata as well saying to run it on full-on successor mode.
 
Couldn't there be a boost mode for an onboard TX1+ by telling it run at higher clocks like with homebrew ?
 
Dekuman said:
So is it possible T239 was cancelled during tapeout?
Click to expand...

Maybe, but its hard to make a great argument for why they would do that. That's years of R&D money down the toilet and back to the drawing board. When you are that far along and have nothing else in the pipeline, why wouldn't you go through with the plan and release the Pro model rather than cancel and have nothing new for the foreseeable future? Even if the Pro model only sold 15-20 million units, that wouldn't be terrible for a Pro model and much better than recouping zero of your R&D dollars. At some point it needs to be considered that perhaps the Pro rumors were just conflating leaks that slip out that were actually related to the Switch successor. People speculated that a Switch Pro would exist from the day Switch launched, so its not hard to believe that people were misguided when little nuggets of info did get about about a new processor for Nintendo from Nvidia that was a lot more powerful than the X1.
 
space-coyote said:
I just want to mention that I don't think die size is much of a relevant consideration when we talk about which process node nintendo will choose for the drake. They could literally double the die size from the Erista and that'd only translate to an additional 5mm increase (or 1/5th of an inch for the americans) in chasis size in each direction. This might be a consideration for smaller form factors such as phones, but for the switch that'd only correspond to a 3% increase in length and 5% increase in height. A larger die would also benefit from increases in passive cooling so there also is much more to the story with regards to heat. I imagine the only major consideration for process node selection is going to be power consumption and the associated battery life implications.
Click to expand...
Height increases?

Body width increases?

So we know it's not 8nm. They'll abandon Joy-Con compatibility when the sun implodes.
 
kvetcha said:
neither of those would preclude JoyCon rails being the same size
Click to expand...
Yeah, but it would make them an aesthetic and proobably ergonomic nightmare while possibly breaking compatibility with cases, non-standard Joy-Con shapes, etc.

I can't see them changing the basic dimensions of this device beyond maybe 1-2mm wider, like OLED Model was. Multiple millimetres in multiple dimensions across millions of devices is a LOT of material they could save by making it smaller, no matter what that takes. We've already seen them miniaturise the motherboard significantly, and all Switches currently have an extremely overprovisioned cooling system. Even in a worst case scenario 8nm chip, there would be no reason for it, and plenty of reason against it.

Plus if we talk more about it getting "bigger", @Raccoon will lose the plot. Rightfully so, I'd say.
 
Serif said:


Thought this was an interesting look at a Switch like device, with a camera and 4 AR sensors on the tablet body, and these 'magnetic' joy-cons with analog triggers.
Click to expand...

Noice.

How does the Snapdragon chipset compare to the one rumoured to be used in Switch 2?
 
ITSMILNER said:
Question regarding backwards compatibility, if Nintendo went the route of including a TX1 in the Switch 2 so when you play a Switch game the system reverts into ‘OG Switch mode’ would that then mean any type of boost mode or upgrade patches for Switch games would be impossible?
Click to expand...
No matter the solution, patched games will essentially be Switch 2 ports, even if unoptimized, so they can benefit from the extra power just from that.

Anyone correct me if I'm wrong, but none of previous BC were done by putting the entire chipset there, they just integrate necessary parts to the new chipset to act identical to the previous console. But in theory, it they put an entire X1, unpatched games are very unlikely to perform any better.

A more common option would be adding TX1's GPU to Drake, which would means a much better CPU, reaching the fps cap in many games, better loading times and a substantial boost to battery life in BC mode, but resolution would likely be the same.

NVidia can also customize Drake GPU to support all instructions present in the TX1 GPU. This would be the best for unpatched games, as they will perform better, including max dynamic resolution, and have amazing battery life.

Finally, there's emulation or translation layer. That have a headroom though and I'm not sure how much performance and battery life would be lost compared to the above, but I'm very confident it's not to the point Nvidia can't do it, even with conservative specs.

The thing is, Myamoto said it was easier than ever to bring games forwards. I doubt he would say this if Nintendo was going to just use one of these hardware solutions like they did in the past. If you believe Nintendo doesn't cares about losing thousands of third party games, you can join MVG theory they could patch 1st party games and call it a day, but I don't - they fought really hard to get where they're Today to just cheapen out there.
 
Last edited:
For my simple mind, can someone make a list, from best to worst, of the process nodes that Drake could be on? My mind reels with all the 4N and N4 and Samsung this and TSMC that… I’m assuming the list would start with 3nm TSMC being best (but unlikely), and 8nm Samsung being worst, right? Where do all the 4Ns and N4s fit between there?
 
Klax0nKlax0ff said:
Based on the leaked info and the speculated testing being done on the chip, is that possible or is consensus now that a node change is unlikely?
Click to expand...
There is no consensus on the node. The Orin numbers are… pessimistic but not definitive and there are at least two technologies that we know reduce battery life on Drake that Orin does not have.

There is no consensus on the journey to pick the node either. I’ve speculated about a circuitous route to a node change relatively late in the T239 process but that’s just a theory and only interesting to me because I dig the history of the thing.

There is pretty strong consensus about battery life and performance targets. Nvidia put 6x the GPU cores and 2x the CPU cores in there - CPU cores which have a ~3x perf gain clock per clock, so 6x.

They’re not going to vastly underclock that 6x number for battery reasons, they’d be better off making a smaller chip, something they would have known years ago. I doubt they’re going to hugely overshoot it either. An 8x leap would be a fucking blast! 10x though? I can buy a Series S if that’s what I want. I bought a Switch for Zelda and for handheld mode

Somewhere in that 6-8 range is the sweet spot, and the node just shifts that one direction or another.
 
0
Reader mode
You must log in or register to reply here.
Please read this new, consolidated staff post before posting.

Furthermore, according to this follow-up post, all off-topic chat will be moderated.
Last edited by a moderator:


Back
Top Bottom