• 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)

There were rumours 3D was bolted on relatively late, after SoC had already been selected, and it impacted battery life. My own assumption is it likely also impacted performance, as clearly there is a performance cost to 3D.
Not true. 3D was chosen quite early in the development of the machine. Nothing bolted on (Specially because the custom 3D screen solution was quite novel).
The original selection was apparently nvidia Tegra 2 but the chip was too power hungry so Nintendo went with another custom design, which had an underpowered CPU and a proprietary GPU that while powerful was sort of a turnkey design from 2005 and isn't not like there's a v2 of the DMP PICA, it seemed like a one and done thing so there isn't even a clean roadmap to a 3DS successor BC outside of just including the SoC in the device.
The OG SoC was meant to be a Tegra MCM or Tegra 2. But rumors say that it failed due to Nvidia not meeting Nintendo standards for power efficiency.

The PICA 200 design wasn't the 2005 one but rather the revised 2008 with bigger fillrate and Mpixel/s. The CPU IP, while underpowered, would probably be the same be it Tegra or CTR that was in the device.

There was a roadmap for 3DS successor from DMP. They even had solutions that were BC with 3DS GPU IP and were researching things like real time Global Illumination at low cost with fixed shader for possible future Nintendo Portable contracts.

It's just that, as we now know, Nintendo was thinking of reusing Wii U GPU IP into the 3DS successor and bolting on the 3DS SoC for 3DS BC. Due to the lack of contract wins, DMP downsized and turned their efforts into niche applications. But they were still rumored to be into a contract with Nintendo even at Switch era, before the confirmation of Nvidia Tegra solution being used.
 
Isn't the whole point of a CPU to do generalist tasks? That doesn't really fit with fixed functionality. And as far as I can tell, the decompression blocks in the PS5 and Xbox Series chips are on the I/O complex, not the CPU complex.

I mean, yeah, but CPUs also seem to have stagnated at being able to do directional graphs much better and I don't think GPUs can do that at all.
 
When do you think realistically AMD will have a shot at getting Nintendo business again?
Who knows. Nintendo is free to choose their vendor, but Nvidia pros outweigh the cons. Besides, Nintendo has been wanting to work with Nvidia for the longest time.
So MLID rumour of AMD trying to make a deal with Nintendo isn't true right?
Nobody knows. Maybe AMD pitched something to Nintendo and was rejected. Soundwave SoC, however, isn't a repurposed Nintendo design but rather another MS custom order, like Van Gogh/Aerith.
 
When do you think realistically AMD will have a shot at getting Nintendo business again?
3 years from now.

I think Nintendo is likely to genuinely consider AMD anytime they're looking at new hardware. The Nvidia partnership is currently very strong, but Ko Shiota, head of the hardware team, has close relationships with AMD, having worked very closely with the silicon engineers on the Wii and the Wii U (and likely, but less clearly, the GameCube).

AMD might have been able to offer Nintendo better raw performance this time around. But it couldn't deliver DLSS, it couldn't deliver high ray tracing performance, and it would have had a significantly harder time offering backwards compatibility, or the sort of software investment that Nvidia has.

But AMD is actively trying to close the feature/technology gap with Nvidia, they have a much stronger custom chip division (where Nvidia's is essentially Just Nintendo), and they're hungry for it in a way that, by the third generation, Nvidia might not be.

AMD could come in and say, look, we've closed the gap on AI, on upscaling, on frame generation, on RT. But we're ahead of Nvidia on chiplets by a huge margin - we can integrate WiFi, Bluetooth, cartridge readers, your whole motherboard into a single SOC. We've got custom ARM cores we've been building on top of Zen's execution pipeline. We've moved to new process nodes faster than Nvidia, and are willing to give you chips at a razor thin margin.

Maybe it's only a 25% chance they pull that off, but that ain't 0%.
 
It just feels same-ish. The 3DS and Switch 2 are both following up extremely extremely successful paradigm shifting platforms and attempting to iterate off of them. If rumors are to be believed, Switch 2 "could" also have a high price tag just like 3DS did at its time.

I've always felt like the switch 2 is a trap waiting for Nintendo and they should proceed carefully with the lessons learned from the 3DS in mind. There is one big advantage that the switch 2 will have over 3DS.

Unified software development teams. The most underrated, biggest, and best game changer to how Nintendo operates. With time, It will prove to be better than the blue ocean strategy pivot Iwata did /hot take
It’s not often I can talk with Nintendo employees (When I go to Japan for events) and when I do I have to be careful and they have to be careful but sometimes we can get some interesting tidbits about past events. The 3DS being a gut punch is something they’ve all expressed as if it was a company wide shock.

Another interesting one is apparently back when the Wii-u was winding down. Apparently there was a company wide initiative to allow for more younger developers to take an active role in idea creation and overall game direction. I vaguely remember this maybe being in an Iwata asks segment or something but they took it very seriously and that was viewed as a major success as the years went on. So much so the employees I spoke with were thrilled about it.



He might have. I have no clue who planned it specifically or when but it’s like a big fruit tree that just keeps on giving. I don’t think many will argue with me when I say the switch had some of the highest quality software in Nintendo’s history. Even with a global pandemic. I feel this is a big reason for that and what I mentioned in the above quote about younger developers being allowed to take more of an active role in game direction.
Thank you for sharing this. Even thought Nintendo had some rough years following the success of the Wii, the long-term effects on what they began to do paid off by the time the Switch was released.
 
Not true. 3D was chosen quite early in the development of the machine. Nothing bolted on (Specially because the custom 3D screen solution was quite novel).

The OG SoC was meant to be a Tegra MCM or Tegra 2. But rumors say that it failed due to Nvidia not meeting Nintendo standards for power efficiency.

The PICA 200 design wasn't the 2005 one but rather the revised 2008 with bigger fillrate and Mpixel/s. The CPU IP, while underpowered, would probably be the same be it Tegra or CTR that was in the device.

There was a roadmap for 3DS successor from DMP. They even had solutions that were BC with 3DS GPU IP and were researching things like real time Global Illumination at low cost with fixed shader for possible future Nintendo Portable contracts.

It's just that, as we now know, Nintendo was thinking of reusing Wii U GPU IP into the 3DS successor and bolting on the 3DS SoC for 3DS BC. Due to the lack of contract wins, DMP downsized and turned their efforts into niche applications. But they were still rumored to be into a contract with Nintendo even at Switch era, before the confirmation of Nvidia Tegra solution being used.
The initial 3DS SoC was more of a Tegra predecessor, the lead designer of which's next project was the first Tegra.
 
eighting made a unreal 5 tech demo

25-1024x576.jpg

12-1024x576.jpg

22-1024x576.jpg


pikmin 5 gonna look amazing lol
They should turn that into a real game tbh
 
Not true. 3D was chosen quite early in the development of the machine. Nothing bolted on (Specially because the custom 3D screen solution was quite novel).

The OG SoC was meant to be a Tegra MCM or Tegra 2. But rumors say that it failed due to Nvidia not meeting Nintendo standards for power efficiency.

The PICA 200 design wasn't the 2005 one but rather the revised 2008 with bigger fillrate and Mpixel/s. The CPU IP, while underpowered, would probably be the same be it Tegra or CTR that was in the device.

There was a roadmap for 3DS successor from DMP. They even had solutions that were BC with 3DS GPU IP and were researching things like real time Global Illumination at low cost with fixed shader for possible future Nintendo Portable contracts.

It's just that, as we now know, Nintendo was thinking of reusing Wii U GPU IP into the 3DS successor and bolting on the 3DS SoC for 3DS BC. Due to the lack of contract wins, DMP downsized and turned their efforts into niche applications. But they were still rumored to be into a contract with Nintendo even at Switch era, before the confirmation of Nvidia Tegra solution being used.

Just to add to this but Nintendo was experimenting (seriously, not just tinkering) with 3D back in the GameCube era as an attachment.

And uh...while we are on the topic....

The GC-3D thing sounds exactly about Nintendo and (Non-cardboard) AR/VR/XR. I mentioned this on here awhile ago. Major grain of salt and/or "noshitsherlock.


Thank you for sharing this. Even thought Nintendo had some rough years following the success of the Wii, the long-term effects on what they began to do paid off by the time the Switch was released.

(y). I get cool nuggets every now and then. Anything about the future is super spotty/Limited but I do learn cool things about past events by actually talking with investors/employees in Japan.
 
Probably when they offer to build a soc with so damn good price/ performance ratio that it outweighs all the downsides.
Hardware is only one half of the equation. Software is the other half of the equation, which is as important as hardware. Nvidia advertised developing the NVN API after the Nintendo Switch was announced. And I think NVN is one of the key aspects for the Nintendo Switch's third party support being quite good.

I imagine AMD also has to offer Nintendo a hardware and software package that's comparable to what Nvidia offers to have Nintendo seriously consider going back to AMD. Only time can tell if AMD can do so.

I can't see the future, but Nintendo probably value Nvidia partnership when they were at their lowest point and Nintendo using a cutsom made chip from Nvidia is quite telling.
Also helps that competitors probably couldn't offer a hardware and software package comparable to Nvidia's offering when T239 probably started development in 2019.
 
3DS did not fail because of its hardware approach. Sure the 3D gimmick may have limited appeal, but the machine was solidly more powerful and that was a good foundation for success. The 3DS launch failed because of software strategy and continued belief in being the casual gaming alternative, when that window was closing and people were hungry for full gaming experiences on the go. The pivot Nintendo did with 3DS was very successful and something we don't see pulled off too often. Focusing on full course gaming experiences, rich 3D games, 3rd party 'hardcore' franchises like Monster Hunter and other RPGs, made it one of Nintendo's better libraries - and sales picked up. The lesson learned was not to avoid similari-ish hardware following up a success, but that Nintendo's core appeal and fanbase should not be forgotten.
 
3DS did not fail because of its hardware approach. Sure the 3D gimmick may have limited appeal, but the machine was solidly more powerful and that was a good foundation for success. The 3DS launch failed because of software strategy and continued belief in being the casual gaming alternative, when that window was closing and people were hungry for full gaming experiences on the go. The pivot Nintendo did with 3DS was very successful and something we don't see pulled off too often. Focusing on full course gaming experiences, rich 3D games, 3rd party 'hardcore' franchises like Monster Hunter and other RPGs, made it one of Nintendo's better libraries - and sales picked up. The lesson learned was not to avoid similari-ish hardware following up a success, but that Nintendo's core appeal and fanbase should not be forgotten.
I'm not sure this is entirely accurate. A lot of the 3DS library from Nintendo is 2D platforming games straddling the line of red and blue ocean. It found its audience in a world on the cusp of throwing out handhelds altogether, and that's a win.

Switch 2 definitely seems to be in a more favourable position. What's its competition? Xbox Cloud Gaming? PlayStation Portal? Steam Deck? All systems with user counts in the ~10M or below mark.

All these products exist because the DEMAND for performant handhelds exist, but none have provided the holistic approach of Nintendo Switch yet. And none have been backed by exclusive software from major franchises. Nintendo has a market hungry for their device category but are weary that it might be a faint illusion, which is fair.
 
Hardware is only one half of the equation. Software is the other half of the equation, which is as important as hardware. Nvidia advertised developing the NVN API after the Nintendo Switch was announced. And I think NVN is one of the key aspects for the Nintendo Switch's third party support being quite good.

I imagine AMD also has to offer Nintendo a hardware and software package that's comparable to what Nvidia offers to have Nintendo seriously consider going back to AMD. Only time can tell if AMD can do so.
That's one of the things I implied with "all the downsides". Currently AMD is no match for Nvidia at the software aspect. But as Oldpuck said, they're catching up.
 
Didn't Iwata plan this though? I think they were trying to consolidate their business way earlier than the Switch, but the tech just wasn't there yet. They couldn't give up the handheld market due to how successful it is.

Yeah, Iwata talked about this a few times. The only difference is that he was expecting more devices with a shared library, not a Hybrid form factor.

Switch is definitely Iwata inheritance imho.
 
Anyway, it would be really cool to see dedicated hardware for directed graphs because it feels like NPC AI has stagnated badly.
I'm not in game dev but directed graphs does not seem like anything you would need dedicated hardware for.

what is a directed graph? and what does it do for npcs?

Kinda like decision nodes for NPC dialogue in general

 
I'm not in game dev but directed graphs does not seem like anything you would need dedicated hardware for.



Kinda like decision nodes for NPC dialogue in general

Hardware acceleration of dialogue trees. 💀
 
I think 3DS would have been more successful if:

1) Dropout 3DS gimmick and system be called 2DS
2) Screen resolution to 480p
3) Feature both second stick and ZR/ZL buttons
4) have made a more powerful SoC capable of running Wii games as 1:1 ports but with modern tools like programable shaders. It would be as powerful as Xbox 1.
5) Possible priced as $200
 
Question for those more familiar with Nvidia's architecture design, would it be possible and/or feasible to add a separate clock domain to
specific portions of the SM, like Tensor or RT Cores? Or are those logic blocks too tied into the Ampere/Lovelace SM structure for this to be viable without a significant architecture redesign? Thanks!
 
what is a directed graph? and what does it do for npcs?

NPC behavior is generally controlled by stuff called "finite state machines" or "behavior trees" which are just types of directed graph. Directed graphs have nodes that follow an order. NPC behavior has the NPC's brain following the instructions in the directed graph where it acts as demanded in the certain node until a condition happens that causes it to go to another node.

eg

image.png
 
Does anyone think there will ever be fixed function CPU cores that only do directed graphs to massively improve gaming NPC performance?

I don't know if there's any fixed function CPU stuff out there at all...

Unless the decompression chips found in modern game consoles are fixed function CPU stuff? I don't think they are, but maybe I'm wrong...
NPCs isn't something that needs new hardware to improve. It's not even a hardware problem. Devs have to just... make better NPC routines
 
NPC behavior is generally controlled by stuff called "finite state machines" or "behavior trees" which are just types of directed graph. Directed graphs have nodes that follow an order. NPC behavior has the NPC's brain following the instructions in the directed graph where it acts as demanded in the certain node until a condition happens that causes it to go to another node.

eg

image.png
ah ok. intresting, are state machines particularly cpu intensive? how and what kind of dedicated hardware could help with such a thing?
 
yeah, zelda, 2d mario, fire emblem, splatoon, pikmin, xenoblade, prime 5 must be all in prototype phase while the launch games and last switch games like fe4 remake,mario party and prime 4 are almost done, probably next year they gonna ramp up


not all of nintendo, unless they finishing a launch window this is really weird, above all for the dev of one of switch's graphic showcase

ZOXEJyb.png
With how pinchy Nintendo can be with money its probably that Nintendo does not want to pump more cash into their western studios right now with how weak the yen is compared to the dollar. If they are going to expand their studios soon it will probably just be in Japan.
 
0
Yeah, I have no idea which is why I'm asking, just a little disappointed by seeing DD2's NPCs eating up so much CPU and wondering if this will be the same for all games with a lot of NPC-NPC interactions.
yeh i mean, now im thinkin bout it, if all the ai is just doing the same kinda behavior check loops i dont see why that couldnt be parallelized with a dedicated hardware. itd at least let a lot more npcs work at the same time
 
0
ah ok. intresting, are state machines particularly cpu intensive? how and what kind of dedicated hardware could help with such a thing?
not really? when you have to update a fuckload of them, sure. I don't think dedicated hardware could help, mainly because of latency. latency penalties are quite severe with CPU related tasks. you're better off dedicating cores to the task than offloading it to offsite silicon
 
Yeah, I have no idea which is why I'm asking, just a little disappointed by seeing DD2's NPCs eating up so much CPU and wondering if this will be the same for all games with a lot of NPC-NPC interactions.
That is why classic game design is having NPCs that stand still and only says the same thing over again when interacting with the player. That doesn't put any pressure on the CPU.
 
0
They should turn that into a real game tbh
Maybe this could work as a tech-demo for the Switch 2, like Metroid Prime Hunters: First Hunt on the DS, serving as a proof-of-concept for the hardware's capabilities. They could also reskin this as a kind of futuristic Zelda tech-demo, it would satiate people curiosity as to how would a next-gen Zelda game look like.
 
Speaking as a dev with an ongoing project with ML NPCs, I can say that to reach their fullest potential they can't operate on simple graphs and trees. You need to fundamentally change and optimize their functions at the engine level. The whole point is to simulate real intelligence, preferably in multiple modalities, not merely navigate a set of instructions. They will need "vision", "hearing", "tactile experiences", possibly "proprioception" to reason about the input data from these modalities and believably interact with their environment and other entities open-endedly (how they do or say something should depend on what they've inferred from their "senses"). Hooking an LLM to them is not enough, nor is scripting their behavior. You need inference at the local level for this to feel realistic and we just aren't there yet. However, I estimate that this problem will be solved in a about a year, based on the trajectory of current machine learning developments.
 
Hey Fami! Time for a vibe check. How are the vibes? Are we back, or is it Nintendover?

Pretty good I think considering the recent shipment data discoveries, games coming to Switch 2, and we’re nearing June direct which is supposedly the last Switch 1 only Direct, then floodgates should open?


Which node process are those numbers for? 5nm right?
 
3 years from now.

I think Nintendo is likely to genuinely consider AMD anytime they're looking at new hardware. The Nvidia partnership is currently very strong, but Ko Shiota, head of the hardware team, has close relationships with AMD, having worked very closely with the silicon engineers on the Wii and the Wii U (and likely, but less clearly, the GameCube).

AMD might have been able to offer Nintendo better raw performance this time around. But it couldn't deliver DLSS, it couldn't deliver high ray tracing performance, and it would have had a significantly harder time offering backwards compatibility, or the sort of software investment that Nvidia has.

But AMD is actively trying to close the feature/technology gap with Nvidia, they have a much stronger custom chip division (where Nvidia's is essentially Just Nintendo), and they're hungry for it in a way that, by the third generation, Nvidia might not be.

AMD could come in and say, look, we've closed the gap on AI, on upscaling, on frame generation, on RT. But we're ahead of Nvidia on chiplets by a huge margin - we can integrate WiFi, Bluetooth, cartridge readers, your whole motherboard into a single SOC. We've got custom ARM cores we've been building on top of Zen's execution pipeline. We've moved to new process nodes faster than Nvidia, and are willing to give you chips at a razor thin margin.

Maybe it's only a 25% chance they pull that off, but that ain't 0%.
At this point I feel vendor lock in will be because of BC. Emulation of Switch 2 games will be painfully difficult to achieve in a manner that is transparent for the consumer.
 
Pretty good I think considering the recent shipment data discoveries, games coming to Switch 2, and we’re nearing June direct which is supposedly the last Switch 1 only Direct, then floodgates should open?



Which node process are those numbers for? 5nm right?
Yeah, TSMC normal 5nm, TSMC 4N should be even better in terms of energy efficiency
 
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