• 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.
  • Do you have audio editing experience and want to help out with the Famiboards Discussion Club Podcast? If so, we're looking for help and would love to have you on the team! Just let us know in the Podcast Thread if you are interested!
  • General system instability
    🚧 We apologise for the recent server issues. The site may be unavaliable while we investigate the problem. 🚧

StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (Read the staff posts before commenting!)

Not really, anyone can license arm. That's how ARMs business mode works. Also, since this is a handheld and a successsor to an arm console, it would make more sense to stick with arm, no matter who the vendor is.
While is very possible for AMD to use ARM, they had done before, I do think that if AMD had an AMD soc ready to go for 2022-2024, we would have hear about it. Specially if they didn't won Nintendo's contract. To be specific, the part that I find hard to believe is that AMD went all in to design a chip only for Nintendo, with the design shelved when that failed. I'm sure AMD would have known it was a long shot to win Nintendo over.

Or maybe the pitch was closer to "Hey Nintendo! You know, we can build you an ARM SoC. Give us a call if you intersted! Bye." than a full design.
 
Last edited:
I assume oldpuck thought that miamoto's talking about Nvidia's flavour of TSMC's N3E process node, especially if Hopper Next's still planned for launch in 2H 2024, and TSMC's N3E process node should start high volume manufacturing (HVM) this year, which is correct, not about how implausible the possibility of Nintendo and Nvidia using TSMC's 4N process node to fabricate Drake is, which I've already debunked here.
If you read the post he was replying to, its clear that it was in the context of 8nm vs 5nm.
 
While is very possible for AMD to use ARM, they had done before, I do think that if AMD had an AMD soc ready to go for 2022-2024, we would have hear about it. Specially if they didn't won Nintendo's contract. To be specific, the part that I find hard to believe is that AMD went all in to design a chip only to Nintendo, with the design shelved when that failed.
Let's say the bidding was done in 2018 or even before that. The chip woudnt have to exist at all, except on paper.
 
0
That AMD pitched an ARM design part is hard to believe. Also, Is doubtful that MLiD is privy to details of such communications.
allegedly, AMD is working on an ARM chip for MS. that said, I don't think it would be used for low power offerings as that would step on Samsung's toes
 
If you read the post he was replying to, its clear that it was in the context of 8nm vs 5nm.
I don't disagree. But to be fair to oldpuck, miamoto's post was really long. And I could see myself easily missing the context of miamoto's post, unless I re-read multiple times, if I were oldpuck.
That AMD pitched an ARM design part is hard to believe.
I don't believe so, considering AMD mentioned being ready for designing Arm based SoCs for customers. And as ILikeFeet mentioned, there was a Reuters article mentioning that AMD's planning on designing an Arm based SoC for laptops running Microsoft Windows.
 
Surely it was in Nintendo's best interests to field a design from AMD if only to leverage it against Nvidia in negotiations.

As far as I know this so called "20 year partnership" with Nvidia has never come from an official source? which means it's probably not true.

MLID definitely has credible sources. He's been right waaaay more times than most of the so called videogame 'insiders'. He rarely talks about Switch 2.
 
Please not, I don’t want to catch up on 20+ pages.

Oh I would hope for such luck lol. It depends on context of the hack though. We haven’t really gotten much “new” legitimate info or leak in ages.

Soonish we should get Vietnamese customs data for January but I feel it’ll be yet another month of nothingburger
 
Surely it was in Nintendo's best interests to field a design from AMD if only to leverage it against Nvidia in negotiations.

As far as I know this so called "20 year partnership" with Nvidia has never come from an official source? which means it's probably not true.

MLID definitely has credible sources. He's been right waaaay more times than most of the so called videogame 'insiders'. He rarely talks about Switch 2.
It comes from official source, from Nvidia CEO, and No MILD is not reliable
 
Surely it was in Nintendo's best interests to field a design from AMD if only to leverage it against Nvidia in negotiations.

As far as I know this so called "20 year partnership" with Nvidia has never come from an official source? which means it's probably not true.

MLID definitely has credible sources. He's been right waaaay more times than most of the so called videogame 'insiders'. He rarely talks about Switch 2.
The "20 year" thing actually came from Jensen himself, based on how long Nintendo had stuck with ppc750. There's no way we can know what's in their actual contract.
 
The "20 year" thing actually came from Jensen himself, based on how long Nintendo had stuck with ppc750. There's no way we can know what's in their actual contract.

Yeah I mean that's something he can talk about as a PR line but I doubt it was anything written in stone. If Nvidia was offering a crappy deal or crap chip Nintendo could go to say AMD or whoever as they please.
 
miamoto said:
The chip was done in 2022? It's not going to use a high end process from 2024. It's going to use whatever its sister-chips of the same architecture are being made on, which started sales in the last year or two.

You buy this? Those Lovelace chips, also taped out at the same time as Drake. And its a handheld where performance per watt is a lot more important, than even laptop chips.

I mean, I would buy the basic logic about sister chips (not timing, that seems more suggestive of 5nm), if we didn't have info about 8 a78, 12 SM but we do.
I think we're on the same page, let me rephrase for clarity

I think Miamoto's logic here matches mine from 2 years ago - don't be surprised when a chip built from 8nm technology, and sister to an 8nm chip is also 8nm. LiC supplemented that idea with "don't be 100% sure that Orin's power numbers match Drake's, especially when we know that there are power related customizations." Which I also think is smart.

Thraktor's power analysis shows that the amount of savings necessary is... very high. There is some circumstantial evidence that Drake shares a production line with Ada. Even @miamoto here simply doesn't believe that a chip T239's size is viable at 8nm - they just assume that it's not that size, rather than assume it's not 8nm. I don't buy that.

Supplementing what I said before -

In the 8nm hand, all this power draw analysis was really set on "not larger than the existing Switch." Which seems to be falling apart as an assumption, as we get more data. 13% larger device. Maybe smaller wifi chip, more efficient IO substrate, could we actually go for a 20% larger chip than the original TX1... maybe 20% more power draw too, supported by a larger battery? Larger fan? Maybe with a new formfactor, the power savings don't need to be quite as high over Orin as they looked on paper.

Returning to the 5nm hand - sure, maybe a bigger device makes 8nm more viable. But wouldn't that also rule out a smaller model? With a larger device, wouldn't a smaller model be even more appealing? Yes, the Lite sells slightly less than the OLED model, but by less than a couple percent, and it's more profitable. Not just because of the hardware, but because Lite users are more likely to be first time buyers, and thus buy more software. Would Nintendo leave that market (kids, women, exclusive handheld players, the budget minded) behind?

I look at all this and other considerations, (like the long term cost of a node shrink) and I think, 60/40 it's gonna be 5nm. But I wouldn't bet on a horse at those odds.
 
While is very possible for AMD to use ARM, they had done before, I do think that if AMD had an AMD soc ready to go for 2022-2024, we would have hear about it. Specially if they didn't won Nintendo's contract. To be specific, the part that I find hard to believe is that AMD went all in to design a chip only for Nintendo, with the design shelved when that failed. I'm sure AMD would have known it was a long shot to win Nintendo over.

Or maybe the pitch was closer to "Hey Nintendo! You know, we can build you an ARM SoC. Give us a call if you intersted! Bye." than a full design.
I don't how exactly how it would work, but I doubt they need to be complete plans. What Nintendo would need to know was ballpark performance, ballpark plans, ballpark price and ballpark when it would be ready. They would probably need some detail for it to be believable, but not a 100% finished design.


And considering how much money is at stake, I could see AMD putting some serious effort into it even if it was a long shot.
 
0
I am supplying the chip for Nintendo. We decided the flavor years ago, but I'm not at liberty to divulge that information.
slice-potato-chip-png.png
Is it the Nintendo Switch game cartridge flavor chip?
 
Some clarifying stuff about the Nvidia hack and the Linux leak

With less and less to analyze, I feel like my primary job is to be "institutional memory" about some of these things. The hack was 2 years ago (almost to the day!), and many folks have joined since then, and the people who saw some of the leaked files at that time have moved on from being thread regulars, so I'll just bring some of those very old discoveries forward.

Nvidia Hack
Nvidia got hacked with a ransomware attack, and the attackers dropped some of the found data onto the internet to get Nvidia to move. Some of that data was apparently stuff like "employee addresses and social security numbers." The majority of it was not Nintendo related. What we know about the hacked data mostly comes from Nintendo-centric grey-hatters who have sifted through, but except for a few screenshots and tiny excerpts, the actual data is out on the darkweb somewhere.

The Linux "leak"
Nvidia also published drivers for Drake in their Linux distribution. Those drivers were clearly in-progress before they were moved out of the public source code repository. But almost everything that came out of the hack was later confirmed by the Linux driver, and a lot more interesting data besides. The only thing that there wasn't a scrap of info about in the Linux driver was the size of the GPU. That remains basically the only thing we need to trust people who saw the NVN2 code to know about.

The GPU Size
I don't know what all was contained in the hack, but what mostly gets talked about is essentially the GPU driver/core API for Switch 2. This isn't a chip design document or spec sheet. It does list some properties of the chip, but that's not what's relevant here. What's relevant is that the code is designed to run on any Nvidia graphics card so that developers can use their workstations without needing a hardware devkit. But while running on those cards, it ignores the cards actual hardware and behaves as if the system has 12 SMs.

Rich at Digital Foundry called that "our smoking gun" when it came to GPU size, and I think that's about right. Whatever size we get, that was the target for the final chip.

File Age
First mention about NVN2 was from 2019 some people said, so there is no way AMD work on something for Nintendo when Nvidia aleardy start work

2019?? That's way too early.
Not sure about accuracy of 2019 mention itself, but APIs can exist before hardware does and NVN2 might have been tested against existing nvidia hardware that has DLSS capability in 2019, before T239 was taped out, etc.
NVN was forked for NVN2. They are substantially similar. Some of the files apparently have 2019 dates on them, not because it represents the start of NVN2 development, but the last time those files changed in the original NVN.

The version of DLSS integrated into NVN2 had a build date (if I'm remembering correctly, searching the thread is tough) of February 2022, days before the data was exfiltrated from Nvidia. So we know the code was up-to-date at the time of the leak

Chip Tape Out
The chip tape out timeline has nothing to do with the hacked data, and is instead something I found while diving into the Linux code. There is a bugfix about PCI timing in the Linux driver from "SSG" - that's Nvidia's Silicon Solutions Group. That sorta bug seems like the sort of thing that comes up when you are off simulator. Subsequent fixes also seemed "real hardware" related. I guessed at a tape-out in August of 2022 based on those commits.

That timing has since seemingly be confirmed by customs data, with the first chips shipped in August IIRC.

With the Linux drivers being updated all the way to tape-out, with tape-out happening only 6 months after the hack, and with the hack and the linux drivers agreeing on all points it seems highly likely that the few things in the hack that weren't in the Linux drivers - the GPU size - was still accurate at time of tape out.
 
Do y'all think there's a high possibility that Metroid is getting a soft delay for next year so that the switch 2 has two notable Nintendo Ip that's cross gen or might we see a holiday release, because jeff grub has said that their might be a Metroid marketing push in april.

Like Metroid 4 in my opinion would make for a great graphic showcase when it comes to ray tracing and fidelity plus i can see the switch running prime 4 similar to prime remaster (Stupidly good looking game might i say)

Also lastly which games would satisfy y'all needs for this year when it comes to switch releases, like for me having a prime 2-3 remaster and windwaker/twilight princess port would make me enough satisfy (Double points if retro made a Donkey Kong country return port from the Wii and the rumored fire emblem 4 remake)
 
0
Epic got hacked..... now I am not saying I hope there's some material on Switch 2.... but umm..... I wouldn't look the other way
You'll get internal dates for the EGS release of FF7 Rebirth and whatever Disney's plans with their future partnership at best.

Maybe a game or two that might be developed with UE4 or 5 if we're being optimistic, no hardware stuff though.
 
If there is one IP beside Zelda that can be used to show off technical capabilities of hardware, that's Metroid.
It would be very silly to keep it for Switch 1 only imo.

It's extremely unlikely that Prime 4 was designed with the Switch 2 in mind and thus a Switch 2 version of Prime 4 would just be a Switch 1 game with high resolution, TAA, less pop-in, better textures, better texture filtering, and shorter load times.

Which would be great, but wouldn't really be showing what the Switch 2 could do.

Mario Kart and 3D Mario are the most impressive franchises graphically from Nintendo other than Metroid and will probably be built from the ground up for Switch 2.
 
Genuinely expect the next mario kart to be one of if not the most impressive game they've ever made

I think people forget because... I mean, holy shit, it's been 10 fucking years, but Mario Kart 8 was easily the best looking game Nintendo had ever made when it released (with 3D World as it's only competition) and probably was the best looking game Nintendo had released until like...... IDK, Luigi's Mansion 3 over half a decade later?
 
You'll get internal dates for the EGS release of FF7 Rebirth and whatever Disney's plans with their future partnership at best.

Maybe a game or two that might be developed with UE4 or 5 if we're being optimistic, no hardware stuff though.
I mean more of a mention of switch 2. The code name or whatever.
 
I don't disagree. But to be fair to oldpuck, miamoto's post was really long. And I could see myself easily missing the context of miamoto's post, unless I re-read multiple times, if I were oldpuck.
That's what I said several pages before that Miamoto post are so long. I didn't finish read it because it's so long, not I having any agenda. Didn't understand why I got warning about it
 
0
Hey Fami Fam decided to pop my head in to ask is this Moores law guy legit and also can someone remind why we don't want 8nm?
You can browse the last few pages to see people talking pros and cons of what MLID is saying. The short version of 8nm is that compared to something better it would be bigger, hotter, more power hungry; and almost certainly also run slower to help compensate for those last two.
 
What I find really amazing about his speculation is that he still thinks the GPU is between 3 to 4 TFLOPs, but that the CPU is on par with the XB1's CPU... Which Switch's CPU at 1GHz is already ~2/3rds of the performance, while A78 cores are 3 times the performance per clock than the cores inside Switch... It's amazing how his logic block here looks like a sponge.
I believe he mentioned xbone x cpu, which is 30% faster than base xbone. Yeah I agree about saying switch 2 only matching that would be absurd, considering the 3x IPC gains you mentioned, and we'll likely be getting 8 cores too of course

You lost me at 2/3rds performance part though. Switch's 3 A57 gaming cores at 1GHz vs PS4's Jaguar's 1.6Ghz 6.5 cores for gaming already gives a 3.5x speed disparity on paper. And of course PS4's CPU is slightly lower than xbone.

.
 
That's what I want to know. Where do they get this stuff?
He doesn't
Let's give MLID the benefit of the doubt.
Let's not. He's a proven somewhat informed* liar.
*He knows just enough tech to attract the clueless crowd's attention. But he's no insider nor does he have special contacts, he only makes shit up for easy clicks and shrugs it off when it doesn't happen like he said it would.

The moment he said "oh and btw, AMD tried to get nintendo do let them make a switch 2 SoC for them ;)" when we knew YEARS ago that nintendo had already settled with NVIDIA for switch 2, that's the moment he completely lost any credibility (that he didn't have to begin with).
The 8nm bs and the 1024 CUDA cores is just him being uninformed.

Like, just think for a while: you're nintendo, you have an API that is made to work with NVIDIA hardware (NVN), the shader code you wrote for your games is essential for them to run many animations and effects, all that shader code would need to be recompiled and even then, many issues could arise if the code ran under an AMD iGPU (regardless if the CPU was ARM and not x86, what matters here is the GPU). Unlike NVIDIA, where they could (and probably already did) find a way for older shader code to run in a compatibility mode under the new iGPU.
AMD probably already knew that, and if they both knew nintendo would probably stick with NVIDIA, why the FUCK would AMD try betting hard against NVIDIA to get to make nintendo SoC??? what even is the logic here?

inb4 "maybe AMD wanted to make a console again with nintendo under the premise that it's not backwards compatible or maybe it is but uses emulation"
well, it's unlikely switch 2 won't be backwards compatible and emulation even if it's great would lead to a bunch of edge cases that nintendo probably doesn't want to deal with.
 
Anyone who posts a video a week or more with the word “leak” in it is a bullshitter. Or extremely gullible. Or a gullible bullshitter.

Either MLID is full of shit or the greatest prolific leaker in tech history has fewer followers than my grandma’s Instagram.
 
Ultimately I haven't really seen anything to dissuade me from the same opinion I had like 3 years ago -- Switch 2 will be a full generational leap from the Switch 1. So the Switch 1 was like a PS3.5 kind of that was capable of running a fair number of PS4 ports. Switch will probably be a PS4.5, something that can a fair deal of PS5 ports, probably more because DLSS will really help low resolution ports.

Most Nintendo consoles are actually a full generational upgrade from their previous system, the only exceptions were really the Wii, which had a revolutionary controller that comes around once every 40 years maybe and Wii U to Switch, though that also had a killer of a hardware feature in the hybrid concept, and Switch is also the successor to the 3DS and was clearly more than generational leap there.

All of Nintendo's portable machines have all been large upgrades from each other too, GB to GBA was a huge upgrade, GBA to DS was a big upgrade too, even with the 3D screen, 3DS was a full generation ahead of the DS. Switch obviously is way more than even a generation above the 3DS.

Unless they have some magic controller miracle (read: I'm just going to state this as probable fact -- they don't), the Switch 2 is going to be a generational leap over the Switch 1, because otherwise what exactly is the sales pitch? "Buy our new Switch that has no gimmick, and also a modest hardware upgrade"? Yeah good luck selling that to the tune of 150 million.
 
Outside of BG3, is there like a single game released this gen that is genuinely CPU constrained.
What’s the standard for “genuinely” CPU constrained? What CPU? How constrained? Are you only counting games that hit some threshold for quality, or do legacy multithreading implementations count?

Plague Tale: Requiem, Starfield, Gotham Knights all have their frame rates primarily set by CPU load on modern consoles, all for different reasons.

Plague Tale’s limit is objects on screen + IK animations. That’s why the 60fps patch halved the pixel counts but still needed lock most objects at 30fps, with distant objects even at 20fps.

Starfield is simulations limited, which is why an unlocked frame rate is highly variable but potato resolutions do nearly nothing.

Gotham Knights “fucking mess” limited, which maybe doesn’t count. But the core issue is “multithreading is hard”, and that’s something that virtualized geometry doesn’t fix, just puts a bandaid on.
 
What’s the standard for “genuinely” CPU constrained? What CPU? How constrained? Are you only counting games that hit some threshold for quality, or do legacy multithreading implementations count?

Plague Tale: Requiem, Starfield, Gotham Knights all have their frame rates primarily set by CPU load on modern consoles, all for different reasons.

Plague Tale’s limit is objects on screen + IK animations. That’s why the 60fps patch halved the pixel counts but still needed lock most objects at 30fps, with distant objects even at 20fps.

Starfield is simulations limited, which is why an unlocked frame rate is highly variable but potato resolutions do nearly nothing.

Gotham Knights “fucking mess” limited, which maybe doesn’t count. But the core issue is “multithreading is hard”, and that’s something that virtualized geometry doesn’t fix, just puts a bandaid on.

I'm talking about games that cannot hit 60 FPS on PS5 due to CPU constraints.

Gotham Knights and Starfield are good examples.

Anything that can run at 60 FPS on PS5 is probably not an issue for downporting to Switch 2 (assuming the Switch 2's CPU is at least 50% as powerful as the PS5's) as most AAA PS5 games will probably be downgraded to 30 FPS on Switch 2 to preserve the visual intent of the geometry and lighting other than games built around 60 FPS like Call of Duty and Tekken.
 
Please read this staff post before posting.

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


Back
Top Bottom