• 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!)

Maybe the engine is compatible, but its defining features are not.
All of their features except Raytracing are compatibles, and Epic use all of them on the fortnite patch.

Anyway, the conversation (and the thread) is about Drake, which could be compatible with raytracing and other things, not about Switch.
 
Last edited:
AFAIK, Dragon Quest XII uses Unreal Engine 5. I don’t think this is going to run nicely on the Switch if Square wants to publish the game only for PS5 / Series X/S.
We are talking about DQXII being a Drake launch exclusive though I would hope and expect Drake runs UE4/5 games a lot better than what the original Switch does, also if DQXII skips the Switch that’s the biggest suicide move ever done by a big publisher, even in the 3ds days (in which Nintendo was a lot less dominant in Japan) SE was forced to make a 3ds version mid development to avoid losing the sales from Japan.
 
Last edited:
Even though this has to be taken with a grain of salt, that is already two of them. When the first one came out a few weeks ago, I said we might be due to more of them soon. If this one is true I guess I was correct.
I believe that we are talking about the same source. They made a few additional comments in the past two weeks. I was also hesitant to repost until now.
 
All of their features except Raytracing are compatibles, and Epic use all of them on the fortnite patch.

Anyway, the conversation (and the thread) is about Drake, which could be compatible with raytracing and other things, not about Switch.
Nanite and lumen running on last gen? Seems like that would have been big news.
 
Lumen is raytracing.

Nanite works but with compromises and worse than on more powerful computers, such as a much smaller draw and polygonal rendering distance compared to more powerful computers.
There is a minimum threshold for storage speed for nanite to work decently, not to mention its cpu heavy.
 
I believe that we are talking about the same source. They made a few additional comments in the past two weeks. I was also hesitant to repost until now.
That is interesting, because if the machine is in production or close to starting production, its launch will not be especially far away (I think thqt could be Q1 2023 ), and sooner rather than later more information should leak.
 
0
I get it but I don’t think Yuij’s team is going for a « Fortnite » look for their next mainline entry.

I guess we’ll have to wait and see but Yuji alreadly stated that the game would target the next generation of consoles (aka PS5 / Series X/S -weren’t revealed back then).
I know this is off topic but I don't recall hearing that Hori ever said anything like that. He explicitly didn't give any hint about the platforms IIRC. Do you have a source for him saying it would target next gen consoles?
 
We are talking about DQXII being a Drake launch exclusive though I would hope and expect Drake runs UE4/5 games a lot better than what the original Switch does, also if DQXII skips the Switch that’s the biggest suicide move ever done by a big publisher, even in the 3ds days (in which Nintendo was a lot less dominant in Japan) SE was forced to make a 3ds version mid development to avoid losing the sales from Japan.
Ok, I was under the impression that Drake would just be a « Switch Pro » revision and not the next Nintendo console.
 
I know this is off topic but I don't recall hearing that Hori ever said anything like that. He explicitly didn't give any hint about the platforms IIRC. Do you have a source for him saying it would target next gen consoles?
The rumor started due to this job application:
https://attackofthefanboy.com/news/...ring-for-a-next-generation-dragon-quest-game/

“The Dragon Quest XI team within the Second Development Division has begun work a new project, starting up on a challenge unique to a Dragon Quest for the next generation.

The mission of this project is to “deliver a new appeal of Dragon Quest to fans worldwide.”

In order to achieve that, we are looking for new “party members” with the same desire.

The planning stages of development have wrapped up, and we are now entering the phase of further improving the game quality.

As a new high-definition Dragon Quest title, why not try to create a new Dragon Quest history that moves players like never before with Yuji Horii and the development team?

We await the applications of developers who love trying new things!”

Ok, I was under the impression that Drake would just be a « Switch Pro » revision and not the next Nintendo console.
A revision can have exclusives (Game Boy Color/New 3ds) and this was all hypothetical anyway of the only 3rd party game that could be important enough to help carry sales at the start.
 
Last edited:
Ok, I was under the impression that Drake would just be a « Switch Pro » revision and not the next Nintendo console.
However Nintendo positions it is just marketing, but Nate has heard about third party exclusives and the specs that leaked is definitely in line with what you would expect for next gen hardware.

EDT: Third party exclusives. Typo.
 
The leap from Wii to Wii U was huge and the Switch leap from Wii U was much smaller in comparison. Both are next gen Nintendo systems despite the power discrepancies.

Switch could have just been a Wii U pro. It was marketed as a successor. A reverse could happen.
 
Last edited:
The rumor started due to this job application:
https://attackofthefanboy.com/news/...ring-for-a-next-generation-dragon-quest-game/

“The Dragon Quest XI team within the Second Development Division has begun work a new project, starting up on a challenge unique to a Dragon Quest for the next generation.

The mission of this project is to “deliver a new appeal of Dragon Quest to fans worldwide.”

In order to achieve that, we are looking for new “party members” with the same desire.

The planning stages of development have wrapped up, and we are now entering the phase of further improving the game quality.

As a new high-definition Dragon Quest title, why not try to create a new Dragon Quest history that moves players like never before with Yuji Horii and the development team?

We await the applications of developers who love trying new things!”


A revision can have exclusives (Game Boy Color/New 3ds) and this was all hypothetical anyway of the only 3rd party game that could be important enough to help carry sales at the start.

A Dragon Quest for the next generation != built for next generation consoles.

I’d have only ever interpreted that as them wanting to evolve the franchise.
 
I think people forget about much of huge leap Wii to Wii U was and how small the Switch leap was from Wii U. Both are next gen Nintendo systems despite power discrepancies.

Switch could have just been a Wii U pro. It was marketed as a successor. A reverse could happen.
Drake would be massively underutilized by having its software always limited by tx1.

Nintendo isnt the company to release overpowered hardware for what its supposed to do.
 
They already did with Mariko.
Mariko was a simple die shrink. More or less every console ever made has had at least one. Also it was necesary to make the lite work at all. The main reason its such a huge improvement, is that 20 nm was a really bad node for mobile hardware.
 
0
The rumor started due to this job application:
https://attackofthefanboy.com/news/...ring-for-a-next-generation-dragon-quest-game/

“The Dragon Quest XI team within the Second Development Division has begun work a new project, starting up on a challenge unique to a Dragon Quest for the next generation.

The mission of this project is to “deliver a new appeal of Dragon Quest to fans worldwide.”

In order to achieve that, we are looking for new “party members” with the same desire.

The planning stages of development have wrapped up, and we are now entering the phase of further improving the game quality.

As a new high-definition Dragon Quest title, why not try to create a new Dragon Quest history that moves players like never before with Yuji Horii and the development team?

We await the applications of developers who love trying new things!”
Gotcha, thanks, so it's a misunderstanding of a translation. Boy that happens a lot.
 
A Dragon Quest for the next generation != built for next generation consoles.

I’d have only ever interpreted that as them wanting to evolve the franchise.
I agree with you that DQ isn't skipping the Switch and is more trying to keep the franchise evolving (like we have seen them already do with the mobile games) I just tried to post from where the rumor of DQXII being next-gen only (or skipping the Switch at least) started I should have clarified it.
 
I agree with you that DQ isn't skipping the Switch and is more trying to keep the franchise evolving (like we have seen them already do with the mobile games) I just tried to post from where the rumor of DQXII being next-gen only (or skipping the Switch at least) started I should have clarified it.

Yeah it was less about you and more about some previous posters talking like they’d be targeting only the highest end hardware
 
0
A Dragon Quest for the next generation != built for next generation consoles.

I’d have only ever interpreted that as them wanting to evolve the franchise.
At what point do we stop calling them next gen? It’s been 18 months and they are rolling towards 40 million out to the public across both PlayStation and Xbox.

At what point is stuff just targeting the current consoles especially stuff that won’t be out for a bit.
 
Wasn't this from the 2021-02-01 Q&A? The OLED Model was later announced on 2021-07-06, merely 5 months after. The "plans" changed I guess?
Right, I think that was the point -- "no plans" isn't a disconfirmation on its own.

Although this thread has seen it swing too far in the other direction with people saying that saying "no plans" in the next Q&A would actually be an implicit confirmation that something is launching this year, because they didn't choose a stronger form of denial. And that is just not true.
 
Right, I think that was the point -- "no plans" isn't a disconfirmation on its own.

Although this thread has seen it swing too far in the other direction with people saying that saying "no plans" in the next Q&A would actually be an implicit confirmation that something is launching this year, because they didn't choose a stronger form of denial. And that is just not true.
In all fairness, the one time they responded to a question asking about new hardware in a year where they didn't release new hardware they did explicitly say something like "we are not releasing any new hardware this year" which was in 2020.

But yeah I don't think we should assume they'll say the same thing either way.
 
0
Right, I think that was the point -- "no plans" isn't a disconfirmation on its own.

Although this thread has seen it swing too far in the other direction with people saying that saying "no plans" in the next Q&A would actually be an implicit confirmation that something is launching this year, because they didn't choose a stronger form of denial. And that is just not true.
Sorry to ask again, but have you seen any rtx io/ decompression functionality in nvn 2?
 
I feel so bad because I try my best to keep up with this thread but a LOT of it goes well over my head haha.
For now, for entertainment purposes, only look at the hidden comment and that should be more up your alley.

Remember, too much sodium chloride isn’t good for your health, so take in moderation.
 
I feel so bad because I try my best to keep up with this thread but a LOT of it goes well over my head haha.

Let me give you a quick, easy to understand summary of information:

1. Nvidia is developing a chip. Nintendo will use some version of said chip.
2. We are getting a new Nintendo system someday.
3. Insiders claim dev kits for some new Nintendo hardware exists for 1.5+ years.
4. People are "hoping" new hardware is coming by next Spring.

That is a basic summary of the last 4 years of speculating about new Nintendo hardware. That's not every detail ever but the highlights in my opinion.
 
if tape out starts in October or Novembe wouldn’t news media report of a new switch console and people will wait out for that instead of buying the current models?
Tape out does not need to be public. It could've happened already for all we know.

Mariko tape out was never reported or leaked IIRC.
 
I doubt the people making up the bulk of Switch's sixth holiday hardware sales are the type to be reading successor/revision rumor reports.
 
EDIT: The following post is for educational and entertainment purposes only. Unless something goes unprecedentedly and horrifically wrong, Nintendo's next system will NOT be using the specs discussed below.

A78 x4 @ 1Ghz = roughly 80% of PS4, 60% of PS4 Pro*
6SM @ 768Mhz = 1.18TFLOPS, PS4 and Pro are 1.84 and 4.2 TFLOPS respectively, take with grain of salt because of architectural differences
88GB/s is exactly half the bandwidth PS4 has

Those percentages are based off the assumption A78 is three times more performant per clock than Jaguar. To reach that estimate was super simplified, so it's possibly more inaccurate than the rest. I've read that Jaguar performs similarly per gigahertz to A57, and separately that A78 is 3-3.5x performant vs A57. Also, in calculating the rough numbers I accounted for one core in each system being reserved by the OS. I have no clue how much cache each A78 would have or I would have liked to compare that too, but it's probably too granular a point to bother with anyway since this is all hypothetical.

In the alternate universe we were living in a few months ago when we believed these specs likely, I would be shocked if Nintendo clocked things that low.

TL;DR: If you set your expectations for almost PS4-level performance in docked mode, maybe at worst 720p where PS4 is 1080p, that will definitely be safe :) That's without accounting for DLSS, which, with those specs, would likely be impressive but not mind-blowing.
Thank you so much!
720 p with then added DLSS is actually pretty good!
Ironically, the weak point of this system would again be the memory bandwidth, anyway this will be my expectation from now on.
Prepare to be disappointed hahaha
 
if tape out starts in October or Novembe wouldn’t news media report of a new switch console and people will wait out for that instead of buying the current models?
A tape out isn’t public, it’s private. No one is supposed to know when a tape out happens.

From Tape out to release can span 1 quarter or 3 quarters.

It just depends on the logistics of the situation.

XBox wasn’t taped out until very late and released a few months after

PS5 didn’t have any announcement of a tape out, but just that production started in June for the November launch.

A tape out information is based on a leaker and if they care for that, some do while others don’t. Why? Because some have an interest in the consoles and for the console space of technology, others have an interest of only exclusively the PC side and couldn’t care less for leaking console related things like that. Doesn’t help that console space garners too much interest compared to the PC space.


In essence, I wouldn’t expect Kopite7kimi to announce a tape out of Drake when it happens.
 
Thank you so much!
720 p with then added DLSS is actually pretty good!
Ironically, the weak point of this system would again be the memory bandwidth, anyway this will be my expectation from now on.
Prepare to be disappointed hahaha
Yea, unfortunately there's not a whole lot that could be done about that. Various approaches of increasing cache here and there help, but don't completely solve the issue.
Fundamentally, the issue is that the active transfer of data costs energy. And Thraktor has a post here making some estimates. To quote the specific paragraph I had in mind:
"The capacity of the memory shouldn't impact the performance, just the speed, width and type of the interface. I calculated a little while back that LPDDR5 probably has around 4 picoJoules per bit (pJ/b) of energy consumption, based on claims from manufacturers on LPDDR4X consumption and efficiency improvements for LPDDR5, but it's only a rough estimate, and likely to vary from manufacturer to manufacturer. If it is 4 pJ/b, though, then the power consumption for 102.4GB/s would be a bit over 3.3W. For LPDDR5X, Samsung claims 20% less power consumption than LPDDR5, so we could expect about 3.2 pJ/b, which would result in a bit over 3.5W to hit the maximum 136GB/s. For comparison, I'd say the LPDDR4 in the original Switch consumed between 1.5W to 2W for 25.6GB/s of bandwidth, so it's an increase over the original model either way. My guess is that the memory will be clocked quite a bit lower in portable mode to accommodate for this."

And the Switch's usage & form factor realistically allows for only so many watts. To recap, citing this article, the OG Switch itself needs about 11 watts while docked and 7.1-8.9 watts while undocked, depending on screen brightness.
That's not to say that Drake will necessarily have the same power consumption numbers. I just listed the above as an example.

My personal speculation bouncing around in my head for fun (please don't take this as hard expectations):
~15 watts docked, ~8-9 watts undocked (excluding screen).
Earlier I said 2 watts for the CPU; I think that CPU clocks will remain the same in both modes.
I expect either eMMC or eUFS as internal storage. Ergo, I assume 0.5 to 1 watts when active (that is, I assume operating conditions to vary from 25 Celsius to... somewhere higher than 25 C, but nowhere near 85 C).
Docked: be it either LPDDR5 or 5X, I expect full speed, so pencil in low to mid 3 watts. I think that we're at ~6 watts now? Alright then, 8-9 watts for the GPU. With 12 SMs, that's 2/3 to 3/4 of a watt per SM.
Undocked: RAM may be clocked down to... about 7/8th of its LPDDR5 max or 2/3 of its LPDDR5X max, so ~2.8 or ~2.33 watts. Shit, that's a wider range than I'd like given this much room to work with. That's like... 5 to low 5 point something watts from CPU + RAM + actively using internal storage. Ehhh, I guess 3-4 watts for the GPU means 1/4 to 1/3 of a watt per SM.
Yaknow, in certain contexts, 12 SMs is freaking huge, man.

-----

Re: Tape out
Yea, I wouldn't expect to hear about it whenever it happens, or if it has happened already. The specific source who told us that that chip wasn't taped out as of early 2021, has been under NDA since like, summer of last year. I don't think that we've heard anybody else speak on this particular topic.
 
if tape out starts in October or Novembe wouldn’t news media report of a new switch console and people will wait out for that instead of buying the current models?
The real question is how many people would even believe it after last year's OLED mix up?
 
0
Yea, unfortunately there's not a whole lot that could be done about that. Various approaches of increasing cache here and there help, but don't completely solve the issue.
Fundamentally, the issue is that the active transfer of data costs energy. And Thraktor has a post here making some estimates. To quote the specific paragraph I had in mind:
"The capacity of the memory shouldn't impact the performance, just the speed, width and type of the interface. I calculated a little while back that LPDDR5 probably has around 4 picoJoules per bit (pJ/b) of energy consumption, based on claims from manufacturers on LPDDR4X consumption and efficiency improvements for LPDDR5, but it's only a rough estimate, and likely to vary from manufacturer to manufacturer. If it is 4 pJ/b, though, then the power consumption for 102.4GB/s would be a bit over 3.3W. For LPDDR5X, Samsung claims 20% less power consumption than LPDDR5, so we could expect about 3.2 pJ/b, which would result in a bit over 3.5W to hit the maximum 136GB/s. For comparison, I'd say the LPDDR4 in the original Switch consumed between 1.5W to 2W for 25.6GB/s of bandwidth, so it's an increase over the original model either way. My guess is that the memory will be clocked quite a bit lower in portable mode to accommodate for this."

And the Switch's usage & form factor realistically allows for only so many watts. To recap, citing this article, the OG Switch itself needs about 11 watts while docked and 7.1-8.9 watts while undocked, depending on screen brightness.
That's not to say that Drake will necessarily have the same power consumption numbers. I just listed the above as an example.

My personal speculation bouncing around in my head for fun (please don't take this as hard expectations):
~15 watts docked, ~8-9 watts undocked (excluding screen).
Earlier I said 2 watts for the CPU; I think that CPU clocks will remain the same in both modes.
I expect either eMMC or eUFS as internal storage. Ergo, I assume 0.5 to 1 watts when active (that is, I assume operating conditions to vary from 25 Celsius to... somewhere higher than 25 C, but nowhere near 85 C).
Docked: be it either LPDDR5 or 5X, I expect full speed, so pencil in low to mid 3 watts. I think that we're at ~6 watts now? Alright then, 8-9 watts for the GPU. With 12 SMs, that's 2/3 to 3/4 of a watt per SM.
Undocked: RAM may be clocked down to... about 7/8th of its LPDDR5 max or 2/3 of its LPDDR5X max, so ~2.8 or ~2.33 watts. Shit, that's a wider range than I'd like given this much room to work with. That's like... 5 to low 5 point something watts from CPU + RAM + actively using internal storage. Ehhh, I guess 3-4 watts for the GPU means 1/4 to 1/3 of a watt per SM.
Yaknow, in certain contexts, 12 SMs is freaking huge, man.

-----

Re: Tape out
Yea, I wouldn't expect to hear about it whenever it happens, or if it has happened already. The specific source who told us that that chip wasn't taped out as of early 2021, has been under NDA since like, summer of last year. I don't think that we've heard anybody else speak on this particular topic.
In an ideal scenario, they'd give at most 3.25W just to the CPU, I wonder if 5X is on the table to have it clocked to give 88GB/s in portable mode and in docked it goes to 136GB/s

LPDDR5 would be 88GB/s in portable and 102GB/s docked, but 5X would be 112GB/s portable (I think?) and 136GB/s docked.

Perhaps if they used the lower profile that results in 88GB/s when going with LPDDR5X? and the full in docked?

What can the frequency be for the CPU if it is given 3.25W of room btw instead of 2W?
 
At what point do we stop calling them next gen? It’s been 18 months and they are rolling towards 40 million out to the public across both PlayStation and Xbox.

At what point is stuff just targeting the current consoles especially stuff that won’t be out for a bit.
I definitely agree with you. Series X|S and PS5 are current gen. I was only using the language I used because I was addressing how some might have taken this listing to mean they were targeting only cutting edge tech ie. 'next generation'.

“A Dragon Quest for the next generation” in the listing was 100% using generation to describe people and fandom, not technology. The entire rest of the listing focuses on creating something to appeal to a new era of fans. If they wanted tech to be the goal, they'd have been far more explicit.

Anyway, the thread has moved on...
 
In an ideal scenario, they'd give at most 3.25W just to the CPU, I wonder if 5X is on the table to have it clocked to give 88GB/s in portable mode and in docked it goes to 136GB/s

LPDDR5 would be 88GB/s in portable and 102GB/s docked, but 5X would be 112GB/s portable (I think?) and 136GB/s docked.

Perhaps if they used the lower profile that results in 88GB/s when going with LPDDR5X? and the full in docked?

What can the frequency be for the CPU if it is given 3.25W of room btw instead of 2W?
Assuming 8 cores and all at the same clock? (ie simplifying by not letting the OS core do its own thing)
(reminder to readers that this is just rough/simplified, ballpark estimation!)

Take 3.25w, then divide by 8 to get 0.40625 (that is, target power per core)
I then flip it around by hitting the 1/X button in Calculator to get ~2.46154 (ratio of 1 watt:target power per core)
Then I hit the sqrt button to get ~1.56893 (ratio of clock@1 watt:clock@target power per core)
Then I hit the 1/X button again to get ~0.63738 (flip it around to get clock@1 watt at the bottom)

If N7, then assume 1 watt = 2.6 ghz per ARM's claims
Multiply 2.6 by ~0.63738 to get 1.657 ghz per core

If N5, then assume 1 watt = 3 ghz per ARM's claims
Multiply 3 by ~0.63738 to get 1.912 ghz per core
 
Assuming 8 cores and all at the same clock? (ie simplifying by not letting the OS core do its own thing)
(reminder to readers that this is just rough/simplified, ballpark estimation!)

Take 3.25w, then divide by 8 to get 0.40625 (that is, target power per core)
I then flip it around by hitting the 1/X button in Calculator to get ~2.46154 (ratio of 1 watt:target power per core)
Then I hit the sqrt button to get ~1.56893 (ratio of clock@1 watt:clock@target power per core)
Then I hit the 1/X button again to get ~0.63738 (flip it around to get clock@1 watt at the bottom)

If N7, then assume 1 watt = 2.6 ghz per ARM's claims
Multiply 2.6 by ~0.63738 to get 1.657 ghz per core

If N5, then assume 1 watt = 3 ghz per ARM's claims
Multiply 3 by ~0.63738 to get 1.912 ghz per core
Thank you so much for showing the math on this! Posts like this one are how I learned to calculate theoretical performance for GPUs, but power usage has always seemed so opaque that I could never find a starting point to dig into it.
 
0
No prob.
And yea, chip power usage isn't the most straightforward thing for me either. I needed a bit to notice that the curve was quadratic (it's predominantly voltage squared). But there's the other little details that make it such that what I wrote above are ballpark figures/approximations. Which is still sufficient for casual discussion at least.
 
Thinking about getting a Steamdeck while we wait for Drake. Anyone know how many times more powerful Steamdeck is than the current Switch?
 
Assuming 8 cores and all at the same clock? (ie simplifying by not letting the OS core do its own thing)
(reminder to readers that this is just rough/simplified, ballpark estimation!)

Take 3.25w, then divide by 8 to get 0.40625 (that is, target power per core)
I then flip it around by hitting the 1/X button in Calculator to get ~2.46154 (ratio of 1 watt:target power per core)
Then I hit the sqrt button to get ~1.56893 (ratio of clock@1 watt:clock@target power per core)
Then I hit the 1/X button again to get ~0.63738 (flip it around to get clock@1 watt at the bottom)

If N7, then assume 1 watt = 2.6 ghz per ARM's claims
Multiply 2.6 by ~0.63738 to get 1.657 ghz per core

If N5, then assume 1 watt = 3 ghz per ARM's claims
Multiply 3 by ~0.63738 to get 1.912 ghz per core
No prob.
And yea, chip power usage isn't the most straightforward thing for me either. I needed a bit to notice that the curve was quadratic (it's predominantly voltage squared). But there's the other little details that make it such that what I wrote above are ballpark figures/approximations. Which is still sufficient for casual discussion at least.
Even then it’s really interesting and cool work to see! Since it’s an AI supercomputer that is actually designing Drake, and it is much faster designing Drake than a human could, it could actually perhaps find an even more efficient design that doesn’t consume more power but can edge out more frequency even at the finest level.


Speaking of, I wonder how long it takes to design a chip like this using this method. I know it’s faster, but not sure how much faster.



Not to mention that due to it being derived from an already existing chip, I wonder if this can make the process of switching nodes, while not at all cheap, quicker and easier to do.
 
0
Please read this staff post before posting.

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


Back
Back
Top Bottom