• 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!)

Reader mode
The last 10 pages summed up

look-at-this-fucking-nothing.gif
 
Hermii said:
Don't tell others how to spend their money.

That's for a year, a month would be plenty for us. That's 3k/12=250. So adjusting for a year being the cheaper option say 300. 30 of us donate 10$ and we're there.
Click to expand...
A little fiscal responsibility can go a long way. (I know you're probably joking, but still)

That's actually a pretty smart plan. 👀
 
D36 said:
I don't think anyone here has that kind of money to burn, anyway and if they did, it certainly shouldn't be for something like this. Also, they probably wouldn't be hanging around here.
Click to expand...
Yeah, lol. I would save the $3k for Switch 2, whatever the next-gen version of Switch Pro is (unless they're keeping it for Switch 2, cuz I love Switch Pro so much) and roughly 40 games
 
ItWasMeantToBe19 said:
It’s much more expensive in Japan (60000 yen) than in the US to try to compensate for how weak the yen is and it’s still not working for Sony. Most of the systems seem to be being resold overseas based on how bad the PS5’s software sales in Japan are.
Click to expand...
It’s more expensive compared to the US just like every other country that has a PS5 that isn’t the US. This isn’t a Japan only issue.

It’s partially not more expensive in the US, because Xbox is competitive in the US, they can’t get away with this.

PS5 is more expensive in many regions where Xbox is frankly irrelevant to the market, it’s not so in regions where Xbox puts up a noticeable fight in the market.
 
Thanks to P4blo's shenanigans, this thread has been moving faster than it has been for the past few weeks. We're definitely gonna hit 2000 before Halloween, hell, we might even get there before Wonder releases.
 
ReddDreadtheLead said:
It’s more expensive compared to the US just like every other country that has a PS5 that isn’t the US. This isn’t a Japan only issue.

It’s partially not more expensive in the US, because Xbox is competitive in the US, they can’t get away with this.

PS5 is more expensive in many regions where Xbox is frankly irrelevant to the market, it’s not so in regions where Xbox puts up a noticeable fight in the market.
Click to expand...

Okay so I’m talking about the relentless scalping issues in Japan that seem to be strongly affecting the PS5 is the thing. Software sales are just super strangely low relative to hardware sales and those sales plus the extremely weak yen seem to point to PS5s being shipped out of Japan.

This is a pretty serious issue for Nintendo if they want the Switch 2 to have a large Japanese user base for games like Dragon Quest etc.
 
mjayer said:
I'd still love @Thraktor to take another look at this. These estimates are for a drake inside a Switch sized device... how might he adjust it in a device sized up to SteamDeck or bigger? Id also love to see a power curve chart for 4nm, would that change things?
Click to expand...
Not to speak for Thrak, but he was trying to game out the process node. Moving to a larger device opens up new process node possibilities, but actually doesn't change the performance estimates. Let me see if I can summarize why, since I genuinely think Thraktor's post is so brilliant, I wanna fangirl on it a bit.

TL;DR: Thraktor's analysis is basically the same, regardless of form factor or the node.

We know Drake is 12 SMs. At 420 MHz, that is 1.29 TFLOPS. Thraktor's power curves let us guess at the power draw on 8nm, 5.76 Watts. From this, folks were trying to make 8nm work, by pushing clocks lower and lower until we got something plausible. "Nintendo will clock is super low" was an article of faith after the Switch.

Except Thraktor realized the obvious thing which is that Nintendo clocked the Switch super low, because clock speeds are the only thing they could control, because the chip was pre-existing. With a custom chip they could alter the size of the GPU. So what would a different chip look like?

Let's imagine a much smaller Drake, like 8SMs. And let's push the clock speeds up higher to compensate for it's smaller size, 624Mhz. That's 1.27 flops, basically the same as our previous experiment. And with Thraktor's power curves... 5.8 Watts. Also basically the same.

Bigger chips cost more and Nintendo controls the size of the chip this time. So why would Nintendo go for the more expensive option (the larger GPU), which requires lower clocks (because of battery life) that don't actually offer extra performance?

They wouldn't. Obviously. In fact, if you were designing a GPU for max performance with a tight power budget, on a well known architecture, you'd do the opposite. You'd lock in the clock speed first, at peak efficiency, and then dial in the number of cores till you hit your maximum, or ran out of money. In Ampere, 12SMs happens to just under a line where you there is a sudden spike in the complexity and size of the chip.

12 SMs at sub 522 Mhz clocks doesn't make any financial sense. 12 SMs at higher clocks on 8nm draws too much power for a Switch. Ergo, not 8nm. That was Thraktor's primary conclusion.

Unfortunately, we don't have data that would allow us to extract a similar power curve for other nodes. Thraktor's reasonable guess was that 4N would provide similar-or-better power curves, which is the basis of his clock estimates. But it's possible that's wrong (thought doubtful). It's also possible that the power curve for Orin Ampere differs from gaming Ampere in some way. Recently, I saw some power testing numbers from an Ampere laptop, and Thraktor's curves were spot on, so I doubt that's true as well, but it's still possible.

Changing to a Steam Deck sized device also doesn't change this analysis much. A bigger device makes 8nm more viable (bigger battery, bigger fan) but it doesn't change where the sweet spot is for the handheld, or where the bandwidth limit exists in docked mode. A bigger device however costs more - a larger battery, a bigger heatsink, and the shipping costs of a larger device. Is the cost savings of 8nm beat the cost increases of a larger piece of hardware?

Probably not, that's why Playstation Slim exist - a smaller node lets you make a smaller device, which cuts shipping and manufacturing costs, and saves money.

I will say that I disagree with Thraktor's conclusions slightly. As @LiC likes to point out, marginal gains are still gains. Nintendo might start at peak efficiency when planning, but then the actual physical hardware comes together and not everything is going to line up exactly. If there is slack in the design, or if the devs really need that extra performance, then clocks can be pushed past peak efficiency. And if it comes down to it, if battery life isn't where it needs to be, or the heat is too high, clawing back the performance will be necessary.

I think Thraktor is pushing the top of that line with his predictions, so I expect to come below that as the back and forth of design wears on. Note that I consider this optimistic not pessimistic. One of the reasons that Switch kept getting handheld clock bumps late in development (and after release, effectively) was because they chip just couldn't deliver at the battery life they wanted. I think that T239 is well designed and that sort of late stage clock pushing won't be necessary.
 
0
The issue with a lot of the “Samsung 8nm wouldn’t make sense, it would cost more while being way heavier” analysis is that it looks at current and actual costs right now instead of projected costs back in 2021.

Yes, Samsung 8nm would end up making this cost a lot more.

But if NVIDIA gave extremely expensive projections for the costs of 4N chips in 2024, the projected cost for Samsung 8nm could have been much lower relatively.

If this is Samsung 8nm, it’s largely just because NVIDIA or Nintendo screwed up projecting demand for 4N chips, but there were tons of projection screwups in 2020-2021 due to COVID.
 
Last edited:
ItWasMeantToBe19 said:
The issue with a lot of the “Samsung 8nm wouldn’t make sense, it would cost more while being way heavier” analysis is that it looks at current and actual costs right now instead of projected costs back in 2021.

Yes, Samsung 8nm would end up making this cost a lot more.

But NVIDIA gave extremely expensive projections for the costs of 4N chips in 2024, the projected cost for Samsung 8nm could have been much lower relatively.

If this is Samsung 8nm, it’s largely just because NVIDIA or Nintendo screwed up projecting demand for 4N chips, but there were tons of projection screwups in 2020-2021 due to COVID.
Click to expand...
Source for those 2024 numbers? Are you saying 4N in 2024 would offer less bang for buck than 8N? If I recall, 4N is 2.7x denser than 8N, is 4N in 2024 also more than 2.7x expensive?
 
darthdiablo said:
Source for those claims? Are you saying 4N in 2924 would offer less bang for buck than 8N? If I recall, 4N is 2.7x denser than 8N, is 4N in 2024 also more than 2.7x expensive?
Click to expand...

What in the world are you talking about?

I’m saying that many companies in 2020 and 2021 expected massive demand for chips to continue indefinitely.

In this time period, they may have estimated the cost of near cutting edge wafers to be extremely expensive due to the extremely high projected demand.

I am saying nothing about current prices and make that pretty clear?

The Switch 2 is only Samsung 8nm if NVIDIA screwed up projecting demand in 2021… But a lot of companies did that.
 
ItWasMeantToBe19 said:
The issue with a lot of the “Samsung 8nm wouldn’t make sense, it would cost more while being way heavier” analysis is that it looks at current and actual costs right now instead of projected costs back in 2021.

Yes, Samsung 8nm would end up making this cost a lot more.

But if NVIDIA gave extremely expensive projections for the costs of 4N chips in 2024, the projected cost for Samsung 8nm could have been much lower relatively.

If this is Samsung 8nm, it’s largely just because NVIDIA or Nintendo screwed up projecting demand for 4N chips, but there were tons of projection screwups in 2020-2021 due to COVID.
Click to expand...
That still wouldn't explain the chip size. As Oldpuck explained a few posts above, a smaller chip with higher clocks would have similar if not better performance per watt as a 12 SM soc with ultra low clocks on 8nm.
 
Hermii said:
That still wouldn't explain the chip size. As Oldpuck explained a few posts above, a smaller chip with higher clocks would have similar if not better performance per watt as a 12 SM soc with ultra low clocks on 8nm.
Click to expand...
Yes, but Nintendo could have prioritized computing power over weight based on market research. Will people be okay with a 22W monster with a Steam Deck like weight? Maybe, a lot of people are okay with the Steam Deck’s weight currently.
 
0
ItWasMeantToBe19 said:
What in the world are you talking about?

I’m saying that many companies in 2020 and 2021 expected massive demand for chips to continue indefinitely.

In this time period, they may have estimated the cost of near cutting edge wafers to be extremely expensive due to the extremely high projected demand.

I am saying nothing about current prices and make that pretty clear?
Click to expand...
I'm sorry, no you didn't make anything clear.

Assuming 4N is 2.7x denser than SEC8N, your hypothetical would be a concern only if it also is 2.7x or above more expensive. Are you thinking we are past that point?
 
ItWasMeantToBe19 said:
Okay so I’m talking about the relentless scalping issues in Japan that seem to be strongly affecting the PS5 is the thing. Software sales are just super strangely low relative to hardware sales and those sales plus the extremely weak yen seem to point to PS5s being shipped out of Japan.

This is a pretty serious issue for Nintendo if they want the Switch 2 to have a large Japanese user base for games like Dragon Quest etc.
Click to expand...
Nintendo won’t have this problem because Japanese people actually want to buy Switches.
 
darthdiablo said:
I'm sorry, no you didn't make anything clear.

Assuming 4N is 2.7x denser than SEC8N, your hypothetical would be a concern only if it also is 2.7x or above more expensive. Are you thinking we are past that point?
Click to expand...

?????????

I am saying it’s very possible NVIDIA (in 2021 when demand was at its highest) made a projection that the price gap would be larger than 2.7x in 2024 and several years afterwards.

You are asking for sources about hypothetical projections that would explain the 8nm rumors and that does not make any sense.
 
ItWasMeantToBe19 said:
What in the world are you talking about?

I’m saying that many companies in 2020 and 2021 expected massive demand for chips to continue indefinitely.

In this time period, they may have estimated the cost of near cutting edge wafers to be extremely expensive due to the extremely high projected demand.

I am saying nothing about current prices and make that pretty clear?

The Switch 2 is only Samsung 8nm if NVIDIA screwed up projecting demand in 2021… But a lot of companies did that.
Click to expand...
As I said in the thread previously, but even in 2021, if you are going to design an SoC focused on a minimally portable gaming device, you don't design it in a WORSE architecture than that of the consoles launched in 2020, it doesn't make any sense, in the At least NVidia would have planned something at TSMC 7nm.
 
ItWasMeantToBe19 said:
?????????

I am saying it’s very possible NVIDIA (in 2021 when demand was at its highest) made a projection that the price gap would be larger than 2.7x in 2024 and several years afterwards.

You are asking for sources about hypothetical projections that would explain the 8nm rumors and that does not make any sense.
Click to expand...

I'm trying to add something more substantial to your mental exercise, thus why I asked for sources or numbers. Based on your previous reply, I gathered you were talking purely hypothetical. Which is perfectly fine.

I'm just saying, for Nintendo and Nvidia to have made a bad projection back in 2020/2021, the cost (ie: 2.2x in 2022) would have had to go beyond the density improvement (ie: 2.7x in 2022) we see going from SEC8N to 4N, if that makes sense.

I seriously doubt Nintendo/Nvidia treated semiconductor prices as a static thing when they did an analysis/decision back then. They most likely made their decision, factoring for future potential changes in the market.
 
Last edited:
LinkURL said:
As I said in the thread previously, but even in 2021, if you are going to design an SoC focused on a minimally portable gaming device, you don't design it in a WORSE architecture than that of the consoles launched in 2020, it doesn't make any sense, in the At least NVidia would have planned something at TSMC 7nm.
Click to expand...
I don't think that would have factored into it. Nvidia achieved a similar performance per watt on 8nm with ampere than AMD did with 7nm. Performance per watt is all that matters here, not how it's achieved.

Which is why 12 SM on 8nm doesn't make sense , unless Drake is somehow so efficient it makes every other comparison point irrelevant.
 
Hermii said:
I don't think that would have factored into it. Nvidia achieved a similar performance per watt on 8nm with ampere than AMD did with 7nm. Performance per watt is all that matters here, not how it's achieved.

Which is why 12 SM on 8nm doesn't make sense , unless Drake is somehow so efficient it makes every other comparison point irrelevant.
Click to expand...
What I've said is, it's either 4N at standard efficiency, or it's 8nm but so power efficient it knocks everything else out of the water, to the point it might as well be 4N.

Ultimately, it probably doesn't matter, even if one is more likely than the other. The NG Switch will probably use comparable power consumption to... Nintendo Switch. Because it's still a handheld marketed to children.
 
Last edited:
oldpuck said:
Maxes are probably 3GHz, since that's the default max from ARM. Technically, an integrator can alter the power curves to push higher, but a quick check of Qualcomm and Samsung SOCs shows that no one pushed those cores past 2.8GHz.
Click to expand...
Mediatek has the max frequency for the Cortex-A78 on the Dimensity 8300, which is fabricated using TSMC's 4 nm* process node, set at 3.1 GHz.
* → a marketing nomenclature used by all foundry companies

But saying that, I don't think the Cortex-A78's max frequency (3.1 GHz) is for sustained performance, considering the performance CPU cores on smartphone SoCs generally only run at high frequencies in short bursts of time.
 
Is there a path from Samsung 8nm for a future revision of the Switch 2 like there was with the Switch going from the TX1 to TX1+ (20nm to 16nm)? If not, then it really wouldn't make sense to go with that, right? We're talking about an SoC that is going to be in use for the next 6+ years. Nintendo didn't have any real control when they first contracted the use of the Tegra X1, but with the T239 being custom with budgets and expectations on their behalf, they have a LOT more.
 
P4bl0 said:
Now I want a Luigi Burger ;-)

[Luigi Burger Deleted for the good of Humanity]
Click to expand...
I made myself a chilli chicken wrap and it made my day better. I don't want your damn vegetables Luigi Burger.

Tbh, the biggest justification for the new tech in a Nintendo System is for Monolith Soft to be forced to show a Xenoblade character actually eat food instead of eating indestructible sandwiches and soup.
 
Concernt said:
What I've said is, it's either 4N at standard efficiency, or it's 8nm but so power efficient it knocks everything else out of the water, to the point it might as well be 4N.

Ultimately, it probably doesn't matter, even if one is more likely than the other. The NG Switch will probably use comparable power consumption to... Nintendo Switch. Because it's still a handheld marketed to children.
Click to expand...
That might be true for the GPU, but the CPU is a known quantity at 8 vs 7-4NM.
 
0
I have been thinking that if it's using TSMC 4N, then Nintendo may want a battery life around 4.5-5 hours while running the next AAA zelda.
Then, in 2026 they could release the lite model using the same node, but having ~3h on zelda. It would be the inverse they did when comparing switch v1 vs Lite on this matter, with the latter had advantage.

So I believe a handheld clock in the range of 460-520 is a safer bet [for me]
I'm ok with the performance it will give.
 
Discostew said:
Is there a path from Samsung 8nm for a future revision of the Switch 2 like there was with the Switch going from the TX1 to TX1+ (20nm to 16nm)? If not, then it really wouldn't make sense to go with that, right? We're talking about an SoC that is going to be in use for the next 6+ years. Nintendo didn't have any real control when they first contracted the use of the Tegra X1, but with the T239 being custom with budgets and expectations on their behalf, they have a LOT more.
Click to expand...
no, Samsung stopped developing 10nm further. guess they can always make a further enhanced variant, but I doubt anyone would spend money on that
 
Kise Ryota said:
I have been thinking that if it's using TSMC 4N, then Nintendo may want a battery life around 4.5-5 hours while running the next AAA zelda.
Then, in 2026 they could release the lite model using the same node, but having ~3h on zelda. It would be the inverse they did when comparing switch v1 vs Lite on this matter, with the latter had advantage.

So I believe a handheld clock in the range of 460-520 is a safer bet [for me]
I'm ok with the performance it will give.
Click to expand...

I don’t think cutting a few ounces of battery weight would be a very appealing product. The Lite for the Switch 1 already sold not great so making a Lite that has a much worse battery and is only a couple bucks cheaper and a couple ounces smaller could really struggle.
 
Reader mode
You must log in or register to reply here.
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