• 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!

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

Reader mode
Simba1 said:
Don't take PR statements like absolute truth, they are saying things that fits them currently most, for instance like they were saying that Switch is not replacement for 3DS, or no new 3DS model and than 2 week later new model is annouced..
So he will definitely not say it has higher profit margin than regular Switch and in same time has $50 higher price point.

I really have hard time believing that basically OLED screen and 32GB more internal memory makes even $50 difference not to mention more than that,
OLED screens are now cheaper than ever, I mean you now have phones with 4GB/128GB (RAM/internal) and 6.5" OLED screen for around $150.
My guess is that difference cost is maybe around $30.
Click to expand...
Simba1 said:
Well if they lied before, I don't see why wouldnt again, I mean they cant prove it in any case.

Dont forget that Bloomberg reported that Switch OLED costs are only around $10 higher compared to regular Switch,
of course Nintendo deny that report (because raises question why selling than $50 more!?), but I really cant see how can cost $50 more than regular Switch.
Click to expand...
Simba1 said:
Yeah, Switch was selling with profit from day one, now 6 years later, even if Covid and chip shortages made some prices go up,
Nintendo cost currently are probably around $200 (but less than $250 in any case) per Switch unit.

Thats why probably some people expecting $499 price point for new Switch, they think Switch is not selling at profit or with minimal profit margin,
so they think that new Switch hardware need to sell at $499 to be sold with profit.
While reality is that Nintendo has huge profit margin on current models.
Click to expand...
They can't lie to investors. This isn't Miyamoto saying there's no 3DS XL and then Nintendo announcing it a week later. This was a investor Q&A. They are abide to the truth.
And to answer, as people here seems to think current market trends and environment is the same as of old consoles, it's perfectly possible Switch OLED has the same profit margins and the $50 extra was actually needed. Heck, I wouldn't doubt Switch Lite and Switch V2 profit margins have been reduced. Remember, we are in unprecedented times. TSMC 16nm prices has been increased 20% higher than pre-pandemic, shipping costs are 5-6x higher than pre-pandemic and they will increase even more as there's a spike in petroleum/oil/gas prices worldwide, the prices of metals have been increased a lot compared to pre-pandemic, the prices and availability of components have increased and are more scarce, etc, etc, etc.
People need to forget that old thinking "consoles become more cheaper to manufacture as time passes". No, it's not applicable anymore and the savings are very small nowadays or in some cases the costs of producing the machines increase and that affect margins. Microsoft literally stated that's one of the reason they released a cheaper Xbox at launch: Consoles don't become as cheaper to manufacture anymore. There's no savings to be done waiting.
 
Ghostsonplanets said:
They can't lie to investors. This isn't Miyamoto saying there's no 3DS XL and then Nintendo announcing it a week later. This was a investor Q&A. They are abide to the truth.
And to answer, as people here seems to think current market trends and environment is the same as of old consoles, it's perfectly possible Switch OLED has the same profit margins and the $50 extra was actually needed. Heck, I wouldn't doubt Switch Lite and Switch V2 profit margins have been reduced. Remember, we are in unprecedented times. TSMC 16nm prices has been increased 20% higher than pre-pandemic, shipping costs are 5-6x higher than pre-pandemic and they will increase even more as there's a spike in petroleum/oil/gas prices worldwide, the prices of metals have been increased a lot compared to pre-pandemic, the prices and availability of components have increased and are more scarce, etc, etc, etc.
People need to forget that old thinking "consoles become more cheaper to manufacture as time passes". No, it's not applicable anymore and the savings are very small nowadays or in some cases the costs of producing the machines increase and that affect margins. Microsoft literally stated that's one of the reason they released a cheaper Xbox at launch: Consoles don't become as cheaper to manufacture anymore. There's no savings to be done waiting.
Click to expand...

I personally don't believe corporations everything they say, especially when they statement cant be proved, or in this case when I dont see nothing in Switch OLED that says that makes cost $50 higher compared to regular Switch, not to mentioned that Bloomberg reported that cost difference is around $10.

Like I wrote, profit margin was reduced around last year or two, but until than profit margin was getting higher from launch, and Switch was selling at profit from day one.
It most likely that Nintendo doesnt have same profit on Switch units currently like they had year or two ago, but again that should be very healthy profit in any case.
 
It seems pretty hard to believe the actual cost of an OLED screen and additional storage really adds up to $50. Maybe that includes the R&D costs, plus the costs of getting an additional SKU into production.
 
Simba1 said:
I personally don't believe corporations everything they say, especially when they statement cant be proved, or in this case when I dont see nothing in Switch OLED that says that makes cost $50 higher compared to regular Switch, not to mentioned that Bloomberg reported that cost difference is around $10.

Like I wrote, profit margin was reduced around last year or two, but until than profit margin was getting higher from launch, and Switch was selling at profit from day one.
It most likely that Nintendo doesnt have same profit on Switch units currently like they had year or two ago, but again that should be very healthy profit in any case.
Click to expand...
Quick intervention. Just like another poster stated it already, a margin could be a reference to a percentage. In this case, it going down doesn't necessarily mean that the OLED makes less money for Nintendo.

That said, everybody is having trouble with the 2-hit combo resulting from the chip shortage and the Ukraine-Russia war. I can see how Nintendo could make less money from its hardware from now, even with a price hike, like in the case of the OLED model.
 
Branduil said:
It seems pretty hard to believe the actual cost of an OLED screen and additional storage really adds up to $50. Maybe that includes the R&D costs, plus the costs of getting an additional SKU into production.
Click to expand...
It's also the increased cost of components during the pandemic, I'm pretty sure they specifically alluded to that.
 
Now that Oil prices are increasing globaly... (even more than before) Moving stock will BE EXPENSIVE holy shit.
 
Hermii said:
You clearly made your mind up, so I don’t know why im bothering. But your way to hung up about that one quote about AMD engineers figuring out ways Nintendo didn’t think of. That was at the micro level, where they wanted to include a circuit for Wii, but AMD found a way to do it with just 1 circuit. It wasn’t like AMD changes Nintendos mind about BC method, it was just they found slightly more elegant engineering solutions than Nintendo would have thought of.

Nintendo wanted a near 100% accurate BC solution that worked on every game out of the box. And the way they achieved that, was with ppc750 and a custom gpu. End of story.
Click to expand...

And you are way too hung up on them saying off hand they wanted bc, and then claiming this is the only way they would have or could have done it, and the sole reason they stayed with a bunk partner who was losing business because they could not deliver competitive products against Intel, and designed the entire console the way they did. Nintendo was no stranger to bc at the time, they had no issues simply putting an old processor in systems like the gba, or DS.

Not a single person has addressed my actual point yet, which is a mile up the road from this traffic jam that keeps happening. They were already going to be working with ibm and using a ppc750 derivitive long before any talk of how to best implement the bc took place. It was never going to be anything else and bc certainly wasn't the deciding factor. The fact they were still working with ibm was the deciding factor. That would be like saying bc was the sole reason the Wii was designed the way it was.
 
Last edited:
Quoted by: MP!
1
Simba1 said:
Well if they lied before, I don't see why wouldnt again, I mean they cant prove it in any case.
Click to expand...
Lying to investors and/or shareholders is literally illegal. And investors and/or shareholders can literally collectively sue Nintendo for securities fraud if investors and/or shareholders reasonably suspect Nintendo's lying. (I imagine the last thing Nintendo wants to do is to anger investors and/or shareholders to the point where investors and/or shareholders decide to stop investing in Nintendo.)

The media's obviously a different story.
 
Last edited:
Ghostsonplanets said:
Yes. But Nintendo would need to support it. And given the portable screen will probably cap out at 60Hz, I can see them just staying at the trued and test 30/60Hz.
Click to expand...
Forgive a silly question, but is it possible for a screen to support 30/40/60Hz, without needing to go up to 120Hz in portable?

I'm thinking how the PS5 can hit 120fps but they were marketing Ratchet & Clank as supporting 40fps locked because the TVs could output that and it's at least a higher framerate count than 30 without unnecessary screen tearing.
 
Crusters said:
And you are way too hung up on them saying off hand they wanted bc, and then claiming this is the only way they would have or could have done it, and the sole reason they stayed with a bunk partner who was losing business because they could not deliver competitive products against Intel, and designed the entire console the way they did. Nintendo was no stranger to bc at the time, they had no issues simply putting an old processor in systems like the gba, or DS.

Not a single person has addressed my actual point yet, which is a mile up the road from this traffic jam that keeps happening. They were already going to be working with ibm and using a ppc750 derivitive long before any talk of how to best implement the bc took place. It was never going to be anything else and bc certainly wasn't the deciding factor. The fact they were still working with ibm was the deciding factor. That would be like saying bc was the sole reason the Wii was designed the way it was.
Click to expand...
It’s been answered like a thousand times you just keep pretending like it’s not an answer
The real answer for any of this is “no one knows”
 
Brofield said:
Forgive a silly question, but is it possible for a screen to support 30/40/60Hz, without needing to go up to 120Hz in portable?

I'm thinking how the PS5 can hit 120fps but they were marketing Ratchet & Clank as supporting 40fps locked because the TVs could output that and it's at least a higher framerate count than 30 without unnecessary screen tearing.
Click to expand...
yes, it's possible. I don't think anyone would bother though
 
MP! said:
It’s been answered like a thousand times you just keep pretending like it’s not an answer
The real answer for any of this is “no one knows”
Click to expand...

I don't disagree with this. I know we don't know, I really want some kind of light shed on the situation, but I've never claimed we knew the answer. Although I don't see where that has been the answer a thousand times. What I have seen, is a lot of people state they know, and have stated the same reason that doesn't really add up to addressing this.

That being said, while I enjoyed the discourse, and I want to answer as many people as I can (sorry if I missed anyone) so as not to seem rude, it was taking place at a time in this thread where it seemed to me interesting discourse of the subject at hand was in a lull, and now it seems to have picked up
 
Last edited:
Quoted by: MP!
1
Mr Swine said:
Is there any possibility for Drake to run Switch games at 4K through Emulated BC?

The CPU side should be native but the GPU part could be emulated (like SMG on the Mario collection) out is that not even feasible?
Click to expand...
It's very unlikely. It's not something that can be done in a generic way for a console like the Switch.
 
FernandoRocker said:
Let me correct you right there.

If the leaks are true, it seems to be at least 3 times more powerful (in theoretically brute force), and that's without DLSS in play.
Click to expand...

I was talking about portable mode being basically the same.

Every Switch game on this model is going to be developed for the portable mode.
 
Pokemaniac said:
It's very unlikely. It's not something that can be done in a generic way for a console like the Switch.
Click to expand...
What they could do is use AI upscaling when docked to a 4k TV, ideally giving it a better 4k IQ. But yeah I don't see them being able to change the render resolution in BC.

Patched games should get there though.
 
0
Kenka said:
Quick intervention. Just like another poster stated it already, a margin could be a reference to a percentage. In this case, it going down doesn't necessarily mean that the OLED makes less money for Nintendo.

That said, everybody is having trouble with the 2-hit combo resulting from the chip shortage and the Ukraine-Russia war. I can see how Nintendo could make less money from its hardware from now, even with a price hike, like in the case of the OLED model.
Click to expand...
As far as I understand it, a profit margin is a percentage. If you're talking about an absolute amount, then you just call it profit. Nintendo stated that the profit as a percentage of revenue on the $350 OLED model is lower than the profit as a percentage of revenue on the $300 2019 model. As far as I'm concerned, that simply reinforces my belief that Nintendo make a very comfortable profit on Switch hardware, as it tells us that the margin on the $300 base Switch is higher than the margin they make on swapping out the OLED components and charging an extra $50. And the latter isn't really that difficult to calculate. The OLED panel is cheaper than ever, and the eMMC module is a cheap commoditised part. There are a few changes in the dock as well, but that's mostly swapping out cheap commoditised parts for slightly less cheap commoditised parts. Bloomberg may not be exact on the $10, and the chip shortage may have had an effect, but there's no way they're not making a comfortable profit on adding those components. They wouldn't have bothered making the OLED model if they didn't.

And if they're making a comfortable profit on the extra OLED model components, and they've told us they make an even higher profit on the base model, then obviously they must be making a comfortable profit on the base model. Which shouldn't really surprise anyone, given they're selling it at the same price they were at launch, 5 years and 100 million units later. They're not making as much profit as they would have without the chip shortage, but that doesn't mean they're not making a healthy profit.
 
Crusters said:
I don't disagree with this. I know we don't know, I really want some kind of light shed on the situation, but I've never claimed we knew the answer. Although I don't see where that has been the answer a thousand times. What I have seen, is a lot of people state they know, and have stated the same reason that doesn't really add up to addressing this.

That being said, while I enjoyed the discourse, and I want to answer as many people as I can (sorry if I missed anyone) so as not to seem rude, it was taking place at a time in this thread where it seemed to me interesting discourse of the subject at hand was in a lull, and now it seems to have picked up
Click to expand...
You don't seem rude it just seems like you're going in circles... or the discussion is anyway
 
MP! said:
You don't seem rude it just seems like you're going in circles... or the discussion is anyway
Click to expand...

Thanks, the discussion definitely was going in circles, as it wasn't addressing (or speculating) the area I was, the area we don't know about, but an area further down the road.

What I meant about not wanting to be rude was leaving anyone out who responded to me, as though I had fun with the blast from the past discourse, the threads discourse of the subject at hand has picked up and become more interesting again, and I'm more interested in focusing on that again now.
 
There's often talk about how Nintendo will handle software with new hardware hardware--keep most things playable on everything, transition over the course of a couple years, what. Anyway, I was thinking about a theoretical situation where both current Switch and next Switch were coexisting since 2017, which games do I think would release playable on the lower model. There are other maybes, but I figure these are the low hanging fruit, the definites, the kind I can see still releasing on base Switch five years after new hardware arrives.

Snipperclips
1-2 Switch
Flip Wars
Super Mario Maker 2
Ring Fit Adventure
Labo
Brain Training
Clubhouse Games
Famicom Detective Club
Game Builder Garage
Big Brain Academy
Advance Wars 1+2
Switch Sports

Rereleases/multiplat (NES/SNES/N64, BOTW, MK8, Pokkén, Bayonettas, DKC TF, Hyrule Warriors 1, Captain Toad, NSMBU, TMS#FE, SM3DAS, Pikmin 3, SM3DW, Miitopia, SSHD, Sushi Strikers, Pokémon freemiums)
 
0
Crusters said:
Thanks, the discussion definitely was going in circles, as it wasn't addressing (or speculating) the area I was, the area we don't know about, but an area further down the road.

What I meant about not wanting to be rude was leaving anyone out who responded to me, as though I had fun with the blast from the past discourse, the threads discourse of the subject at hand has picked up and become more interesting again, and I'm more interested in focusing on that again now.
Click to expand...
Speculating about the business relationship of Nintendo and IBM is kind of hard when there is nothing to go on.

None of us were at the Wii U engineering team and knows exactly the rationale for every decision, we just have to take the publicly stated reasons as what they are.

Could just including the Wii chipset on the die have resulted in a better product in the end? Quite possibly yes. But no way to know for sure.
 
Hermii said:
Speculating about the business relationship of Nintendo and IBM is kind of hard when there is nothing to go on.

None of us were at the Wii U engineering team and knows exactly the rationale for every decision, we just have to take the publicly stated reasons as what they are.

Could just including the Wii chipset on the die have resulted in a better product in the end? Quite possibly yes. But no way to know for sure.
Click to expand...

Accurate and responsible speculation about the business relationship is hard.

Scintillating and sensational speculation about the business relationship is much easier.
 
0
Just in the context of chip costs, I think it's quite interesting that Apple announced the new iPhone SE yesterday, and in particular that it's using the A15, which is a ~15 billion transistor chip on the TSMC N5 process, and starts at $429. Obviously this isn't directly comparable to Nintendo, as Apple use their own SoCs, rather than buying them from a vendor like Nvidia, which brings down costs. It's also got relatively low-end specs outside the SoC, in terms of RAM, screen, camera, etc, but Apple's expectations on profit margin are going to be a lot higher too.

Anyway, I just thought it was a pretty interesting that using a 15 billion transistor SoC on a bleeding edge node in a $429 phone is actually a thing. Particularly so from a company like Apple, who would usually have pretty high profit margins on their hardware.
 
Thraktor said:
Just in the context of chip costs, I think it's quite interesting that Apple announced the new iPhone SE yesterday, and in particular that it's using the A15, which is a ~15 billion transistor chip on the TSMC N5 process, and starts at $429. Obviously this isn't directly comparable to Nintendo, as Apple use their own SoCs, rather than buying them from a vendor like Nvidia, which brings down costs. It's also got relatively low-end specs outside the SoC, in terms of RAM, screen, camera, etc, but Apple's expectations on profit margin are going to be a lot higher too.

Anyway, I just thought it was a pretty interesting that using a 15 billion transistor SoC on a bleeding edge node in a $429 phone is actually a thing. Particularly so from a company like Apple, who would usually have pretty high profit margins on their hardware.
Click to expand...

I think it's somewhat directly comparable right? Semi direct? Nvidia and Nintendo are in a multi year partnership, so while Nintendo aren't doing it themselves like apple, I would think it would be more cost effective than if they were just shopping around with vendors they don't have a partnership with.
 
Dakhil said:
Just to offer a small correction, the Apple A15 Bionic's fabricated using TSMC's N5P process node, not TSMC's N5 process node.
Click to expand...

Thanks for the clarification, it's even more bleeding-edge than I thought!

Crusters said:
I think it's somewhat directly comparable right? Semi direct? Nvidia and Nintendo are in a multi year partnership, so while Nintendo aren't doing it themselves like apple, I would think it would be more cost effective than if they were just shopping around with vendors they don't have a partnership with.
Click to expand...

Yeah, I'm sure Nintendo get a much better deal than if they were just buying "off the shelf", but there's still some margin for Nvidia in there to make the deal desirable from their end. So if, in the purely hypothetical scenario that Drake is manufactured on TSMC N5P and is about the same size as A15, Nintendo would be paying more than Apple do for the A15, but I don't know if it's possible to say by how much.
 
Crusters said:
Nvidia and Nintendo are in a multi year partnership, so while Nintendo aren't doing it themselves like apple, I would think it would be more cost effective than if they were just shopping around with vendors they don't have a partnership with.
Click to expand...
To play devil's advocate, always being in contact with different companies who Nintendo could potentially work with in the future is always a good idea, especially since Nintendo's partnership with Nvidia can always end abruptly at any time for any reason. And I imagine at least one company Nintendo's always in contact with is a company Nintendo has worked with before (e.g. AMD).
 
Dakhil said:
To play devil's advocate, always being in contact with different companies who Nintendo could potentially work with in the future is always a good idea, especially since Nintendo's partnership with Nvidia can always end abruptly at any time for any reason. And I imagine at least one company Nintendo's always in contact with is a company Nintendo has worked with before (e.g. AMD).
Click to expand...

That makes a lot of sense. But doesn't really seem like a devil's advocate for this situation: cost of purchasing a chipset from Nvidia as a unrelated vendor, vs cost of purchasing a chipset from Nvidia as a partner, but rather a competitor offering a sweetened deal to steal Nintendo away from Nvidia. I mean that certainly could be more cost effective, but an exceptional situation outside the bounds of the comparison.

Wait can they even do that? Would Nintendo breach some kind of contract if they exited Nvidias partnership? Wasn't it a seven year deal?
 
Following on from the A15, I thought it would be interesting to consider what kind of die size we might be looking at for Drake on the different manufacturing processes that are available. Irrespective of cost, power consumption, feasibility or whatever, just "how big would it be".

We obviously don't know how many transistors Drake is, but for the sake of a ballpark figure, I'm going to go with 10.5 billion, for no other reason than it's exactly half of Orin, and that just feels about right to me. The GPU is only 25% smaller, and the CPU perhaps 33% smaller, but we'd be completely eliminating stuff like the DLA and PVA accelerators, and there would be lower-level savings from the single GPC, reduced tensor core size, etc. It's a super-rough ballpark figure, but enough to give a very rough idea of what we could be looking at.

For the transistor densities I'm not going to use the advertised densities, or the densities achieved by mobile SoCs, as Nvidia typically aren't designing purely for maximum density, and you can see on Samsung 8nm and TSMC 7nm processes that they aren't hitting the same density as the mobile SoCs on the same nodes do. For the Samsung nodes, I'm taking the 8nm density as 45.6 MT/mm2 (same as Orin), and then using the density improvements from this Samsung slide. Foundry density claims aren't necessarily 100% reliable, but for this one at least it's a consistent real-world logic block they're comparing from 10nm to 4nm, so there isn't the issue of mixing and matching scaling factors which may be measured in very different ways. Samsung also don't specify the exact process here (eg 5LPE vs 5LPP), so I'll just refer to them as Samsung 5nm, etc.

For TSMC N7, I'm using 65.6 MT/mm2, as it's the density of Nvidia's A100. For N6 I didn't have a real-world comparison, so I used TSMC's claim of 18% density improvement over N7 (real world may be lower). For N5, I compared the relative density of Apple's A12 (N7) and A14 (N5), and then applied that scaling factor to Nvidia's A100 transistor density. Effectively, I assume that Nvidia won't hit the same density as Apple on N5, but they will achieve similar scaling from N7.

So, the very, very rough, please-don't-take-too-seriously die sizes would be as follows:

ProcessDensity (mT/mm2)Drake size (mm2)
Samsung 8nm45.6230.3
Samsung 7nm59.2177.3
Samsung 5nm83.4125.9
Samsung 4nm109.795.7
TSMC N765.6160.1
TSMC N677.4135.6
TSMC N5106.199.0

On Samsung 8nm it would be (unsurprisingly) a pretty big die. On Samsung 5nm or TSMC N6, though, it's not necessarily that far off of the original TX1 die size (121 mm2), and I wouldn't rule out either of those nodes at this stage.
 
FernandoRocker said:
I think they mentioned they have a 20-year partnership.

Edit: Correction. Nvidia expects a 20-year partnership.

www.pcmag.com

Nvidia Expects 20-Year Relationship With Nintendo Thanks to Switch

Nvidia CEO Jen-Hsun Huang calls the Switch an 'incredible game console.'
www.pcmag.com www.pcmag.com
Click to expand...
Dakhil said:
Jensen Huang only said he thinks Nvidia's partnership with Nintendo is likely to last two decades. So no one knows exactly.
Click to expand...

Holy crap, that's a commitment.

So.... What happens if Nintendo wants out like 10 years down the line? Like, Do they have to pay a fine? get a divorce? Get a spanking or.... What?

Are there any typical examples to look to?
 
Crusters said:
Holy crap, that's a commitment.

So.... What happens if Nintendo wants out like 10 years down the line? Like, Do they have to pay a fine? get a divorce? Get a spanking or.... What?

Are there any typical examples to look to?
Click to expand...
there doesn't seem to be any hard agreement. if Nintendo bounces, that's it. they walked out for milk and never returned
 
Thraktor said:
Following on from the A15, I thought it would be interesting to consider what kind of die size we might be looking at for Drake on the different manufacturing processes that are available. Irrespective of cost, power consumption, feasibility or whatever, just "how big would it be".

We obviously don't know how many transistors Drake is, but for the sake of a ballpark figure, I'm going to go with 10.5 billion, for no other reason than it's exactly half of Orin, and that just feels about right to me. The GPU is only 25% smaller, and the CPU perhaps 33% smaller, but we'd be completely eliminating stuff like the DLA and PVA accelerators, and there would be lower-level savings from the single GPC, reduced tensor core size, etc. It's a super-rough ballpark figure, but enough to give a very rough idea of what we could be looking at.

For the transistor densities I'm not going to use the advertised densities, or the densities achieved by mobile SoCs, as Nvidia typically aren't designing purely for maximum density, and you can see on Samsung 8nm and TSMC 7nm processes that they aren't hitting the same density as the mobile SoCs on the same nodes do. For the Samsung nodes, I'm taking the 8nm density as 45.6 MT/mm2 (same as Orin), and then using the density improvements from this Samsung slide. Foundry density claims aren't necessarily 100% reliable, but for this one at least it's a consistent real-world logic block they're comparing from 10nm to 4nm, so there isn't the issue of mixing and matching scaling factors which may be measured in very different ways. Samsung also don't specify the exact process here (eg 5LPE vs 5LPP), so I'll just refer to them as Samsung 5nm, etc.

For TSMC N7, I'm using 65.6 MT/mm2, as it's the density of Nvidia's A100. For N6 I didn't have a real-world comparison, so I used TSMC's claim of 18% density improvement over N7 (real world may be lower). For N5, I compared the relative density of Apple's A12 (N7) and A14 (N5), and then applied that scaling factor to Nvidia's A100 transistor density. Effectively, I assume that Nvidia won't hit the same density as Apple on N5, but they will achieve similar scaling from N7.

So, the very, very rough, please-don't-take-too-seriously die sizes would be as follows:

ProcessDensity (mT/mm2)Drake size (mm2)
Samsung 8nm45.6230.3
Samsung 7nm59.2177.3
Samsung 5nm83.4125.9
Samsung 4nm109.795.7
TSMC N765.6160.1
TSMC N677.4135.6
TSMC N5106.199.0

On Samsung 8nm it would be (unsurprisingly) a pretty big die. On Samsung 5nm or TSMC N6, though, it's not necessarily that far off of the original TX1 die size (121 mm2), and I wouldn't rule out either of those nodes at this stage.
Click to expand...
More fuel to the "Dane was 8nm then they changed to Drake which is 5nm" fire.
 
Skittzo said:
More fuel to the "Dane was 8nm then they changed to Drake which is 5nm" fire.
Click to expand...
Or it was just never 8nm at all. Had they started design as an 8nm chip, they would have known it was going to be well over 200mm2 from the beginning, and power consumption can't have been that much of a surprise either. If it turns out it's not 8nm, then my guess is that it never was, and kopite7kimi just mistakenly assumed that because it's in the same family as Orin, it would use the same process.
 
Thraktor said:
Or it was just never 8nm at all. Had they started design as an 8nm chip, they would have known it was going to be well over 200mm2 from the beginning, and power consumption can't have been that much of a surprise either. If it turns out it's not 8nm, then my guess is that it never was, and kopite7kimi just mistakenly assumed that because it's in the same family as Orin, it would use the same process.
Click to expand...

I’m guessing that SoC’s made on Samsung 8nm is more than enough for being in cars (better cooling perhaps?) while a handheld console it’s more important with a smaller node I guess?
 
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