If they did change modes, it probably wasn’t a 12sm chip when it was 8nm. Then at some point, they thought it was worth it to delay for a more powerful chip.
-
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!)
- Thread starter Dakhil
- Start date
While that’s true, there are other major considerations that still make 8nm attractive. Price and capacity. Nintendo wants to produce millions of these during a chip shortage, and ideally make a profit on each unit.
- Pronouns
- He/Him
If Dane was 8nm then it likely didn't have anywhere near 12SMs. Probably 6 at the very most.
Crusters
Koopa
- Pronouns
- He/him
Pretty sure this is just a size thing.
Fitting those dimensions somewhere in a car with all it's stuff isn't really an issue compared to fitting inside something roughly the size of a switch, with all the stuff that also needs to be in there.
- Pronouns
- He/Him
I don’t understand this comment. Where are you getting 50% from.
Yes, but I think only in docked mode
I think it can in theory, depending on the CPU, play possibly Genshin Impact at the 120 setting (not necessarily hit it all the time) but in docked mode, again.
Portable mode they’d need to have the screen for that. Which I doubt.
I think it’s hard to say really as the A15 Bionic has the 5G modem which Apple has to pay royalties for to Qualcomm if I’m not mistaken for every soc they sell.
So it could be about the same or lower since I hear it isn’t cheap to pay for that modem!
Last edited:
ShadowFox08
Paratroopa
whats the performance gains and power draw savings in NP5 vs regular 5nm?
Thraktor
"[✄]. [✄]. [✄]. [✄]." -Microsoft
- Pronouns
- He/Him
Higher transistor density generally goes hand-in-hand with better power efficiency. It's not an exact relationship, and there are differences between foundries, with TSMC's processes generally being quite a bit more efficient than Samsung's, all else being equal.
I feel like if there was a change from 6 or 8 SMs on 8nm to 12 SMs on 5nm, then that's functionally a new chip design and that would be reflected by a change from the T239 identifier. Having a whole load of code and design files that refer to an 8nm 6 SM part named T239, and then having a whole other load of code and files referring to a 5nm 12 SM part as T239 just seems like a recipe for confusion. We'd also surely see the old version of the chip referenced somewhere in the leak, but the T239 with the 12 SM GA10F GPU is the only new chip in there. The name Dane is in there once under a list of codenames, but without linking it to any particular chip. For all we know someone just added it in after kopite's tweets without realising it's not actually one of their products.
I think it’s hard to say really as the A15 Bionic has the 5G modem which Apple has to pay royalties for to Qualcomm if I’m not mistaken for every soc they sell.
So it could be negligible or slower since I hear it isn’t cheap to pay for that modem!
That's a good point, I had forgotten about the modem. It's not actually part of the A15, it's a separate chip, but a would certainly add to the cost of the iPhone SE alright.
OP
OP
My Tulpa
Bob-omb
That’s…literally the graph i referred to in my post.
And it crystallizes my point…the R&D spent during Drake creation isn’t as big of a deal as the poster I was responding to was suggesting.
The size of R&D budget doesn’t suggest something HAS to be treated as a true successor hardware, any more than spec differentials do.
The % of revenue stat should be obvious, which that graph also uses it to point at…while R&D costs and earned revenue both go up…Nintendo is actually using less % of their revenue on R&D
This is important.
Especially if the argument you are making is that Nintendo is spending so much money on R&D for Drake, Drake has to be positioned as a next gen console! They wouldn’t spend all that money if it wasn’t!
They are pulling more in than they are spending compared to when they were making Lite and TX1+ revision hardware. They can easily treat Drake as a mid gen revision model and not sweat about the R&D any more than they usually do.
They care very much about their games running at native resolutions to the screen/tv and at the same refresh rate as the screen/tv.
This Drake will certainly alleviate that and give them headroom to play with other graphics design they do care a lot about like lighting.
You are asking me to believe Nintendo went to Nvidia and told them they need to have Orin/Ampere and tensor and RT cores! What? 6SM’s?? No way, we want 12SM’s! What? 8nm? No way, give us the good stuff, let’s go 5nm! We need all that juicy power because we need it to make BotW3 and Prime 4 better! We have so many ideas beyond just playing Switch games at steady resolutions and framerates!
Hyperbole aside, i think these choices came more as happenstances during design mode and I firmly believe Nvidia incentivized them.
I think Nvidia told them how DLSS was the future of gaming where you never have to render at native resolutions ever again…that you can have lower variable resolutions and gain performance…it’s almost as if this was designed with Nintendo in mind. Nintendo immediately hired people and delved into it. I’m sure they wanted to start the process cause they see that as the future of their Switch design as well.
If Nintendo are going to start doing this (as I’ve mentioned, they lament they should have moved over to HD development sooner during the Wii years), don’t they kind of have to use ampere? Wouldn’t Nvidia say hey, we have new ampere mobile SoC with enough tensor cores to actually make this work!
So they kind of have to use a variation of Orin when discussing this in 2019 or so, no?
And then they try to figure out what’s the minimum they can do and still get a reasonable function of DLSS with 4K/60fps output for…say…the BotW2 game they are building or the new Monolith/Platinum game.
But it has so be able to clock extremely low while doing. Like, appealing to Nintendo’s standards of really low heat and power draws.
Doesn’t going wide make sense to this? Couldn’t the design process and prototypes have had trouble, maybe running too hot? (Wasn’t there suggestions by insiders that there were issues like this?)
So maybe, in the end, 12SM’s and moving to 5nm (if true) was a necessity in the end, rather than an ask.
As far as money, again as I stated above, Nintendo is in a place of clear success now and eyeing for even more growth and investment in this ecosystem. Paying for these changes is easier now than it would have been in…say…2016.
So, when people tell me this is wayyyy too powerful to be just used like it’s just an RTX 2070 for Nintendo games…offering to game with better resolution and performance than you were getting with some headroom for higher graphics sliders and some light ray tracing…I kind of shrug about it because I don’t think it’s extravagantly too powerful to be used as just that in a Switch form factor.
Look, I get why people don’t like me saying Nintendo is going to be conservative on how they use this right now…but people wanted “a catch” to these leaks seeming too good to be true? Here it is. It’s not going to be used and positioned by Nintendo how you think this hardware should be.
I do believe Nintendo is interested in trying some offshoot games utilizing the tensor cores for some unique gameplay, don’t get me wrong.
- Pronouns
- He/Him
Out of curiosity, but how did you get such a large size for 8nm?
Even the GA106 is at 276mm^2 and that has 30 SMs while GA10F only has 12SMs, and the GA106 is actually less dense than the ORIN SoC (43 vs 45MTr).
Even accounting for 8 big CPU cores (A78), and the larger cache, I’m not sure how you got a size so large. Do you mind giving the breakdown?
Are you accounting perhaps the more memory channels?
Or is it just a raw calculation based on being 10.5B Transistors?
Also yes I noticed the bolded
, and RennaNT explained it was simply based on being 10.5B not really based on a specific thing! It’s clear.
Last edited:
RennanNT
Bob-omb
Not Traktor, but it is purely based on 10.5B and the node density.
10,500/45.6 = 230.3
10,500/59.2 = 177.3
and so on.
Ghostsonplanets
Ghost in the Sales
- Pronouns
- He/Him
GA107 is a Samsung 8nm 20SM die and it's 19xmm²(x means a number between 1 and 9). I think this estimation for 12 SM + maybe 6 - 8 A78 cores seems too big?
I gotta agree that Samsung 8nm doesn't seems well suited for a gaming portable SoC though. From both a energy consumption and area PoV. I really hope Nintendo agree and is fabbing it at a more modern node.
I gotta agree that Samsung 8nm doesn't seems well suited for a gaming portable SoC though. From both a energy consumption and area PoV. I really hope Nintendo agree and is fabbing it at a more modern node.
Last edited:
Mr.Gamerson
Octorok
I figured it out guys! There is an 8nm Dane for the eventual console variant and then a 5nm version of the chip for the hybrid model and handheld only version. This will allow them to sell the home console only version at $299 because of going with 8nm. The Hybrid model will retail at $450 to account for using a newer node. /s
- Pronouns
- he/him
Yo at least put down a towel if you’re gonna leak like that
My Tulpa
Bob-omb
Pretty sure I read an interview with the studio that ported the Switch version and they said Witcher 3 took about 1 year? That’s pretty impressive.
They will announce it a few months ahead and drop it, like Lite and OLED. They tend to build up the next gen console hardware, not revisions of a system.
You had me at 2000.
Different world man. Consoles now are iterative and BC and previous gen libraries aren’t locked to previous gen consoles.
Try looking at console sales in the past 20 years. Ps3 selling 3 million after the next gen model released is the best case scenario.
LiC
Member
I've also thought of this as a plausible explanation. If it was in every list I might think differently, but it's only in one. And even if it's a reference to a scrapped chip, I don't think they would remove it from their scans. There are still things like "mariko" on those lists -- if something isn't public, it can still be on there even if it's from some years past.
This is not a response to @Simba1 post. I just want to comment on that oft-cited Bloomberg report. Many forgot that it was a pre-release scoop, not a post-release analysis. The $10 cost increase was merely a supposition by Bloomberg's source(s) based on the incomplete info suggesting that the OLED Model only differs in the display and memory. After the model was actually released, we all learned that there are other improvements: better wifi and bluetooth antennas, stronger kickstand, partially metallic enclosure, various new chips, etc. That $10 delta estimation immediately fell apart. Even if we put that flimsy premise aside, the notion that a 7" OLED panel (custom made by Samsung, not some B-tier manufacturer) only costs a few dollars more than a 6.2" LCD panel was extremely dubious.
Edit: typo
Last edited:
- Pronouns
- He/Him
I don’t think his sources would know exactly as they seem to be software developers, and they probably assumed 8nm only because Ampere is already on 8nm. Otherwise they have no idea of how to know that unless it came out that the die size and the number of features are optioning to a denser node, unsure of which exactly, but simply a node other than 8nm. Not to mention, these are probably earlier devkits that are based on the Tegra ORIN (234) and not the final silicon which could be 8, 7, 6, 5 or 4nm(doubt on this one).
ShadowFox08
Paratroopa
I wonder how the dev kits look like. Could it be small PC box with an 8nm ampere gpu card.. or an actual early module of dane/Orion nx? I can't imagine the latter, or that they would know the node size of the final product. Though early kit of the original Orion maybe... Dev kits using 8nm is certainly likely, but final node design? Not necessarily.
Thugstas
Bob-omb
last year yes....
- Pronouns
- He/Him
Considering it is apparently portable in the current devkits, I think there’s reason to believe it has an orin chip inside it that was modified for the devkit. If it’s releasing this year then the devkits have to look more closer to the final silicon at least in form similar to what happened with the PS5 Ariel to Oberon.
ie, it has to be able to be taken off for optimization in portable mode.
If they changed the process node recently, along with a major redesign of the chip, that would guarantee a long delay.
Thugstas
Bob-omb
we will see.
Z0m3le
Bob-omb
It would have happened in the very beginning of last year, given that is when Nvidia had gotten TSMC's 5nm secured (reports are from February 2021. That gives Drake around 2 years for a node shrink... Mariko chips from August 2018 exist in retail units, just a year and a half after Switch. the idea of a delay is actually not likely, we've even seen Nvidia change Orin from 17B transistors to 21B in a similar time frame. We need to stop looking for the doom and gloom here so openly, it's fine to expect something to fall, it's another thing to create a scenario with such confidence. Like even though I believe they moved to 5nm now, I am still very open to the reality that that is an assumption, we can't be sure of anything that isn't in the data leak directly.
ShaunSwitch
Moblin
Looking at Thraktors estimated chip size, if Drake is a similar size to the TX1 on Samsung 5nm is it reasonable to expect around V1 switch battery life at around (or slightly higher) clocks? Just thinking, its likely Nintendo chose portable clocks with an optimal balance between battery life and performance in mind, and I understand there is a floor for chip voltage. So I would imagine in this instance the delta between portable and docked will be similar, unless Nintendo does indeed turn off some of the SM, it will be interesting to see how this unit is configured and marketted, I can see a couple of options.
If its going to just play switch games but at higher fidelity, then could we see a 1080p OLED display and the difference between docked and portable profiles is the same as current switch.
If its designed to play both switch games and exclusives, it may retain portable clocks as is, but have more aggressive cooling to hit higher clocks in docked mode to increase the gap. We have evidence for this in how the OLED dock can deliver more power to the switch. In this case still a 720P screen.
We also have the option to disable SM in portable mode, this would have a larger performance gap between docked and portable, 720p screen again.
I am leaning more towards option 2, I think it makes the most of the powerful silicon Nintendo Would be deploying for this model. There is just the question around cooling really as I expect if clocks go above 1.2GHZ for GPU the console will start to get quite warm to touch when taken out of the dock.
If its going to just play switch games but at higher fidelity, then could we see a 1080p OLED display and the difference between docked and portable profiles is the same as current switch.
If its designed to play both switch games and exclusives, it may retain portable clocks as is, but have more aggressive cooling to hit higher clocks in docked mode to increase the gap. We have evidence for this in how the OLED dock can deliver more power to the switch. In this case still a 720P screen.
We also have the option to disable SM in portable mode, this would have a larger performance gap between docked and portable, 720p screen again.
I am leaning more towards option 2, I think it makes the most of the powerful silicon Nintendo Would be deploying for this model. There is just the question around cooling really as I expect if clocks go above 1.2GHZ for GPU the console will start to get quite warm to touch when taken out of the dock.
Last edited:
- Pronouns
- He/Him
Just because they’re making more money doesn’t mean they have money to waste and are gonna make it rain for the sake of it. If we want to talk about actual un-Nintendo behaviour, that ranks really high up there. Making more money doesn’t make the amount you’re spending immaterial, especially if the money you spend isn’t being or will not be adequately made back commensurate to that expenditure.
An SoC with a Volta-based GPU would give them headroom, too, even accounting for the need to scale back the standard Xavier SoC config it would be based on to reduce die size (its base config is frankly more than what is necessary and, like the base configurations of Orin, Xavier features too many autonomous vehicle components that they could gut out and/or replace). As a matter of fact, my first post in this thread was confusion about the mention of Orin, as I expected a Xavier derivative alone to be next-gen enough for Nintendo’s next hardware when compared to the TX1 and (at the time) seemed a more realistic choice were it not for the persistent rumours of Orin.
Volta features Tensor cores that enable DLSS.
It‘s well past sampling with tested real-world performance figures to work with, having everything they need to up-sample Switch games and more.
They could have used a perfectly acceptable GPU for a custom SoC that’s cheaper, already commercialized, on a smaller process node and still a major step up from the TX1 for a piece of stop-gap hardware.
But even if we concede to the idea that Orin was somehow the only choice or the easiest/cheapest method to get what Nintendo wanted, this is going to be a custom SoC either way, so the core count on the GPU that's being bandied about (1536, 6 times the GPU cores of the TX1) is, given what your expectations are, extravagant to the point of being grotesque.
To achieve the goal of an iterative revision you envision (4 year lifespan before a true successor, upscaling Switch games to 4K, GPU power to spare to try out graphics techniques in a tiny handful of exclusives for said iterative successor), they could easily slash that core count in half and STILL achieve ALL of that, while having even more battery life improvements by virtue of the smaller process node providing a smaller die size (possibly smaller than Mariko is) and kicking off less heat in portable mode. And why on earth wouldn't Nintendo want a 4K Switch AND more battery life in portable mode AND more room in the chassis due to a smaller SoC die size?
But again, that's not what we appear to be getting. There was a cheaper (and in some respects, better and more consumer-positive) way to achieve everything you're suggesting, no matter which way you look at it. The spec information we're gleaning from the leak is incredibly excessive for such requirements, with cheaper and more battery-efficient options Nintendo could easily take on a custom design, since no one (not 3rd parties, not Nintendo) are going to jump at the chance to use all that horsepower making more than the tiniest handful of exclusive titles for some iterative revision with the constant install base cap of 20% of all hardware sold that will be outmoded in 3-4 years.
Your hyperbole suggests that Nvidia had the power to demand Nintendo make a chip with far more power than its intended goals, when it's the other way around; Nintendo is the client in this scenario, a business relationship Nvidia wants to keep for another 15 years, there are ways to give them what they want if your suggestion is all they wanted it to be, so if they wanted a device that just upscaled Switch games to 4K and Drake as we understand it is what they're getting from Nvidia, then Nintendo's not that good of a hardware designer and Nvidia swindled Nintendo like a bunch of rubes.
With what we are hearing about the leaks, and knowing this would be a custom SoC either way, the design they're getting is a waste of money, an unnecessary increase to power draw and too big in die size if it's only to achieve what you believe Nintendo wants this new hardware to be.
Nintendo have control of what they're getting from Nvidia. So if what we are seeing from the leak is what Drake is? That's not happenstance or some design quirk, it's completely on purpose. And this purposeful design is way more than what's necessary for an iterative revision.
Last edited:
What’s being suggested here isn’t a node shrink. It’s more like a complete reset/ rethink of the entire chip layout. Substantially higher sm count/ power budget for the cpu/ gpu, more Rt cores, possible change from Orins tensor cores to desktop ones who knows what Drake would have looked like at 8nm. Probably something closer to what we have been speculating here this entire time.
To me it sounds more likely that Drake was never an 8nm chip. Or maybe it is indeed still a really large chip at 8nm. The future will tell.
Thraktor
"[✄]. [✄]. [✄]. [✄]." -Microsoft
- Pronouns
- He/Him
A few weeks ago I would have absolutely said that Nintendo would choose the minimal viable GPU for DLSS, but now I don't think that's the case. It's not just the size of the GPU, but their choice of tensor cores. Specifically, when Nvidia designed Orin, which is heavily focussed on machine learning, they designed tensor cores which operate at double the performance of the desktop Ampere variety. When Drake was designed they obviously would have had an option to use the same double-rate tensor cores from Orin, but we know from the leak that they didn't, and are using tensor cores with the same performance as desktop Ampere. That suggests to me that they weren't designing this around maximising tensor core performance for DLSS, but rather were designing a chip with a substantial increase in graphics performance in general, with tensor cores (and therefore DLSS) on top.
It's just a raw calculation based on 10.5B transistors.
Yeah, it's a super rough estimate, but I was basically just looking at the Orin die photo Nvidia showed (which is a couple of years old and may not have even been accurate at the time it was shown) and thinking if you chopped 25% of the GPU, 33% off the CPU, and cut out a lot of the extra logic (DLA, PVA, etc.), how big a die would you be left with. It could well be off by a couple of billion transistors.
- Pronouns
- He/Him
August 2018? Did you mean 19?
I think if it were on 5nm, the voltage scaling goes lower for the same clocks. To a point that is.
And I don’t think they’ll go that high for the GPU. IMO, 460-510MHz and 920MHz-1020MHz, so slightly lower, but to allocate the other resources to the CPU to be as performant as it could be.
So, something like 2.1GHz for the CPU if it’s an 8Core A78 setup (1clocked lower for energy saving concerns and being the OS core.)
Z0m3le
Bob-omb
if it's 5nm, I think the GPU could be clocked higher, otherwise they wouldn't make it that big and just use a higher clock... docked clocks is probably ~1.2GHz for 3.7TFLOPs, the reason is simple... the power curve would naturally move to higher clock speeds, so they could use a smaller GPU at a higher clock, unless the clock is already pretty high and their goal is actually XBSS style performance before DLSS is applied, which an improved Ampere at near 4TFLOPs would get close to XBSS.
- Pronouns
- He/Him
I wasn't even aware of this when I responded, so thanks for that.
Is the likelihood here that Orin's tensor cores are just less power-efficient and offer more deep learning strength than any use case Nintendo would ever use them for? After all, those tensor cores in the standard Orin configs are designed for mission-critical use cases (medical robotics, autonomous vehicles, etc.) and DLSS (or frankly any AI usage for an entertainment product) doesn't demand the kind of absolute rigorous precision Orin's typical use cases would demand to achieve the desired result, so it would stand to reason in my mind that they draw more power and offer way more performance than Nintendo would ever need, while maybe also costing more.
Basically, would it make sense than they went with desktop Ampere tensor cores for the same reason we're all but certain the SoC won't use an A78AE CPU, that it doesn't make sense since it would add needless expense and/or power draw to a product meant to entertain?
It depends on the number of SM. When we speculated the chip was 4 to 8 sm, the performance you get from regular desktop level tensor cores per sm would have been barely/ not sufficient. In this case, the extra performance from orin tensor cores would be much needed.
Thraktor
"[✄]. [✄]. [✄]. [✄]." -Microsoft
- Pronouns
- He/Him
Orin's tensor cores don't offer higher precision, and they don't make any claims that they're more reliable in any way, they're pretty much just the same thing, but run twice as many operations in one cycle. Their autonomous driving systems typically pair Orin with standard GPUs, anyway, so if standard Ampere weren't precise or reliable enough for autonomous driving, then that would certainly be a problem.
One possible issue is that DLSS couldn't effectively use that extra performance because of other bottlenecks (probably on the cache/memory end of things), but I'd imagine the solution there would be to alleviate those bottlenecks with increases in cache size, etc., which they've already done on Orin, and seemingly on Drake too.
- Pronouns
- He/Him
So we keep talking about bigger dies in term of die area, which obviously makes a ton of sense. If we are to entertain the rough calculation from Thraktor, 230mm^2 seems like a lot more than 121mm^2.
However, in terms of how much more space that requires inside the case we're only talking about 4 more millimeters in each direction. ~5 if we account for packaging. Is there really not 5 more millimeters that could be found in each direction inside the Switch OLED casing? Not to mention it is certainly possible this device is wider and thicker by a small amount, as the OLED dock allows a bit of variance in those dimensions.
I think the power usage argument makes a lot more sense for why 8nm doesn't work. Plus price of the die. I think they can fit it in there without too much difficulty.
However, in terms of how much more space that requires inside the case we're only talking about 4 more millimeters in each direction. ~5 if we account for packaging. Is there really not 5 more millimeters that could be found in each direction inside the Switch OLED casing? Not to mention it is certainly possible this device is wider and thicker by a small amount, as the OLED dock allows a bit of variance in those dimensions.
I think the power usage argument makes a lot more sense for why 8nm doesn't work. Plus price of the die. I think they can fit it in there without too much difficulty.
- Pronouns
- He/Him
Orin Drive is expected to be upgradable via software to Level 5 autonomous driving (meaning no attendant human required) according to Nvidia, so that’s what I mean by precision, enough AI strength to make that claim (which Nvidia hadn’t before Orin), and that kind of machine learning capacity would be derived (at least in VERY large part) from these new and improved tensor cores you mentioned and their deep learning capabilities.
But Nintendo makes video games, the accuracy required for Level 5 autonomous driving just isn’t necessary for them, their machine learning use case can absolutely afford to make a mistake. And if Orin’s better tensor cores aren’t efficient for Nintendo’s needs on a performance per watt level and do far more than what they need while the tensor cores in desktop Ampere chips are enough and better on power use... it’s a no-brainer, right?
PedroNavajas
Boo
- Pronouns
- He, him
I do believe that a large 8nm is possible. The main argument against a low clockled large chip is price, but 8nm should be a relative cheap node in the first place. I do not think that the size itself is an impediment. Area scales in quadratic relation with vertical and horizontal size. Just ~22% increase in linear dimensions get you a ~50% of total area increase and an 100% one gives a 400% of area for example.
Last edited:
I can not really see something that would stop Nintendo from using 5nm if they could use it. There are so many benefits.
Really exciting times. After months of silence we are finally back on discussions. Yes we are going in circles sometimes, but that's better than nothing. And right now the possibilities are high to get more information soon.
Really exciting times. After months of silence we are finally back on discussions. Yes we are going in circles sometimes, but that's better than nothing. And right now the possibilities are high to get more information soon.
LoneRanger
Chain Chomp
- Pronouns
- He/Him
I will be “pessimistic” but even if they go for a 5nm, I always expect to have the same clocks as OG Switch. With 12 SMs, that is a very good jump in power and enough for Nintendo’s developers vision. (And also for lots of third parties).
- Pronouns
- He/Him
There are several downsides to them using 5nm. Increased cost of R&D, decreased availability, potentially delayed launch if they indeed changed from 8nm originally (not guaranteed), potentially worse yields (not as mature of a process)...
Currently I'm still expecting 8nm since we have no concrete evidence pointing to 5nm. But 5nm would be nice.
- Pronouns
- He/Him
I think at this point, considering this type of jump, there’s nothing that guarantees they see that as a sufficient jump
.They could adjust the clocks before launch and take it to 3.7TFLOPs for all we know.
May I remind you, the switch launched on 20nm in 2017. That was considered a worse node for mobile tech at the time, than 8 is now.
Yes, but switch was originally planned for 2016 and the SoC was designed even earlier.
8nm would be great too. But than Nintendo would face some problems which could be prohibited with 5nm.
- Pronouns
- He/Him
I think we have to consider the realistic possibilities for this configuration.
A 12SM, 8nm hybrid is not happening. This is the outcome that doesn't make any sense at all due to size and power consumption. Other explanations of the information we have make varying amounts of sense:
A 12SM, 8nm hybrid is not happening. This is the outcome that doesn't make any sense at all due to size and power consumption. Other explanations of the information we have make varying amounts of sense:
- a 12SM 8nm stationary system: this is perhaps the most rational reconciliation of a 2022 Drake on 8nm. The OLED was an upgrade to the handheld experience, now docked-only players can have an enthusiast model. I think that, regardless of its capabilities, such a product has limited sales potential.
- a 12SM 5nm hybrid: expensive, delayed, and limited in production capacity, this is the only potential solution to the 12SM dilemma.
- a different SM count: throw Drake out the window, those numbers aren't happening. Despite the validity of their source, this seems to be the most likely option.
I doubt it would have been delayed for the change to 5nm. these chips are designed in simulation first, they would know from the start if 8nm was unviable for the design and be made for 5nm from the start. Samsung's 5nm looks to be clearing up so I doubt capacity will be an issue
- Pronouns
- He/Him
Nah, 12SM 8nm die is not an issue with size. And power wise shouldn't be an issue in docked mode, they could get away with disabling 4/8 SMs in portable mode just fine too.
8nm 12SM dies will cost more per wafer but a 5nm die will cost more in R&D upfront, and probably have worse availability so may not make back that money as fast/at all. I don't think we can discount either possibility.
D
Deleted member 2551
Guest
What do you consider more likely to happen compared to hardware specs pulled from Nvidia database? Do you have an alternative to suggest that is more reliable than Nvidia itself?
ShaunSwitch
Moblin
This is how I feel too. Both options are possible and both are conclusions we can logically arrive at.
If some of the chatter around samsung struggling to get customers due to TSMC leading the way from an efficiency stand point are true, and with Samsungs own products now moving off 5nm outside of mid range smart phones I could see them offering a decent deal to nvidia for the silicon going into the switch. It's a safe bet and they will have that fab space taken up for years. In the case that this is a successor to the switch then the volume will be high as well.
Remember, Nintendos next device is less of a gamble now than it was, having that confidence that you will be providing huge volumes for years may result in a more favourable deal between parties.
Last edited:
