-
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
kimbo99
Spirit Detective
- Pronouns
- He/him
ShaunSwitch
Moblin
Concernt
Optimism is non-negotiable
- Pronouns
- She/Her
To make a long story short, H1 would still be on the cards with 4N.
OP
OP
Depends on when the SoC's taped out, I think.
not really unless you want some super fresh, new node. but 5nm has been in use for years, and Nvidia has ample product on it for Drake to take advantage of. and Lovelace has been sampled for a while now
- Pronouns
- He/Him
The initial design had to have had a target process node in mind, and that's not something that can easily change at any point after the design has begun. So whatever process node they had chosen back in 2019/2020 which gave us the late 22/early 23 window likely hasn't changed, meaning the window itself likely hasn't changed.
Goodtwin
Like Like
That is a good point, and may be good reason to believe it wont be on 4nm, its just to new. Seeing as how they would have needed to nail this down a year or two in advance of release, 5nm or even 6nm is much more likely. 5nm is been widely used in the mobile market for a couple years now, making it a very mature process is probably the best case scenario for Drake.
- Pronouns
- He/Him
4N belongs to this family. If it’s 4N, it’s 5nm.
SiG
Chain Chomp
Does anybody have a chart of "marketing nomeclature" vs actual node size?
Goodtwin
Like Like
It would still be a "newer" revision of the 5nm process though, right? If a chip were designed around the TSMC 5nm from 2020, would that automatically translate to being manufactured on 4N?
- Pronouns
- He/Him
it’s custom only for nvidia, made by nvidia
ShaunSwitch
Moblin
4N is an Nvidia specific node in the 5nm family, I'd wager 4N is more likely. Especially given Nvidia would have known ADA would be on 4N when the chip was being designed.
BlackTangMaster
Piranha Plant
We have to take into account that 2022 should have already been the year for 3nm if we consider Apple as the lead platform for semiconductors (TSMC) before having been delayed. 5/4nm were not supposed to be the flagship node of early 2023.
- Pronouns
- He/Him
5/4nm is fine.
Also, I don’t think people should conclude that about node all of a sudden…
We are missing a lot of context to assume or jump to a conclusion of a node.
Also, I don’t think people should conclude that about node all of a sudden…
We are missing a lot of context to assume or jump to a conclusion of a node.
Last edited:
kvetcha
hoopy frood
- Pronouns
- he/him
Ah, looks like my hopes and dreams for Drake using IBM's 2nm node are dashed. Doomed, indeed.
(But seriously thank you for the responses)
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
kvetcha
hoopy frood
- Pronouns
- he/him
20A sounds WAY more advanced than 2nm!
Z0m3le
Bob-omb
Great info, for the people reading this, remember Switch was tested at 921MHz and ended up being 768MHz, if we get the 1.125GHz GPU, that is 3.456TFLOPs, it will be the same performance tier when docked as XBSS, though slightly below it, it wouldn't be a major difference, and if DLSS 3.0 is available, they could even push 4K 60fps out of it, even if the games feel 30fps, it would look smoother like 60fps (that is basically what DLSS 3.0 does).
Goodtwin
Like Like
With the clock speeds Nintendo settled on with the X1 for the Switch, people should probably be very conservative with their expectations for the next Switch. They are likely shooting for about 4-5 watts for the SOC in portable mode and 10-12 watts in docked. Nintendo will be far more conservative with clock speeds than most of us would like.
Z0m3le
Bob-omb
the GPU for TX1 was 998MHz for the full clock, it was reduced by ~23%, to 768MHz, this was on the 20nm node that leaked energy, a 1.38GHz clock with a 23% reduction is 1.063GHz or 3.27TFLOPs, but testing on Switch was done at 921MHz and had a reduction of less than 17% for 768MHz, if we are looking at the same, 1.125GHz is pretty close to that, as that is a reduction of slightly more than 18%... Basically 1.125GHz would be in line with reductions we saw to TX1 by Nintendo, it's even slightly more conservative than Switch was.
Well have a pretty good idea though, once we have battery, battery life and clock speeds. At least we know which processes it definitely isn't
Codename_Steam
Cappy
- Pronouns
- He/Him
Does Drake have the necessary hardware for DLSS 3.0?
for frame generations, no
Alovon11
Like Like
- Pronouns
- He/Them
I wouldn't say no quite yet, again, the OFA in Orin is a unknown factor.
While it isn't an explicit no, don't expect it nessecarily.
10k
Nintendo Switch RTX
- Pronouns
- He/Him
We’re getting a 5nm Nintendo Switch Advance. Whew lads.
Z0m3le
Bob-omb
It's unknown actually. Drake's OFA engine is more advanced than Ampere, it's unknown how it stacks up against Ada's, but it is for auto AI, and is the more extreme application, we just don't know what that means exactly, however Frame Generation can technically be done slower on Ampere, and when we are talking about 30fps of frame generations, that is likely possible, given that DLSS 3.0 on Ada can handle 100fps+.
- Pronouns
- He/Him
We don’t really know for sure
And I caution people from jumping to a conclusion at the moment lol.
Not trying to be pessimistic, just that we don’t have a lot of information here.
We have Orin, and we have desktop ampere. Theres no way Drakes effiency gains over those makes 8nm twice the performance per watt over those.We don’t really know for sure
And I caution people from jumping to a conclusion at the moment lol.
Not trying to be pessimistic, just that we don’t have a lot of information here.
- Pronouns
- He/Him
We also don't know if that information used real measured power draw or some kind of target or estimation.
It's not enough info to change our confidence much IMO.
Z0m3le
Bob-omb
running Nvidia's estimation tools for Orin with everything set to low and the gpu off, the system power draw is estimated at 8.5w, with the GPU set to 4TPC and 624mhz, it draws 14.5w, that is 6w for 1024 cuda cores on 8nm at 624MHz, Drake is being estimated to draw just 4.2w at 660MHz on 6TPC or 1536 cuda cores... Without a single doubt, Drake can't be 8nm if the estimation is accurate, and with a disparity this great, it's very very unlikely it wasn't caught.We don’t really know for sure
And I caution people from jumping to a conclusion at the moment lol.
Not trying to be pessimistic, just that we don’t have a lot of information here.
I disagree whole heartedly. There is no way you could estimate Ampere with 12SM to draw 4.2w on 8nm at 660MHz, it's double that.
D
Deleted member 887
Guest
I recently went back through some pre-LAPSUS$ chunks of the thread, which was... a f'ing trip. I have a few of long-lost threads I'd like to pick up.
@Thraktor and others, after Orin started being shown off, seemed convinced that the transistor density was low enough that it was clear Nvidia wasn't using high density mobile libraries. This is outside my area of expertise - what is involved here? Does it change the power/thermal relationship at all, and considering that Drake will swap out the A78AE is there an opportunity for the CPUs to use them?
What the hell ever happened with Orin S/Orin ADAS/Nano Next? It seems pretty clear that Orin Nano is filling the Nano Next slot. At one point, the Orin roadmap had Nano Next launching before Orin X, and there was talk of an Orin ADAS chip that might or might not be Nano launching in "early 2023". Many of us at the time assumed that Nano Next was Drake, or the ADAS chip was Drake or they were one and the same. Obviously Nano isn't Drake, but is it possible the ADAS chip isn't Orin Nano, but the ADAS chip? Is there any reason to believe that there was any change up with Orin Nano that lead to it being delayed?
There was some talk of Samsung having a modified 8nm process specifically for automotive applications that might require higher heat tolerances. @Dakhil , do you have any sense of whether or not Orin is manufactured in such a way? And if there might be some advantages, particularly with regard to density, if Drake is not manufactured in that way?
@Thraktor and others, after Orin started being shown off, seemed convinced that the transistor density was low enough that it was clear Nvidia wasn't using high density mobile libraries. This is outside my area of expertise - what is involved here? Does it change the power/thermal relationship at all, and considering that Drake will swap out the A78AE is there an opportunity for the CPUs to use them?
What the hell ever happened with Orin S/Orin ADAS/Nano Next? It seems pretty clear that Orin Nano is filling the Nano Next slot. At one point, the Orin roadmap had Nano Next launching before Orin X, and there was talk of an Orin ADAS chip that might or might not be Nano launching in "early 2023". Many of us at the time assumed that Nano Next was Drake, or the ADAS chip was Drake or they were one and the same. Obviously Nano isn't Drake, but is it possible the ADAS chip isn't Orin Nano, but the ADAS chip? Is there any reason to believe that there was any change up with Orin Nano that lead to it being delayed?
There was some talk of Samsung having a modified 8nm process specifically for automotive applications that might require higher heat tolerances. @Dakhil , do you have any sense of whether or not Orin is manufactured in such a way? And if there might be some advantages, particularly with regard to density, if Drake is not manufactured in that way?
SiG
Chain Chomp
They were also very flexible with those clock speeds as time went on, introducing Boost Loading clocks and even adding more portable profiles with higher than usual power.
And yeah, what @Z0m3le said about Erista: Those downclocks were more like "post-thermal throttle" clocks.
Last edited:
LiC
Member
For me it's 8nm until it's confirmed otherwise. /shrug
On top of Ampere hardware simply not being supported by DLSS 3, this project was being developed alongside DLSS 2.x, and that's what was integrated into the NVN2 source at the time of the leak. So they were developing with 2.x in mind, and there's no mention of frame generation in the API, and that is not going to be something tacked on in the last minute of development (software planning at the scale of Nintendo and Nvidia just doesn't work that way) even if the hardware was supported.
- Pronouns
- He/Him
Without knowing anything about the context of that chart I really don't think it can inform us much. Is the tool using the 4.2W number to estimate the entire GPU power draw? Or maybe just part of it. Does Drake have some major power efficiency improvements that Orin does not? Could it have been an old out of date test done when the GPU had a different configuration with different power requirements?
There's too much we don't know to draw solid conclusions.
if Nvidia isn't too keen on supporting Turing or Ampere with frame generation, I don't think they'll make the effort to support Drake with it. if Nintendo wants FG, they might look towards AMD's solution when that comes out
- Pronouns
- he / him
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
FWIW, a small gallery of examples of PC Witcher 3, low settings, DLSS 360p->720p versus native 720p.
sometimes I wonder if it's really better to have TAA when the input resolution is so low on the switch. this is another one of those times. the filtering on the textures is so harsh
LiC
Member
They could. I just don't think it's in the cards. That's extra development on NVN2, on the SDK and OS integration/distribution side, and on the DLSS side, since -- possible OFA capability or no -- DLSS 3 just doesn't support Ampere.
Hidden content is only available for registered users. Sharing it outside of Famiboards is subject to moderation.
- Pronouns
- they/them
I really wouldn't go that far.
If we're assuming a launch in the first half of 2023, 5nm and related nodes might still be a bit expensive for a product that's going to have to start at, what, $399? With two controllers in the box? I'm really not sure Nintendo is about to be like "okay kids, here's your new $500 Switch." People like to say things like "5nm has been used in mobile for years" while not really considering the price of the (generally high-end) smartphones that those 5nm SoCs are working their way into. Phones like iPhones take up a huge chunk of available wafers and they have the margins and high selling prices to be able to pay the premium for the latest nodes.
The tenth-generation iPad just went to the 5nm A14 and the price went up considerably, to $449, with the previous-generation model remaining available at an unchanged price as a lower-end option. I get that's kind of what we're expecting to happen with the new Switch — be more expensive with a current Switch still available at $299 — but I'm still not sure I can really see it. Something like TSMC N6 just feels a lot more realistic. Nintendo's not going to put all their money into a fancier SoC. You know they're going to include some sort of new gameplay-enabling features, like cameras and mics or whatever, and those features cost money.
If we're expecting something released in the back half of 2023, like in November, 5nm/4N/whatever might be a little more realistic, since Apple will move the iPhones to 3nm and that will free up a ton of capacity on 5nm.
- Pronouns
- He/Him
My issue with this is that we are making ”pick a point that fits my idea” about the provided data, and assumes that this is perfectly aligned with DRAKE. How do we know it’s for Drake? How do we know it’s not for tool profiling? There’s a lot we don’t know here about the SoC and it’s performance characteristics to draw a reasonable conclusion.
To be fair, you wouldn’t get any confirmation from Nintendo about this
You’d have to make an educated guess based on other factors at play, and even then it’s making an assumption.
LiC
Member
Somehow everybody concluded that TX1+ is using 16nm.To be fair, you wouldn’t get any confirmation from Nintendo about this
You’d have to make an educated guess based on other factors at play, and even then it’s making an assumption.
- Pronouns
- He/Him
UHS-II and UHS-III are, as you point out, quite a bit pricier, as are other options other than UFS Card. M.2 storage has a power consumption and size problem, so I leave that out.
The main impediment for external card storage is, as @Pokemaniac suggested earlier, Nintendo putting in the legwork to establish the market for it that does not yet exist, no matter which microSD UHS-I alternative they choose. That is doable, especially for UFS for the reasons previously mentioned, and they absolutely should choose the option that is cheapest to the consumer, but it operates under an assumption.
That assumption being that, if internal storage and Game Card read speeds increase as we expect they may need to, Nintendo will want to maintain the ability for loading games off external storage and find a solution that grants them read speed parity. But they could just as likely revert to the idea of external storage being storage ONLY and require games to be played off internal storage only, bringing us back to the days of the Wii. Such a decision will ultimately come down to how much internal storage capacity they have, so as to prevent game juggling across storage when possible.
Look over there
Bob-omb
If the SSDs are already getting hot now, then it doesn't seem like there's much overhead left in terms of amount of watts that can be going into a drive of that form factor. So, if there's a goal of increasing raw sequential read in the following gen, one option is work off of presumed improvements in perf/watt of later generations, but not go anywhere near full throttle. Use some p-state in the middle. Alternatively, see how eUFS develops and go with that. Samsung did announce UFS 4.0 this year, claiming read speeds of up to 4.2 GB/s. Fast forward to the time of 'The Next Generation', maybe there's UFS 5.0 with a doubling to 8.4 GB/s. And UFS having the design principles it has, that should easily offer multiple GB/s per watt (I'm guessing that UFS 3.1 already should hit at least 2 GB/s per watt under typical operating conditions).
Speaking of heat in the big consoles, how's the RAM? GDDR7 should be doubling GDDR6's speeds while the change in signaling is alleged to reduce power draw by 25%, so it seems like double the bandwidth for 1.5x the power, before factoring in future node-related improvements. Can the console form factor support cooling full speed GDDR7?
And although it hasn't been stated yet, I think that it's likely that density is doubled. GDDR6 comes in 1 or 2 GB 32-bit chips, so we could be looking at 2 or 4 GB 32-bit chips for GDDR7. Now considering the potential slowdown in increase in raw sequential read, ~doubling the amount of RAM may not necessarily be enough for Next Generation experiences? There are options to RAM quantity (DDR and LPDDR should be beating GDDR handily in that area), but they do involve the tradeoff of relatively far less bandwidth.
(I've probably hinted at it a few times by now, but I am interested in a timeline where the set top boxes evolve towards something like, 256-bit LPDDR6/6X or something with fat amounts of RAM and get back under a hundred watts)
Last edited:
