Gerald
Bob-omb
- Pronouns
- He/Him
Its actually very possible, and it will be 4N, its 5nm but betterIs 5nm a possibility or is that totally impossible?
Its actually very possible, and it will be 4N, its 5nm but betterIs 5nm a possibility or is that totally impossible?
Is 5nm a possibility or is that totally impossible?
Its mostly just a marketing distinction.Its actually very possible, and it will be 4N, its 5nm but better
Ever since the Nvidia leaks revealed the size of the chip, people here have questioned if this makes sense on 8nm.So 4N/5nm is on the table?
While perhaps circumstantial evidence, it is curious to me that T239 appears to have gone through validation, production testing and preparation, taping out and finalisation, all on the same timeline as Ada Lovelace cards, all on 4N.Ever since the Nvidia leaks revealed the size of the chip, people here have questioned if this makes sense on 8nm.
Edit: there's no solid evidence for either node. What speaks for 8nm is the fact that it's the node of almost every other ampere device, including Orin. Insiders have also heard its 8nm, however that info is old now and proven not to be that reliable,
What speaks for 5nm is the sheer size of the soc, and what we know of amperes power characteristics on 8nm.
While perhaps circumstantial evidence, it is curious to me that T239 appears to have gone through validation, production testing and preparation, taping out and finalisation, all on the same timeline as Ada Lovelace cards, all on 4N.
This sort of implies to me, though I could be dead wrong, that even if it isn't Ada Lovelace architecturally, it was treated like part of the Ada family for development purposes and seemingly production. (Fairly accurate Nvidia leaker Kopite7Kimi in fact said it was Ada, but we know now it isn't wholly)
It also has a CUDA version above Ampere, but below Lovelace.
This could all be coincidental but I feel like the evidence is slightly stronger for 4N than 8nm. Personally because of the above, I believe 4N is far, far more likely.
"TSMC introduced a new member of its 5 nm process family: N4. Compared to N5, the node offers 11% higher performance (6% higher vs N4), 22% higher power efficiency, 6% higher transistor density and lower mask count." TSMC 4N its just 5nm+Every other Nvidia product launching in 2023 is 5nm.
Its mostly just a marketing distinction.
4N and N4 are not the same thing, confusingly. N4 is an iterative improvement over their 5nm mode. 4N is a custom version of 5nm just for Nvidia"TSMC introduced a new member of its 5 nm process family: N4. Compared to N5, the node offers 11% higher performance (6% higher vs N4), 22% higher power efficiency, 6% higher transistor density and lower mask count." TSMC 4N its just 5nm+
Source https://www.techgoing.com/nvidia-cl...sed-by-the-rtx-40-gpu-is-a-5nm-process/?amp=14N and N4 are not the same thing, confusingly. N4 is an iterative improvement over their 5nm mode. 4N is a custom version of 5nm just for Nvidia
Ever since the Nvidia leaks revealed the size of the chip, people here have questioned if this makes sense on 8nm.
Edit: there's no solid evidence for either node. What speaks for 8nm is the fact that it's the node of almost every other ampere device, including Orin. Insiders have also heard its 8nm, however that info is old now and proven not to be that reliable,
What speaks for 5nm is the sheer size of the soc, and what we know of amperes power characteristics on 8nm.
We could make theories all day and some of them may be right, but there is no way to confirm them.I kind of wonder if the initial plan was to launch a cut down/heavily underclocked version of the chip at 8nm and then eventually release a "Mariko/Lite" equivalent at 5nm a year+ later ... and as development went on it became clear Nintendo would be launching 1+ year past the initial 2023-ish target that Nvidia (or Nintendo) just said ... "hey how about we just skip the 8nm altogether and start at 5nm?".
While I'm open to the idea that Nintendo/Nvidia shifted nodes at some point in the process, I don't think the plan was ever "launch at 8nm, move to 5nm quickly". There is an unusually high amount of work in that particular jump, both on the chip side and on the software side.I kind of wonder if the initial plan was to launch a cut down/heavily underclocked version of the chip at 8nm and then eventually release a "Mariko/Lite" equivalent at 5nm a year+ later ... and as development went on it became clear Nintendo would be launching 1+ year past the initial 2023-ish target that Nvidia (or Nintendo) just said ... "hey how about we just skip the 8nm altogether and start at 5nm?".
While I'm open to the idea that Nintendo/Nvidia shifted nodes at some point in the process, I don't think the plan was ever "launch at 8nm, move to 5nm quickly". There is an unusually high amount of work in that particular jump, both on the chip side and on the software side.
The situation with the TX1 isn't too comparable because it is off-the-shelf. The TX1 was already going into production. Drake is made for Nintendo and node changes will be costs that Nintendo also incursTrue, though I do think Nintendo would want a full hardware mapping out before even commiting to the first design ... ie: like we plan to launch 8nm Switch 2 in early 2023, but we also know we are going to move to 5nm in 2025 for a revision and Lite model, just like the first Switch (as their plan circa 2020 lets say).
So Nvidia would probably have to have that roadmap figured out before hand, because if you're Nintendo you want to know what hardware options you have not just for the day 1 launch unit, I want to know if I'm Nintendo's president when we can launch revisions in the first half of the product cycle.
Like wasn't Mariko in the Switch firmware that hackers found quite early on? So Nintendo/Nvidia likely had the plan of going to a lower node and releasing Mariko/Lite models a bit after launch all along.
So maybe like the whole project (Switch successor) already got delayed by like a year++ due to COVID, Switch having better sales than expecting, manufacturing crunch, etc. etc. and then it became more feasible to just skip 8nm and start at the 5nm point.
Sort of like imagine for whatever reason if Nintendo launched Switch in 2019 due to a myraid of reasons, well if you're launching in 2019, you might as well probably go with the 16nm Mariko design and forget the previous version.
That might also explain the whole kerfufle about dev kits seemingly going out and then that hardware going into limbo.
We'll probably never know I guess, but if the Switch 2 chip is indeed 5nm day one, that might be a tip off.
Mariko probably wouldn't have existed without Nintendo though.The situation with the TX1 isn't too comparable because it is off-the-shelf. The TX1 was already going into production. Drake is made for Nintendo and node changes will be costs that Nintendo also incurs
For sure. Honestly Nintendo Switch was the difference between Tegra X1 and gaming-first Tegra devices as a whole failing and being the largest single SOC by market share in gaming.Mariko probably wouldn't have existed without Nintendo though.
when you got a partner like Nintendo stuff like the security flaws can't exist. that probably gave nintendo a lot of leverage for Mariko and the costsMariko probably wouldn't have existed without Nintendo though.
Their Q4 for the current FY ends March 31 this year, so what else are you expecting them to drop in the next roughly 40ish days? Iβm strictly referencing the current FY ending March 31, not the calendar Q4 later this year or their Q3 of the upcoming FY. Whereβs the irony of overthinking there?It's a little ironic that you are accusing them of overthinking after typing all of this out. We don't know what is coming Q4 so to say they have nothing to hand their hat on is a bit presumptuous. Also, the President isn't really advocating for potential price increases unless I've missed a quote. He said to investors they aren't looking to change the prices of their hardware or software. And the Zelda price increase so far has been isolated to just that game.
A bit of an overreaction, imo, to say they need a leadership overhaul.
don't know if they can do the same thing twice, for the same reason mid-gen refreshes aren't expected for Xbox and Playstation this gen. die shrinking might not be as cost effective againI do think for the new chip whether it's 8nm, 5nm, or whatever to start ... Nintendo will want not only that chip, Nvidia will have to provide a roadmap for a die shrink because Nintendo will want a smaller model and a model with better battery life anyway.
It worked well for the current Switch, no reason to not want the same for the successor. For consumers we just learn this stuff on the fly, but if you're the president of Nintendo like you don't only want just the (lets say) 2024 launch product, I'd want to know "OK, well when can we release revisions?" as well so you can plan your product cycle and software pipeline accordingly. You need to know this stuff years in advance, you need to know before you even launch the initial model I think.
So yeah maybe the original roadmap of 8nm followed by a 5nm version got changed to 5nm followed a 3nm version later or something like that.
Could just be a byproduct of waiting the extra 12-18 months perhaps past what the original time line was supposed to be. Like I could see Nvidia saying "well you know if you're going to wait this long, you might as well just go with 5nm and then you can pay us for another revision at 3nm later, 5nm and 3nm will give you better performance and battery life" and so on.
Ps5 already had a die shrink.don't know if they can do the same thing twice, for the same reason mid-gen refreshes aren't expected for Xbox and Playstation this gen. die shrinking might not be as cost effective again
6nm is process compatible with 7nm products. it's why TSMC is pushing their customers to move to it. also doesn't offer enhancements over the base systemPs5 already had a die shrink.
First PS5 die shrink spotted as Sony quietly shifts to efficient 6nm chip | Club386
Have the building blocks for Sony's 2023 PS5 redesign already been revealed?www.club386.com
Isnt the reason pro system doesn't make as much sense that cost per transistor doesn't go down like it used to?6nm is process compatible with 7nm products. it's why TSMC is pushing their customers to move to it. also doesn't offer enhancements over the base system
Ok fair, but itβs nintendos responsibility to keep those expectations in check with their guidance. They canβt get their story straight and adequately do that lately. Theyβve had to revise hardware forecast the last two quarters (or possibly longer? I have to check). I expect Q4 ending March 31 2023 will be more of the same.Investors are advocating for a price increase. They see inflation eating into margins, want those margins up and have seen Sony and Microsoft start leading the way, while the highest priced unit is the best swlling
Meanwhile, the Switch is selling better than just about any console in this stage of it's life, is still short of it's 6th birthday, which isn't uncommonly long as generations go, and has released an entry priced unit and an enthusiast unit.
Yes, Q3 under performed relative to expectations, but it was also investors who pushed those expectations up, who believed that there was unslaked demand limited by the component shortage and European shipping problems. That this turned out to not be entirely true is a problem lots of other manufacturers are making.
In terms of software sales, saying that Nintendo "only" has a big 3D Zelda is both underestimating Zelda and not factually accurate. I don't know what Tears will do, but while Sequels tend to do less well on the same console, the install base is much larger than it was in 2017. If it manages to keep pace with the previous game, it will outsell every Pokemon game released in 2021 combined.
I expect Metroid Prime, Kirby, Fire Emblem, and Pikmin to all do numbers roughly in line with their predecessors, down a smidge for the usual reasons. In terms of units sold, that would make 2023 a very good year for Nintendo, and an absolutely bonkers last year for a console. Yes, total software sales will be down, but only because so many Switch's have been sold, they've run out of people to sell Mario Kart to.
Which is exactly what Furukawa was referring to when he said that no hardware strategy would be sufficient to drive up sales. Dropping a new Zelda potentially adds a new evergreen, which has a high attach rate.
Will investors be upset if leadership blows the transition? Yeah, 100%. Do I have total confidence that Nintendo is going to stick the landing? Absolutely not. Is Nintendo currently failing to deliver value to investors? Nope! Is Nintendo behaving in a way that is unusual for them or for the industry? Absolutely not.
Do I think Nintendo is more responsive to investors than ever before, and in a way that is potentially harmful to long term prospects? Yes! But do I think that Saudi Arabia is going to push out the team overseeing the most successful moment the company has had in the 21st century just because a minority of fans are getting their More Powerful Toy Box six months later than they want it? No.
Will they or wonβt they provide a hardware forecast at their Q4 briefing for the following FY Q1? That is what I was referencing and my bad for not being clearer.This is quite literally impossible considering there is no official hardware forecast past next month.
When do you suppose that timeframe happened? Perhaps it was from hearing the bad news about Samsung's 8nm yields despite its "maturity"? I'd like to venture a guess that maybe it took place around 2021 but I do get my dates mixed up...Drake may have switch from 8nm to 5nm at some point in the process, but I highly doubt Nintendo planned on an 8nm launch and to rapidly move to 5nm.
Aren't the AI tools used for architecture changes as opposed to die shrinks ?Nvidia also has better AI tools now for die shrinks now, that they didn't have last time.
Tinfoil hat time.When do you suppose that timeframe happened? Perhaps it was from hearing the bad news about Samsung's 8nm yields despite its "maturity"? I'd like to venture a guess that maybe it took place around 2021 but I do get my dates mixed up...
Interesting narrative. Part of me likes to think Capcom could've chimed in and said something like "16GB of RAM would be nice".Tinfoil hat time.
Late 2020 current gen consoles launches. The series S is of a particular interest to Nintendo, because it's the least common denominator for next gen development. Nintendo has meetings with epic games, with third parties about what it will take for them to support their next gen featuresets to bring their games over to the next Nintendo hardware.
It turns out their 6sm 8nm ampere hardware in development is just a bit shy of that threshold. Meanwhile the crypto mining market crashes, so Nvidia suddenly has a ton more reserved 5 nm capacity than they expect to sell, and the original Switch is still selling like hot cakes.
These two factors combined is the catalyst for the T239 project to undergo a major redesign and become the Drake we know today.
Do I think that's what happened? More than likely not, but it's fun to theorize.
I think someone here shared Dread has only done over a million copies to date? Prime may be different due to nostalgia.
the AI is for layout optimizations. it would help with die shrinks by fitting the shrunken hardware in the most efficient packaging possibleAren't the AI tools used for architecture changes as opposed to die shrinks ?
Shouldn't an eventual die shrink still be Ampere, or is there something I'm missing ?
They are used of Die shrinks too.Aren't the AI tools used for architecture changes as opposed to die shrinks ?
Shouldn't an eventual die shrink still be Ampere, or is there something I'm missing ?
Well, the reason they specifically highlighted ramInteresting narrative. Part of me likes to think Capcom could've chimed in and said something like "16GB of RAM would be nice".
Nvidia's research paper on NVCell seems to be talking about optimising the design for standard cell design rules for an advanced process node.Aren't the AI tools used for architecture changes as opposed to die shrinks ?
Shouldn't an eventual die shrink still be Ampere, or is there something I'm missing ?
In what way is Nintendo's story not straight? Nintendo has been super consistent in their investor messaging. Yes, they were incorrect about demand in the post-component-shortage era, but they're not the only company who got that wrong, and it's not like investors were shouting at them about being wrong the whole time.Ok fair, but itβs nintendos responsibility to keep those expectations in check with their guidance. They canβt get their story straight and adequately do that lately. Theyβve had to revise hardware forecast the last two quarters (or possibly longer? I have to check). I expect Q4 ending March 31 2023 will be more of the same.
Sorry, that was unclear of me. I was referring to how Nintendo's 7th year is going to go and how it is going to look to investors. Even if nothing but announced games come out and perform in line with historical trends, Nintendo will have an unusually good 7th year in terms of software sales, and again, and any underperformance relative to expectations is industry wide and not localized to Nintendo.Also how is TOTK a factor for their current FY Q4? It releases in Q1 of the next FY. Same as Pikmin. Maybe vouchers are a factor? Engage, Prime remake and Kirby may slightly help software sales, but I doubt it will do much to drastically move the needle by March 31. Also take the brick and mortar physical pre-orders with a grain of salt. Physical copies may have had very limited print and I expect it to normalize after the 22nd.
Dread did 3 million to date making it the best selling Metroid game ever. The Prime games have generally outperformed the 3D entries, and though Remastered isn't receiving the marketing push Dread did, considering how other remakes and ports have done on the system and the attention it's getting, a similar number of sales seems achievable.I think someone here shared Dread has only done over a million copies to date? Prime may be different due to nostalgia.
there isFrom the documents about the MS-Activision thing
"Nintendo Switch Online has been excluded from our shares as Nintendo's cloud gaming service is very limited. Nintendo's cloud gaming service is only available on the Nintendo Switch device and [REDACTED]. Nintendo Switch Online gives gamers access to online play and cloud saving amongst other features. We therefore see Nintendo Switch Online as predominately an online multiplayer service rather than a cloud gaming service."
There is a platform censored by nintendo in the documentation, Which is mentioned that the NSO subscription service is available (you know, being able to play online, catalog of retro games, cloud storage, etc.)
Currently, for obvious reasons, the NSO service is only on Switch. It is quite likely that the platform being censored by nintendo is Switch2/Drake/YourFavouriteName.
Unless they plan to put it on mobileFrom the documents about the MS-Activision thing
"Nintendo Switch Online has been excluded from our shares as Nintendo's cloud gaming service is very limited. Nintendo's cloud gaming service is only available on the Nintendo Switch device and [REDACTED]. Nintendo Switch Online gives gamers access to online play and cloud saving amongst other features. We therefore see Nintendo Switch Online as predominately an online multiplayer service rather than a cloud gaming service."
There is a platform censored by nintendo in the documentation, Which is mentioned that the NSO subscription service is available (you know, being able to play online, catalog of retro games, cloud storage, etc.)
Currently, for obvious reasons, the NSO service is only on Switch. It is quite likely that the platform being censored by nintendo is Switch2/Drake/YourFavouriteName.
It makes no sense to put on mobiles a service that consists of saving switch games data in the cloud and charging for playing online.Unless they plan to put it on mobile
Or my theory that Drake Switch hasn't leaked because Nintendo has put 75% of the world's population under NDA is true...surely that must've either
A) been some sort of qualification that was deemed inappropriate (i.e. "and it sucks fucking ass lmao")
B) erroneously said it was on mobile devices which was deemed misleading
It makes no sense to put on mobiles a service that consists of saving switch games data in the cloud and charging for playing online.
Dont forger MS has Mojang, Minecraft and Mojang games are a major partner for Nintendo and they have a pretty good relationship.Or my theory that Drake Switch hasn't leaked because Nintendo has put 75% of the world's population under NDA is true...
They talk specifically about the selection of games, cloud saves and charging for online play.Isn't it on mobile already with the NSO app? The NSO app is what even enables the multiplayer portion with voice chat and has interaction with certain games. The document says "Nintendo Switch Online gives gamers access to online play" so saying "Nintendo Switch and mobile devices" does not seem inaccurate to me. Unless they are being strict with wording and NSO being 'on mobile' is not true when it comes to the retro gaming portion.
this is exactly what happened imoIsn't it on mobile already with the NSO app? The NSO app is what even enables the multiplayer portion with voice chat and has interaction with certain games. The document says "Nintendo Switch Online gives gamers access to online play" so saying "Nintendo Switch and mobile devices" does not seem inaccurate to me. Unless they are being strict with wording and NSO being 'on mobile' is not true when it comes to the retro gaming portion.
Nobody would bother to ask that they censor a part for the reason you say.this is exactly what happened imo
document said mobile, then presumably someone said "that makes it sound like nintendo has xcloud lmao"