How accurate is Nash Weedle? They're posting stuff about Nintendo's private showings at TGS including Switch 2 and Metroid Prime 4 but not sure how reliable they are?
It wasn't about showing Metroid Prime 4 just a status update on Retro and where the game is in development.
(Translated from Spanish)
"Some people have already seen Metroid Prime 4: A person close to Retro Studios commented in a small committee that they are planning marketing. This person also said that the technical deployment of the game is overwhelming but that there are also many playable new features."
Ah that makes sense if they people who saw it are gonna work on its marketing. If that's true then the game is probably scheduled for 2024.
They also said Panic Button are working on a game and that Microsoft even attended the TGS Nintendo private showing and were extremely impressed / gobsmacked with what they saw and "Have no doubt they can port whatever they want"
He said allegedly expect leaks to start flowing about the TGS event from other leakers soon.
So I'm going to wait for other leakers including Nate to report as I'm unsure about Nash Weedle. Pinch of salt and all that. But none of these claims sound too crazy they just kinda confirm what the Matrix Demo said and that was it's a capable system that current gen games can be ported over.
We do know that Nintendo had a closed door event so it's easy to speculate and make things up, but if Eurogamer or reliable sources confirm the same things it sounds exciting!
Last edited:
except switch 1 has been $299 for years and well overdue a price cut. It'll likely be discounted and possibly EoL around launch of siwtch 2 (depending whether Nintendo can still make money on a discounted one).
This will be $299 ideally, $349 would be understandable given price rises generally, and $399 if Nintendo try and capitalise on switch success but I don't think would be sustained - so if its $399 it'll be for a year or two before settling closer to swoled territory
Nate said why would Xbox developers see the Switch 2 at TGS and not at Gamescom or just down the street to the HQ.
I'll put Nash Weedle at a "we'll see".
Yeah it's one of the many reasons (plus I want to spend as little time giving twitter traffic) I didn't post the tweets. I think that the information is easily guessable (it's very safe) and that Microsoft could indeed see the Switch 2 at any time, but I'll just file Weedle as a "Maybe" until we have Eurogamer, Digital Foundry or VGC or Nate have their own leaks come out as I will have a presumption that they'll pretty much line up with one another.
Nobody seems to be backing up Weedle, so I'm taking what is said with a pinch of salt until more leaks from TGS come out and are verified.
I think when there's one leak it's "well that's cool if true" but when it's something like this Matrix Demo that seems to have multiple reports from reputable places it becomes more of "Holy shit that happened and that's awesome!"
It'd be nice to see if Nintendo did a demo on actual hardware rather than target specs.
But there is reason why was $299 for years (6.5 years), and that's inflation and rise of costs, $300 today and 6.5 years ago are totally different values.
So you cant expect that next gen Switch that will definitely be much more expensive for production even compared to Switch OLED have selling price thats even lower than Switch OLED price point.
So $399 is most realistic expecting price point having on mind all above and point that Nintendo will again want profit from day one.
Saying that, later we will have Switch 2 Lite with lower price point and most likely some of current Switch models will be in sale for some time.
Last edited:
How do we know it would be cheaper when we're missing so many key details to compare the two nodes? The only thing we know for sure is approximately how much denser 4N is over SEC8N. The details we are missing for both nodes simultaneously are as follows:
- What do yields of SEC8N look like at this point? I keep hearing conjecture that the yields have not improved in the past three years but there is little evidence to support this conclusion. Do we know for sure with absolute certainty that TSMC 4N has higher yields at this die size range than SEC8N does at that die size tier (whichever die size T239 would be on each respective node)? Do we know the approximate yields for both?
- How much is Nvidia truly paying per wafer for SEC8N and TSMC 4N? We know that Samsung gave Nvidia a true sweetheart deal because they needed big headline customers for their foundry business. Where are people getting the data from for how much Nvidia is paying for each wafer?
- Is opportunity cost being factored into this equation? It's not as simple as doing napkin math to come to the conclusion that the per die cost would be cheaper on TSMC 4N, we also have to consider that Nvidia has likely already purchased tens/hundreds of thousands of SEC8N wafers. Is the cost benefit of TSMC 4N high enough to not utilize those existing wafer contract purchases?
- What is the R&D cost to port an architecture to an entirely different node? All of the existing Orin SoCs are on SEC8N as it stands today, and there is an R&D cost factor into porting it to an entirely different manufacturing process, especially one that has not yet seen Ampere or A78 cores from Nvidia. Compare that to the R&D cost of making a cut down version of Orin on the same node which would be significantly cheaper.
Last edited:
One thing that was reported very early was that N4 was getting very good yields from the start. Samsung 8nm has apparently been generally not great in all areas, including yields, from the start.
But really, the main reason to assume it's N4 is simply that the design of the chip makes no sense at all as an 8nm chip. They wouldn't have made it so large if it were, as an 8SM chip would've been cheaper and more powerful on that node.
Last edited:
Oh noes, my gig's up, I totally made stuff up.
Nah - seriously, you can read for yourself. 2.7x denser, for 2.2x the cost. SemiAnalysis is a publication whose sole function is to do analysis on the semiconductor industry.
I'm starting to think you don't even realize nvidia has already been making their other products using TSMC 4N node process. nvidia using TSMC 4N as one of foundry choices is not a guess or rumored. It's an actual, real thing, they're doing TSMC 4N with their newer products. Is T239 also using TSMC 4N? That's the big question.
No idea. Keep in mind it's a custom SoC. The T239 SoC tape out was rumored to be done mid-2022. A Linkedin profile was found by Doctre81 on YT, with resume attached, outlining project done at Siliconus Techcology in India. It is however unknown whether this is T239 or not (It could be for different nvidia chip). This person's time at Siliconus Tech happens to overlap with the other Linked profile who also worked at Siliconus Technology, the one who has "T239" mention in his profile. Both of those engineers worked in the same city as the person who did that (public) github commit updating info re: T239 in September 2022.
Sure, T239 could very well be SEC8N, but the question is, how did they manage to do it with 12 SM GPU? It just doesn't line up. It just doesn't make sense. The new device would be big like a brick. TSMC 4N makes sense, and TSMC 4N also happens to be a choice available at the time of the rumored mid-2022 tape-out.
By the way, upcoming process node is known years ahead. Nintendo and nvidia could very well have been waiting until TSMC 4N was ready. TSMC 4N entered HVM (high volume manufacturing) Q4 of 2021, in time for the rumored mid-2022 tape-out.
Last edited:
so i was just thinking about something that's been obvious for a while no but didn't really realize the implications until now.
so we know for a fact that switch 2 is a 4k capable machine, because it ran botw in 4k at 60 fps. now yes, i'm fully aware it did that WITH dlss, but the point is, regardless, it's a 4k capable machine! this is something that neither xb1 nor ps4 can do. so knowing this, does this give us a better idea of the theoretically upper limit of the switch 2 combined with everything else we know about it so far?
so we know for a fact that switch 2 is a 4k capable machine, because it ran botw in 4k at 60 fps. now yes, i'm fully aware it did that WITH dlss, but the point is, regardless, it's a 4k capable machine! this is something that neither xb1 nor ps4 can do. so knowing this, does this give us a better idea of the theoretically upper limit of the switch 2 combined with everything else we know about it so far?
That just tells you what output port you have. It says nothing about it's technical abilities. A raspberry Pi 4 that costs €35 and has a cheap broadcom GPU can output to 2 monitors at 4k.
Last edited:
I think it will trade blows with the Xbox One X when it comes to 4K resolution but Switch 2 will be better in just about everything else.
Where the One X has raw GPU power and massive bandwidth it lacks in just about everything else.
Heck, if Cyberpunk 2077 ever comes to Switch (I think it will) it will have light Ray Tracing features, better textures, geometry and denser NPC population thanks to the CPU and more advanced GPU
IMO we talking around PS4 power (but with more RAM and better CPU) in docked mode without DLSS,
so games at end (with DLSS) should be similar to how games look/run on Xbox Series S.
Last edited:
true, but why would anyone bother having the capability of outputting at such resolutions unless they felt comfortable they might be able to utilize them?
There's actually are Samsung UFS memory cards that would be sufficiently fast, and you can make slots that take both UFS and microSD cards.
These things came out in 2016 and never blew up, but they also work in almost no devices. A console line the Switch 2, with the sales it would get, would obviously change that.
These would be great. Anything but proprietary memory cards, please and thank you.
It would be between a PS4 and the Pro in docked before DLSS. PS4ish levels would be handheld like the Steam Deck.
From my limited understanding anyway this is what should be expected.
also, i was having fun reading over this old thread all the way back from 2020:
www.resetera.com
and it's pretty shocking how a bunch of people totally nailed what the CPU was likely going to be 3 years before the leak
Based on how tech has progressed since the Switch, how powerful could a Switch Pro in early 2021 actually be?
I've been told by a few people (mostly on forums and Discord servers but still some people) that a Switch 2 that is 5-10x as powerful as the Switch 1 is not viable until after 2023 and that makes me wonder how much of a boost a Switch Pro could actually be. Like, if this a 2x power increase with...
www.resetera.com
and it's pretty shocking how a bunch of people totally nailed what the CPU was likely going to be 3 years before the leak
If the user puts in a microSD it serves as cold storage, if it's UFS it can run games? Would be really cool. I wonder how Nintendo could make it easy for consumers to know the difference between UFS and microSD.There's actually are Samsung UFS memory cards that would be sufficiently fast, and you can make slots that take both UFS and microSD cards.
These things came out in 2016 and never blew up, but they also work in almost no devices. A console line the Switch 2, with the sales it would get, would obviously change that.
Wouldn't need to:
Samsung has developed a slot that will take both UFS and microSD cards
Samsung and several partners have developed a new socket design that will be able to accept both UFS and microSD cards.
www.androidauthority.com
However, maybe they make it UFS-only to reduce confusion? I could see a lot of people buying the much cheaper microSD cards or moving them from their old Switch and being confused why it doesn't work the same.
Most games on consoles already use dynamic resolution, but PS5/XBox have nothing comparable to DLSS. I don't think AMD has something like DLSS in the making, it is a result of Nvidias AI research and AI is by far the most important market for Nvidia.
Anyway I think DLSS does not work with dynamic resolution, so a steady resolution would be a big plus for me visually, I hate dynamic resolution.
Ultra performance DLSS does not work with dynamic resolution (this would be like 360p -> 1080p or 720p -> 4K, as examples), however performance through quality (540p and up -> 1080p or 1080p and up -> 4K) do work with dynamic resolution scaling.
Though in my experience it's in some ways less noticeable, as the output resolution that it's scaling to never changes, generally causes the flaws in DLSS to become slightly more pronounced moreso than the obvious drop in resolution that comes with standard scaling.
That's just an arbitrary restriction that Nvidia impose on PC. There's no reason that needs to transfer over to Switch 2. Dynamic resolution scaling with DLSS will be common on Switch 2 and it won't be artificially restricted to a 4x reconstruction.
GPU will hardly be on PS4 level in handheld mode, but at end that mostly depends on Switch clocks..
I believe someone in the HW thread did the calculations of what the performance would be with the leaked chip using switch clocks and it put it at PS4 levels handheld. Unless i am mistaken.
Well it goes have 6x the cores and so might be 5-6x the Switch at equal clocks. That sounds about PS4 level, maybe. Add in DLSS so you only have to target 720p or 540p...
Would only really make sense if they were talking about devs from Tango Gameworks, right? Microsoft's first party Japanese studio...
Haven't seen anything yet, but DF said they'll be taking a look at it.
They have a video on RE8 for with Metal FX for Mac already, and I don't think A17 pro is too far off M1. Curious to find out what they're upscaling from on iPhone though.
Inside Nvidia's New T239 Processor: The Next-Gen Tegra For Switch 2?
View: https://www.youtube.com/watch?v=czUipNJ_Qqs
Three other things that can be taken away from that video…
- DF vetting Nate's BotW loading report.
- DF mentioning somewhere between 8 & 12GB of RAM. Couple this with Nate saying nothing below 8GB, we're likely looking at 12GB.
- T239 is capable of HDMI 2.1 throughput.
Looks like there's no deep learning accelerator in the Switch 2 SoC.
View: https://youtu.be/5Y5CBMgAFlU?t=6867&si=ed4llo3K55xWYAhc
View: https://youtu.be/5Y5CBMgAFlU?t=6867&si=ed4llo3K55xWYAhc
Is there a cost effective DisplayPort to HDMI chip out on the market that could power a new dock?
Thats, kinda insane? I feel everyone was expecting DLSS to be a big plus for the Switch 2 to punch above its weight. Hope Rich is wrong about that but he did cite "multiple sources" there.
The stumbling block i always thought when people brought up DLSS for a Nintendo console is Nvidia making one or multiple convincing presentations at Nintendo for them to adopt a new technology. I remember the Nintendo that didn't want to hear about what other online stores were doing, so it would be an uphill battle to make Nintendo consider anything at all from an outsider.
Oh man. Now people will be mad at Nintendo for not offering something they never said they would offer.
The only folks that can convince Nintendo to adopt technology is the Xenoblade team and capcom forcing them to update their online infrastructure to beyond Windows 98 levels for Monster Hunter Rise.
For people who haven't been embedded in the T239 stuff: the Deep Learning Accelerator (DLA) is a separate hardware block that's present on Nvidia Orin (T234) that also accelerates matrix operations. It's in addition to the tensor cores that are on the SMs, not in replacement of them.
T239 will still have tensor cores and run DLSS. It just won't have that additional DLA hardware on the silicon.
T239 will still have tensor cores and run DLSS. It just won't have that additional DLA hardware on the silicon.
They did not say no DLSS. They said it would not be "free".
lmao yep. I told y'all years & months ago to expect nothing with this new switch and don't expect DLSS.