If they're thinking SEC8N for T239 because the other Ampere GPUs were SEC8N, that logic is as bad as saying T210 will be on 28nm solely because other Maxwell GPUs were too. The T210 SoC, which has Maxwell-based GPU in it, ended up being fabbed on 20nm process.
What are we defining as arrogance? And what exactly would compete against them; since the major knock on any product would be the exclusive software they don’t have against Nintendo.
Sometimes we forget we are in a bubble here. DF's reasoning behind Switch 2 using a samsung node, and even 8nm (well, more so the former), is plausible, despite to many of us thinking it's absurd.
They could probably fit in 3 TFLOPs GPU and 1.5GHz CPU.
Quick question which would be the second best option if Nintendo neither go for 8NM or 4TSM.Sometimes we forget we are in a bubble here. DF's reasoning behind Switch 2 using a samsung node, and even 8nm (well, more so the former), is plausible, despite to many of us thinking it's absurd.
8nm in 2025 feels so outdated and backwards though. The size and power constraints is a bit suspect for 8nm, though some think it's still possible on a switch form factor. But a 5nm Samsung is starting to feel a tad more viable to me over 5nm/4nm TSMC. Again my reasoning behind this is because
1. It would be so much easier and cost effective for Nvidia to transfer 8nm Orion layout into 5nm on Samsung, and it would cost far less money and time than making it, vs on a different foundry all together.
2. Samsung offers even steeper discounts if Nintendo chooses screen, RAM, internal storage from them making it more affordable for consumers and more profit margins from them.
I don't know the chip yield for 5nm Samsung off the top of my head, but it would be cheaper than tsm for Nvidia and Nintendo, and meet the chip size and minimum power draw at least.
I want it to be TSMC, but if Nintendo wants to save money (which they love), they could have gone to Samsung. They do care more about battery life on handheld and they would want to make it affordable. I hope I'm wrong and they chose 4nm TSMC. I have no real expectations on power anymore. DF's pessimism reminds me of WUST and the pre switch days when we were hoping for tx2 or 16nm node from the get go with 750-1000 TFLOPs from the get go.
Arrogant in the sense that they increase the prices of their hardware/software products, I don't want that to happen.
It would only be arrogance if the market rejects it throughly. Seemingly the market has not done that yet; so they may be justified in increasing the price even if the market as whole already has done this.
Although TSMC's 20 nm* process node and TSMC's 16FF process node both use ArF immersion lithography (DUV lithography), TSMC's 20 nm* process node uses planar transistors, a different type of transistor from what TSMC's 16FF process node uses, which are FinFETs. So I think transitioning from TSMC's 20 nm* process node to TSMC's 16FF process node is no less complicated than transitioning from Samsung's 8N process node to TSMC's N6 process node or newer, or Samsung's 5LPE process node or newer.
I thought PS5 Pro was rumored to use 4nm, not 6
I ate it
And the crazy thing still moved another 20 million after the DS released, though it admittedly also had the benefit of being only $80 after its successor dropped. GBA impresses me.GBA was pushing Switch level numbers until DS kneecapped it.
GBA was an impressive system: only around for a few years before DS came out and yet it imprinted so heavily on my brain. I really like that portable. While DS was maybe necessary in a world where 3D graphics and the PSP were rising and was at the very least extremely successful, I kind of wish we'd seen a reality where GBA got a few more years of dedicated support. I'd take a GBA 2 right now.GBA was pushing Switch level numbers until DS kneecapped it.
Sounds like some brand of Cup noodles.
Nintendo Switch Attach
I don't know how you are labelling "per generation", but I'm pretty sure some Wii U owners also had a 3DS. You're double counting them as "reliable Nintendo purchasers." And I'm sure a decent chunk of those 3DS users were buying special editions/New 3DSes/second units. Those only happen if the base unit manages to hit a certain level of sustained popularity.
Sony is certainly using N6 for the PS5 Pro. Like @darthdiablo mentioned, these consoles do not have the same power efficiency/size constraints as a portable console. In addition to this, N6 is cheaper per transistor than N5. I spoke with Kepler_L2 briefly about the cost of Viola (PS5P SoC) on N6, and he came to the conclusion that it would be about 320mm² on N6, vs the 260mm² die size of Oberon Plus (PS5 Slim SoC, also on N6).
From what I know, Maxwell was originally designed for 20nm and had a rather rushed port job to 28nm due to the nodes flaws (low yield resulting in higher cost per transistor). Rather (in)famously, Nvidia publicly lambasted TSMC for 20nm's issues, a rather unprecedented break from the norms of customer/foundry relations.
In my opinion, the only probable nodes are Samsung 8N, and TSMC 4N and 7N (Nvidias customized N7 family process).
6nm TSMC I think.
Perhaps. But it's also way easier and cost efficient to stick to the same foundry as t234 vs moving into an entirely different foundry. I'm hoping we at least get 5nm Samsung, which is equivalent in tsmc 7nm in performance. Obviously, 4nm TSMC is the best case scenario.
They're not. They are well aware of the discussions in this thread. Rich is the most skeptical of the bunch when it comes to a newer node and especially from TSMC.
I've been seeing most expect $400, a few are cautiously optimistic about $350, but I don't think I've seen anyone confident about $299. At least not in the past year or so.
No. Most of the discussion here seems to have centered at $400, and I've seen some mention $450. $350 or lower have been comparatively rare from my reading.
I personally don't buy this reasoning. I think Nvidia are very capable of Die shrinks from Samsung to Samsung.
We don't think it will be a Samsung panel.
There are replying to a patreon question. But the whole 299 USD was immediately squashed. Neither Olli nor Rich or Alex believe the Switch 2 will be that affordable.
I would like to stick for a moment to the hypothesis that Nintendo could have a good deal from Samsung if it chose to source several parts/services from them (RAM, foundry, storage, screen, cartridge tech, etc.). I think we haven't fully profiled a Switch 2 yet that would be the result of such collaboration.
I mean, I do think a lot of people confuse the different visual representation of the Matrix demo on switch2 and ps5, but that doesn't stop DF from being overly conservative in their speculation when discussing the T239 process, and that speculation is most likely based on their presupposition that the T239 and orin series are the same process
Yeah, their main audience (mostly PC gamers) can be very aggressive if it comes to graphics. Even if Switch 2 aligns with our most optimistic expectations, they‘d call it underpowered after 4-5 years regardless.
When they put out that witcher thought experiment my heart sunk lmao. They could be right which would suck but reminding myself that the leap in technology will be so good for first party anyway made me feel better lol
LMAO
Nope. Because
Imagine if Nintendo did have a working PS5/Series X version of Matrix Awakens ready to show to developers at Gamescom and a demonstration (either physically or through pre-recorded video) of it on the T239 to show side-by-side. That'd be impressive.
