Nintendo delays Switch 2 preorders to “assess” Trump tariff effects
New console’s planned June 5 launch date is still on for now.
arstechnica.com
Nintendo Switch 2
- Thread starter Shifty Geezer
- Start date
arstechnica.com
that seems to be very close to reality. Additionally, with other offerings nowadays like a future rumoured Playstation's handheld, a Windows based handheld by MS, and the future Steam Deck 2, it won't be the same era for Nintendo, there might be better machines out there, and more customisable too.
MIcroSD Express is the media for the games or for storage expansion of the console or both?
Sega_Model_4
Newcomer
This 8nm (yes it is) processor is drawing less than 10w handheld. Steamdeck OLED is 6nm and is drawing 2-3x that. Only in docked can Switch 2 come close...Optimization and Nvidia hardware can onky make up so much.
Last edited:
I get it's a fit. But that processor is so old. How do we know that they don't have shrunk version of this processor for release? Is there documentation or some sort of manual measurement that happened?
I think Switch 2 hardware was ready at least two years ago and they got a deal from Samsung for 8nm. They have delayed it because Switch 1 was selling and games needed time.
Unfortunate timing for them. They should have just sold it 2 years ago.
DuckThor Evil
Legend
You have a few things mixed up there. FP32 very much has to do with gaming performance. 3TF FP16 is actually 1.5TF FP32 normally. The TF figures go up when you lower the precision. For Example PS4 Pro was quoted having 4.2 TF FP32 and 8.4TF FP16 and the 4.2TF was the meaningful figure.Technically it has 3TF, but FP32 doesn't have much effect on gaming performance, so cutting it in half probably gets you a more realistic number. If it was 3TF FP16 it could run this game at 4k 120fps
Perhaps there is some mix up with the details concerning Ampere instruction schedule pipeline, which went from being capable of issuing 16 FP32 + 16 INT32 instructions per clock from the previous architectures to being able to issue 16 FP32 + 16 FP32 or 16 INT32 instructions, in theory doubling the FP32 output when the pipeline is churning out only FP32 instructions, but in practice normal code is never only floating point instructions, so Ampere didn't actually increase real world performance in relation to the theoretical flops increase.
30mins shorter than the OG Switch. I think this is a great result for the device. And it isnt just a better SoC and more on board memory. The display is much better, the internal speed of the memory is 3x better, better Joy Cons, better sound etc. And power consumption will be more like 1/6 of a PS4 slim in handheld mode.
That is only true in comparision to Turing. But everything prior cant schedule FP32 + INT32 instructions at the same time. Switch 2 compute performance has at least the same efficiency as Maxwell. And Ampere is a much better and robust architecture, too. So the real world improvements will be nearly as high as the theoretical compute increase.
Oh alright i mixed them up :]
Within reason, I'm not sure the extent to which using a very advanced node matters for Nintendo in terms of commercial success? Particularly when balanced against cost and other trades within the system (panel, RAM).
In relative terms, the N64 (1995-1996 production) was fabricated on 350nm; Gamecube (2000-2001 production) was, I think, on 180nm. Both were close to the leading nodes for their time insofar as I understand it. I don't think I had a 250nm CPU until Celeron 300A in late 1998.
And both were, compared to preceding Nintendo machines, generally regarded as less successful (N64 is probably tied for my favourite system, but the market held a different view in aggregate).
In relative terms, the N64 (1995-1996 production) was fabricated on 350nm; Gamecube (2000-2001 production) was, I think, on 180nm. Both were close to the leading nodes for their time insofar as I understand it. I don't think I had a 250nm CPU until Celeron 300A in late 1998.
And both were, compared to preceding Nintendo machines, generally regarded as less successful (N64 is probably tied for my favourite system, but the market held a different view in aggregate).
Last edited:
A more advanced node on a portable would give them both better performance (and they are trying to get third parties on board, so every MHz is important) and better battery, which means people playing your platform for longer, using your store more and using your services. Is saving like 10 dollars really worth it?
An Italian journalist that I follow was at the Switch 2 Paris event and asked the technical director of Hogwarts legacy about the Switch 2 version, and he said that it was using DLSS 2. DF said that there was no trace of DLSS in the footage that got released, so what's going on?
Probably. It's just strange that no game in that event presumably didn't have any dlss at all.
tongue_of_colicab
Veteran
If you plan on selling a 100 million units that's a billion dollars more you're paying. Is that couple of extra fps or 15~30 minutes or whatever battery life really worth it?
They could have put in a bigger battery for probably a lot less. But I guess their Switch usage data shows them this is sufficient. Somebody who plays a lot in portable mode probably hooks it up to a powerbank anyway. It's not like a worse case 2 hours is completely unacceptable while 2.5 hours would be perfectly fine.
By using 8nm for the launch system they make more money now AND get people to "double dip" on hardware in the future.
Like last time they'll use a more advanced node for a lite model and use it the main Switch model as well. The better battery life will get some people to upgrade.
Like last time they'll use a more advanced node for a lite model and use it the main Switch model as well. The better battery life will get some people to upgrade.
