Nintendo Switch Dev Kit Stats Leaked? Cortex A57, 4GB RAM, 32GB Storage, Multi-Touch.

Status
Not open for further replies.
Wii U had 2GB RAM, but only 1GB available for games.

Right, but her source implied it would be the entire system RAM. No mention of RAM allocated to games.
http://n4g.com/news/1070389/rumor-wii-u-ram-talk-mass-production-and-bundles/

I think Emily is talking about the devkit. How would anyone know what architecture the final retail unit uses outside of top Nintendo brass or someone from the manufacturing plant (if Switch production has even begun yet).

I still think Switch is 16nm Pascal if only for better battery life and not increased specs.

She specifically said in a tweet in October that the retail versions will have 4GB of RAM.

https://mynintendonews.com/2016/10/...ntendo-switch-has-4gb-of-ram-in-retail-units/
 
I think Emily is talking about the devkit. How would anyone know what architecture the final retail unit uses outside of top Nintendo brass or someone from the manufacturing plant (if Switch production has even begun yet).

I still think Switch is 16nm Pascal if only for better battery life and not increased specs.

Denial is the first stage of grief.
 
Would Nintendo really go with a 20/28nm process in the handheld? From what I have read it really doesn't make sense unless they had manufacturing issues. I was under the assumption that Pascal was basically the 16nm process plus some minor changes. This has really become very confusing. This would explain the internal fan though wouldn't it?
 
Would Nintendo really go with a 20/28nm process in the handheld? From what I have read it really doesn't make sense unless they had manufacturing issues. I was under the assumption that Pascal was basically the 16nm process plus some minor changes. This has really become very confusing. This would explain the internal fan though wouldn't it?

They would have to convince Nvidia and TSMC, the manufacturer of those chips, to go back on an obsolete process "just cause"...

Don't forget guys, they choose Maxwell because it has less heat in the small handheld form factor than Pascal’s state-of-the-art tech.

ヽ༼ຈل͜ຈ༽ノ
 
They would have to convince Nvidia and TSMC, the manufacturer of those chips, to go back on an obsolete process "just cause"...

Don't forget guys, they choose Maxwell because it has less heat in the small handheld form factor than Pascal’s state-of-the-art tech.

ヽ༼ຈل͜ຈ༽ノ

I don't wanna be one of those people that disregards stuff that is counter to my opinion. But it really strikes me as strange. Like isn't 20nm obsolete in all products lines but the TX1? If it's based off Maxwell I could believe that given when they probably started working on the chipset. But a 20/28nm process just seems so weird. Why wouldn't you go 16nm? Wouldn't Nvidia basically insist? The only thing I can think of is maybe the early yields weren't good?

That article says ~1TFlop. How is that possible without a fan going into overdrive? I had assumed we were looking at something around 500GFlops.
 
As someone who isn't the most knowledgeable about hardware, is it a huge deal if the Switch uses Maxwell instead of Pascal?

Without knowing manufacturing process, it's all meaningless. Difference between 16nm Maxwell vs 16nm Pascal, would be nearly identical outside of a few features (that the gamer will never know the implications).
 
Without knowing manufacturing process, it's all meaningless. Difference between 16nm Maxwell vs 16nm Pascal, would be nearly identical outside of a few features (that the gamer will never know the implications).
Oh okay, so it's not really worth getting worried until we learn more about the system then. I kind of figured as much. It's just that now we have conflicting rumors so I'm kind of just holding out till the system is broken down later on
 
Without knowing manufacturing process, it's all meaningless. Difference between 16nm Maxwell vs 16nm Pascal, would be nearly identical outside of a few features (that the gamer will never know the implications).

More than that, it's a custom chip. I'm leaning on the low end of everything, but there are a lot variables we don't know. I guess we'll find out more next month.
 
Considering that Emily Rogers on a roll, I'm gonna expect our worst case scenario to be the most likely, including the 4GB for retail units confirmation in october. I think we all should after what happened with Wii u(those that were there during the WUST days.

I don't wanna be one of those people that disregards stuff that is counter to my opinion. But it really strikes me as strange. Like isn't 20nm obsolete in all products lines but the TX1? If it's based off Maxwell I could believe that given when they probably started working on the chipset. But a 20/28nm process just seems so weird. Why wouldn't you go 16nm? Wouldn't Nvidia basically insist? The only thing I can think of is maybe the early yields weren't good?

That article says ~1TFlop. How is that possible without a fan going into overdrive? I had assumed we were looking at something around 500GFlops.

According to Laura Kate Dale, at least 2 fans max when docked.

What kind of perf per watt or chipset wattage are you guys expecting? I've been staying out of the spec threads...
Less than 10 watts?

If Nathan is wrong about Pascal, then there's no reason to believe in 5-8 hour battery life with maxwell imo.
 
Why does it seem like there are tablets that are more powerful than this?

Because they're bigger / more expensive / using custom chips that have been iterated on in house for a few generations already / are far more delicate / have larger batteries?

The iPad Pro's a beast, not sure why you would be surprised over that...

Switch will still most likely have better graphics since it's not rendering at that insane resolution, and run on something even more lightweight than Apple's Metal.
 
So with Emily Rogers seemingly confirming the specs in the OP ("90% accurate" - take that however you will), I'm still pretty happy about the Switch hardware. It'll have good third party support and is close enough to home consoles that it won't be shunned like the Vita was.
 
So with Emily Rogers seemingly confirming the specs in the OP ("90% accurate" - take that however you will), I'm still pretty happy about the Switch hardware. It'll have good third party support and is close enough to home consoles that it won't be shunned like the Vita was.

We knew the dev kits were using a Tegra X1 from the word go. Emily confirming it doesn't really add much to the discussion. The question became how the Tegra X1 was customized, and if the X1 was a placeholder for a more customized chip in the final hardware. Whether Maxwell or Pascal was used became a hot button question, because it would allow higher clock ranges and better power efficiency in this form factor.

Maxwell can still perform very well, but if Nintendo is stuck on 20nm and didn't go FinFet, it will run hotter and eat batteries faster than if they had been able to use a more modern 16nm design. Hopefully they were able to move forward with simply die shrinking their custom Maxwell design, if that's what they went with as a result of timing.

We already know the variant used in the final hardware isn't the same as the Tegra X1, because it's sending a video Signal over a USB C connection. Tegra X1 didn't support USB 3.1, so clearly some customization has happened. How much has been done to the core of the SOC remains to be seen.
 
Interesting. Nice observation!

In regards to the ventura beat article saying they are expecting 1 TFLOPS, wouldnt that be impossible if its X1 Maxwell, since its fp16 at 1024 GFLOPS(512 at fp32) only, but it can't be performing at that all the time.
 
Interesting. Nice observation!

In regards to the ventura beat article saying they are expecting 1 TFLOPS, wouldnt that be impossible if its X1 Maxwell, since its fp16 at 1024 GFLOPS(512 at fp32) only, but it can't be performing at that all the time.

It depends on what it actually is. We say X1 because that is the default Maxwell chip so its easy to at least get people to semi understand. Nintendo Tegra could have additional SM for instance. Nintendo Tegra could have almost no similarity to an X1 other than the fact they would be both Maxwell based.
 
1TF at fp32 would be impossible even for Pascal.

Nintendo can't get a higher efficiency than Nvidia with a modern Pascal chip for automotive applications. There is a limit customization can go.
 
1TF at fp32 would be impossible even for Pascal.

Nintendo can't get a higher efficiency than Nvidia with a modern Pascal chip for automotive applications. There is a limit customization can go.

Parker is nothing more than a new manufacturing node X1 with a 128bit memory bus. They did not do anything more than that because Parker is a stop gap for Volta Based Xavier where they will have significant changes.
 
Just...

Are we now moving to Pascal wasn't important anyway rhetoric?



What rhetoric ? Pascal or Maxwell isn't important in term of performances. This wont make Switch a different platform. It'll either mean a slower device or one with less battery life.



1TF at fp32 would be impossible even for Pascal.

Nintendo can't get a higher efficiency than Nvidia with a modern Pascal chip for automotive applications. There is a limit customization can go.



Having 4SMM instead of 2 is a customisation. I'm not saying it's that kind of customization though, I'm just saying core configuration is a simple one.
 
1TF at fp32 would be impossible even for Pascal.

Nintendo can't get a higher efficiency than Nvidia with a modern Pascal chip for automotive applications. There is a limit customization can go.

That's not true. Being a custom design it could just get more CUDA cores. Plus a higher clock.

Not that I believe it is 1TF at FP32, but your statement is false.
 
So the thread just made a step back to Maxwell just to jump into the 1TF territory.

Some people. That's sure the circlejerking nonsense of a Nintendo hardware thread.
 
I think using a 20nm node would be bad for future proofing the Switch and I expect a revision to happen quite early in Switch's life if that's the case.
 
I was wondering when the bubble would burst and someone would find some imaginary theoretical hurdle which Switch wouldn't be able to jump for them. Seems it's 'Pascal'

Either way you're still likely getting a ~1TF handheld. Even if it's 75% of that it's still pretty awesome.

And as far as I've seen, the only people talking about PS4 level performance were the people coming in here saying "lol ppl think it's going to be PS4 level, SMH Nintendo fans" when no-one was actually saying that. (I've still not actually seen anyone expecting more than 1TF max)
 
If Nintendo is forced to go with Maxwell because Nvidia isn't ready to provide Pascal chips then they aren't going to get a Maxwell with all the keyfeatures of the Pascal series either.

There isn't that much too talk. And even if they would get. Getting Maxwell to provide "more than 1 teraflop" in that formfactor would be quite the task. But Nintendo is gonna beat Nvidia in their own business.
 
Can X1 be clocked high enough to do 1TF at fp32 or is that likely impossible? So either way we are looking at a docked 1Ghz X1 with 512Gflops and portable mode 256Gflops?
 
Status
Not open for further replies.
Top Bottom