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

Status
Not open for further replies.
I disagree. All those 3DS handheld gamers who want an upgrade will look at the Switch and think, "do I want to pay X for my next handheld". I don't think a lot of them will give two hoots about whether it's a console or not. Wii U 15 million, 3DS 60 million, we know what Nintendo's customers like from them.
Didn't stop quite a decent amount of people from spending $200 on anemic hardware with the New 3DS XL or even $180 back when the 3DS got its first price drop. For most people, so as long as the system is appealing, has games they're interested in, and doesn't cost a lot, they'll go for it. The 3DS, even after the price drop, was still costly and the Wii at $250 was pretty expensive, considering it was an overclocked GameCube and only $50 cheaper than the Xbox 360
 
Maxwell has color compression techniques and 25GB/s while Wii U does not, has like 12GB/s and needs the EDRAM to even function correctly.

That 25 GB/s optimized color compression is more like 30 or 40GBs(talking about of my ass lol, but basically, color compression and bandwidth saving on maxwell compared to Wii U's ancient architecture is no joke)

BandwidthSavings.png

Assuming 25 GB/s you could estimate that to a "real" bandwidth of 33.3 GB/s. That's less than half of Xbone memory bandwidth even before bringing SRAM into calculation.

I still hope they use the 128-bit memory bus from Parker.
 
I disagree. All those 3DS handheld gamers who want an upgrade will look at the Switch and think, "do I want to pay X for my next handheld". I don't think a lot of them will give two hoots about whether it's a console or not. Wii U 15 million, 3DS 60 million, we know what Nintendo's customers like from them.
You forgot the 3DS had the similar rocky start. Many people did not think about paying more for a new handheld...until a price drop and some killer apps (Kid Icarus: Uprising) started to appear. Eventually the handheld flourished.

In the end it will be title selection.
At the rumoured clocks the CPU is no longer such a huge advantage over Wii U's. The memory bandwidth is the trickiest question here, because on paper it doesn't look good, unless they use the memory controller from Parker.
Not really. Wii U was severely gimped with its pre-GCN architecture to begin with.
 
Assuming 25 GB/s you could estimate that to a "real" bandwidth of 33.3 GB/s. That's less than half of Xbone memory bandwidth even before bringing SRAM into calculation.

I still hope they use the 128-bit memory bus from Parker.

Considering the target hardware's GPU is less than 1/3rd of XB1's GPU even in docked mode, doesn't that mean the bandwidth fits perfectly for the Switch? :)
 
Considering the target hardware's GPU is less than 1/3rd of XB1's GPU even in docked mode, doesn't that mean the bandwidth fits perfectly for the Switch? :)

I thought the whole discussion was how Switch is closer to Wii U than to Xbone or not.

But you're right, it would fit the GPU's raw power ratio. Which again doesn't fit with the "easy to port" feature that Switch supposedly has. Enforced also by the UE4 profiles. Because 1/3 processing power & 1/3 memory bandwidth & slightly poorer CPU (no. of cores) would rather point to some heavy downgrades even for docked mode.
 
Considering the target hardware's GPU is less than 1/3rd of XB1's GPU even in docked mode, doesn't that mean the bandwidth fits perfectly for the Switch? :)

It still does not put it closer to XB1/PS4 than to the Wii-U though:

Wii-U 12.5GB/s (20.8GB/s less than Switch)
Switch : 33.3GB/s effective
XB1: 68GB/s (34.9GB/s more than switch, even without taking into account 200GB/s ESRAM)

It's an impressive spec for a handheld / portable, but I think it's a stretch to claim it's closer to XB1 / PS4 than Wii-U in overall specs, it really is closer to the Wii-U.

The need to inflate it's capability reminds me a lot of the WUST threads, and look how that turned out, I think more realistic expectations should be considered personally.
 
I thought the whole discussion was how Switch is closer to Wii U than to Xbone or not.

But you're right, it would fit the GPU's raw power ratio. Which again doesn't fit with the "easy to port" feature that Switch supposedly has. Enforced also by the UE4 profiles. Because 1/3 processing power & 1/3 memory bandwidth & slightly poorer CPU (no. of cores) would rather point to some heavy downgrades even for docked mode.

Dude, i get what your saying, i really do.

I just think your being very hard on Switch.

The engine discrepancy for Wii u was by far the biggest technical hurdle for the console, because nobody was going to bother trying to port games over wholesale without actual support for the engines involved.

Switch has built in support for the most modern graphical rendering features and the engines that power them.

Even though the hardware is not that powerful, all that means is that the games have to be cut up and compressed to hell to fit.

Sometimes this will be terrible, but other times it will hold up well surely.

Look at Wii's version of Modern warfare. Holds up surprisingly well to the 360 and PS3 versions, and that power discrepancy was way higher than Switch to XB1/PS4.

It still does not put it closer to XB1/PS4 than to the Wii-U though:

Wii-U 12.5GB/s (20.8GB/s less than Switch)
Switch : 33.3GB/s effective
XB1: 68GB/s (34.9GB/s more than switch, even without taking into account 200GB/s ESRAM)

It's an impressive spec for a handheld / portable, but I think it's a stretch to claim it's closer to XB1 / PS4 than Wii-U in overall specs, it really is closer to the Wii-U.

The need to inflate it's capability reminds me a lot of the WUST threads, and look how that turned out, I think more realistic expectations should be considered personally.

Can't you atleast admit that its a mixed bag? Its a mixed bag. RAM Capacity is great, CPU power is pretty good..GPU grunt is eh, GPU architecture is excellent, bandwidth is somewhat concerning.
 
Time to start getting the duct tape, we can duct tape two Wii U's together but we can't duct tape an Xbox One to turn it into a Switch.

I'm joking
 
Can't you atleast admit that its a mixed bag? Its a mixed bag. RAM Capacity is great, CPU power is pretty good..GPU grunt is eh, GPU architecture is excellent, bandwidth is somewhat concerning.

I personally think it's an amazing spec for a portable / handheld.

I just don't think the term used here that it's closer to XB1/PS4 than a Wii-U is in any way accurate and yet it has often been repeated as fact in this and many other Switch threads unchallenged.

I don't think that specific claims stands up to any serious scrutiny, and as a spec thread this is the best thread to either challenge or confirm that perception.
 
Reminder:

http://www.nvidia.com/object/ces-2017.html?ncid=so-twi-cs27-4159

NVIDIA AT CES 2017
OPENING KEYNOTE ADDRESS
This January, NVIDIA will kick off CES and usher in a new year of technology innovation when co-founder and CEO, Jen-Hsun Huang, takes the stage to give the preshow keynote address.
See the industry's most exciting tech unveilings in artificial intelligence (AI), self-driving cars, and gaming.

January 4, 2017 | 6:30 p.m. PST
Location: Venetian, Level 5, Palazzo Ballroom
- See more at: http://www.nvidia.com/object/ces-2017.html?ncid=so-twi-cs27-4159#sthash.steKSw7x.dpuf

We know of rumours and FCC filings that a new Shield TV is going to be revealed at CES. I don't know if it will be revealed during the keynote. (But other things have like Project Xavier during the GTC Europe keynote.)
 
It still does not put it closer to XB1/PS4 than to the Wii-U though:

Wii-U 12.5GB/s (20.8GB/s less than Switch)
Switch : 33.3GB/s effective
XB1: 68GB/s (34.9GB/s more than switch, even without taking into account 200GB/s ESRAM)

It's an impressive spec for a handheld / portable, but I think it's a stretch to claim it's closer to XB1 / PS4 than Wii-U in overall specs, it really is closer to the Wii-U.

The need to inflate it's capability reminds me a lot of the WUST threads, and look how that turned out, I think more realistic expectations should be considered personally.

Switch bandwidth is probably closer to 40-45GB/s with the new texture compression together with pascals color compression which is a bit more efficient than the one in Maxwell. That is if Nvidia adds it to the maxwell hardware
 
That 33.3GB/s already includes the maxwell compression gains, it's 25GB/s without, you can't go and add it again. :P
 
Pretty soon Switch will break all the rules of physics.


Reminder:

http://www.nvidia.com/object/ces-2017.html?ncid=so-twi-cs27-4159



We know of rumours and FCC filings that a new Shield TV is going to be revealed at CES. I don't know if it will be revealed during the keynote. (But other things have like Project Xavier during the GTC Europe keynote.)

Damn, that's late in Europe, I will check the news tomorrow when I wake up. I'm really curious what they will show.
 
So everyone's expecting the new Shield to have the hardware we'd want in the Switch and the Switch to have significantly cheaper, weaker, more conservative specs?
 
Not a dig at NX, but USPTO... what you said is both true and very sad :/.

The USPTO (or any other country's office) has every reason to reject patents that are too broad, but only when what's being claimed has actually been done before. If someone wants to patent something so broad that they'll never be able to use all the coverage, as long as it's a novel and unobvious idea what's the problem?

Anyway this is getting off topic. Basically the point of the patent talk is to say there's no reason for the fan to work exactly like it says in the patent. Especially since it was filed in June of this year in the US, and its parent was filed June of last year in Japan. So many things could have changed.


I'm staying out of other Switch threads so I don't see games leaked, but have we heard anything new about the hardware? Or are we still on 28nm HPC+? Is that die size too big to add a wider bus for RAM?
 
So everyone's expecting the new Shield to have the hardware we'd want in the Switch and the Switch to have significantly cheaper, weaker, more conservative specs?

Well, it can likely afford to. Rumors say it's a shield tv, and thus no screen and handheld limitations to hold it back in terms of spec budget.
 
So why has the 16nm FinFET been disregarded as most plausible choice for the fab process of the SoC again?
 
So everyone's expecting the new Shield to have the hardware we'd want in the Switch and the Switch to have significantly cheaper, weaker, more conservative specs?

No screen, no need for portability and no battery to conserve, so will have much less restrictions on's it's overall design compared to Switch, hard to see why it wouldn't be more powerful.
 
So why has the 16nm FinFET been disregarded as most plausible choice for the fab process of the SoC again?

It wouldn't explain why a fan is necessary even at the docked speeds leaked by DF, unless Nintendo added more SMs or CPU cores which would likely be expensive. We want to keep our expectations low here since, based on those clock speeds, Nintendo doesn't seem to be going for incredible performance.
 
And probably twice the price tag.

Maybe 1.5 x the cost (Guess)...but cost does not equal the wholesale price to distribution channels.

It depends how much Nvidia mark up the shield v Nintendo mark up.

Nvidia have high mark ups on their high end GPU's, but seem very aggressive on their shield.

Nintendo may come out swinging right out of the gate after WiiU failure, but I believe more likely will be Nintendo like profits and possible reduction after 3-6 months before 2017 Xmas ......The first few months will sell out regardless.
 
So everyone's expecting the new Shield to have the hardware we'd want in the Switch and the Switch to have significantly cheaper, weaker, more conservative specs?
Rumors say that specs should be the same as the previous Shield, but with better efficiency thanks to pascal. So it's going to be pretty close to the Switch as well on paper.
 
Rumors say that specs should be the same as the previous Shield, but with better efficiency thanks to pascal. So it's going to be pretty close to the Switch as well on paper.

At this point the previous Shield is not that close to Switch on paper. Half the CPU clock and between 30%/76.8% of the GPU clock makes quite a difference. It's very likely that the new Shield will get even higher clocks considering the efficiency gains and looking at Parker for hints.
 
It wouldn't explain why a fan is necessary even at the docked speeds leaked by DF, unless Nintendo added more SMs or CPU cores which would likely be expensive. We want to keep our expectations low here since, based on those clock speeds, Nintendo doesn't seem to be going for incredible performance.
I just can't picture the worst case scenario being an option if Nintendo is truly intending to have the Switch running the console games/engines of this generation. A Maxwell micro-architecture at around 500~600 GFLOPS @ single precision, if not more when docked would probably fit the bill. That's me hoping though.
 
I just can't picture the worst case scenario being an option if Nintendo is truly intending to have the Switch running the console games/engines of this generation. A Maxwell micro-architecture at around 500~600 GFLOPS @ single precision, if not more when docked would probably fit the bill. That's me hoping though.

To be honest if the rumors about porting any PS4/XB1 game and development in general being fairly easy are true then it doesn't really matter much what kind of flops this thing gets. If they can achieve that with 150-400 GFLOPS then more power to them.
 
That 33.3GB/s already includes the maxwell compression gains, it's 25GB/s without, you can't go and add it again. :P
Erm, the 33GB/s is the effective BW for the average compression gain as shown by several titles. The maximum compression gain is 8:1.

ed: oh, you said 'maxwell'. Ignore my dyslexia.
 
Erm, the 33GB/s is the effective BW for the average compression gain as shown by several titles. The maximum compression gain is 8:1.

ed: oh, you said 'maxwell'. Ignore my dyslexia.

No worries. :)

However, theoretical maximums, especially & specifically when it comes to compression, are of little relevance, the average compression gain in real world use across multiple titles is significantly more relevant.

I mean look at something as simple as compression via RLE, the theoretical maximum compression gain from using RLE can approach unbelievable levels if you're just compressing a string of a billion "A"'s down to just "A10^9", the compression ratio is fantastically amazing, but the real world likelyhood of ever needing to RLE that content is probably zero, and in real world use RLE was surpassed many decades ago, not because it's theoretical maximium has been bettered by other algorithms, but because it's performance in real world usage has been significantly bettered.
 
At this point the previous Shield is not that close to Switch on paper. Half the CPU clock and between 30%/76.8% of the GPU clock makes quite a difference. It's very likely that the new Shield will get even higher clocks considering the efficiency gains and looking at Parker for hints.

CPU wise is a lot better, but the difference in GPU power when Switch is docked should be trivial even if 2SM is confirmed, especially considering that only one of these devices works with Android+shitty libraries. I have little doubts that Switch games will have better graphics than whatever runs in real time on the Shield. All i need to know now is if one of the architectural improvements they borrowed from Pascal is the 128bit bus to get those 51.2MB/s of memory bandwidth+whatever they'll use for cache.
 
CPU wise is a lot better, but the difference in GPU power when Switch is docked should be trivial even if 2SM is confirmed, especially considering that only one of these devices works with Android+shitty libraries. I have little doubts that Switch games will have better graphics than whatever runs in real time on the Shield.

Well, yeah, it's Android. I secretly hope someone will manage to make the Switch OS run on Shield TV as an experiment.

All i need to know now is if one of the architectural improvements they borrowed from Pascal is the 128bit bus to get those 51.2MB/s of memory bandwidth+whatever they'll use for cache.

This is maybe the most important improvement that can happen. And hopefully will.
 
Reminder:

http://www.nvidia.com/object/ces-2017.html?ncid=so-twi-cs27-4159



We know of rumours and FCC filings that a new Shield TV is going to be revealed at CES. I don't know if it will be revealed during the keynote. (But other things have like Project Xavier during the GTC Europe keynote.)

What do you guys think the implications would be if their new Shield products do in fact feature a chipset more capable than what's in the Switch as per Eurogamer's recent report?
 
What do you guys think the implications would be if their new Shield products do in fact feature a chipset more capable than what's in the Switch as per Eurogamer's recent report?

It just makes Nintendo looks worse than they already do. No one will actually buy the Shield but expect a 50+ page thread with a title like "Who is skipping Switch for the new Nvidia Shield?".
 
What do you guys think the implications would be if their new Shield products do in fact feature a chipset more capable than what's in the Switch as per Eurogamer's recent report?

Zero, different product for a different market with a different set of design goals.
 
I just can't picture the worst case scenario being an option if Nintendo is truly intending to have the Switch running the console games/engines of this generation. A Maxwell micro-architecture at around 500~600 GFLOPS @ single precision, if not more when docked would probably fit the bill. That's me hoping though.

I dont think Nintendo is intending that, based on what we heard about the specs.
 
I dont think Nintendo is intending that, based on what we heard about the specs.

I think Nintendo cares about running their games better than last iterations and hoping third parties create unique experiences for the Switch. By unique they may mean custom games catering to the power of the system rather than direct ports of BF1 and COD.
 
To be honest if the rumors about porting any PS4/XB1 game and development in general being fairly easy are true then it doesn't really matter much what kind of flops this thing gets. If they can achieve that with 150-400 GFLOPS then more power to them.

I dont see how those are not mutually exclusive. Either this chip got some secret sauce, or the rumours about easy porting are not true. I would go with the rumours not being true.
 
I'm not as technologically-inclined in this area, so can someone explain to me how two reliable insiders on GAF (Osiris and Matt) confirmed that PS4/XB1 releases could easily be ported over with little problem, but the Switch is much closer to a Wii U power?

Did Nintendo pull a Wii U and lower the actual specs compared to the initial ones given to devs?
 
I dont see how those are not mutually exclusive. Either this chip got some secret sauce, or the rumours about easy porting are not true. I would go with the rumours not being true.
Nobody said the ports would run at the same graphics quality. Guess how easy it is to "port" a PC game to a mid-range laptop?
 
I'm not as technologically-inclined in this area, so can someone explain to me how two reliable insiders on GAF (Osiris and Matt) confirmed that PS4/XB1 releases could easily be ported over with little problem, but the Switch is much closer to a Wii U power?

Did Nintendo pull a Wii U and lower the actual specs compared to the initial ones given to devs?

No but it's Nintendo so we have to keep digging until we can prove the good news to be false.
 
I'm not as technologically-inclined in this area, so can someone explain to me how two reliable insiders on GAF (Osiris and Matt) confirmed that PS4/XB1 releases could easily be ported over with little problem, but the Switch is much closer to a Wii U power?

Did Nintendo pull a Wii U and lower the actual specs compared to the initial ones given to devs?

The architecture of the Wii-U was a bitch, the architecture of the Switch is supposed to be a lot better and require less effort, for porting purposes easy is good, lots of effort is bad. :P.

But you have to keep expectations in check, the power differences mean that 1:1 ports from XB1/PS4 are close to impossible, there will be cuts, games will be scaled back in some areas etc. as typical when porting to a technically inferior platform.
 
I think Nintendo cares about running their games better than last iterations and hoping third parties create unique experiences for the Switch. By unique they may mean custom games catering to the power of the system rather than direct ports of BF1 and COD.
I agree with this. I didn't say the switch isn't a step up from Wii U.
What are you talking about? The most popular engine all we have heard and got confirmed is it is running that engine and getting plenty of games for it. Not the mobile engine the full engine.

Sure UE4 is running on it, that doesn't mean its easy to port any UE4 game to it, if its meant for hardware with 10 times more raw power.
 
I'm not as technologically-inclined in this area, so can someone explain to me how two reliable insiders on GAF (Osiris and Matt) confirmed that PS4/XB1 releases could easily be ported over with little problem, but the Switch is much closer to a Wii U power?

Did Nintendo pull a Wii U and lower the actual specs compared to the initial ones given to devs?

Nintendo never lowered the Wii U specs in the first place, they only bumped the CPU and GPU clocks (respectively from 1GHZ to 1.24 and from 400MHZ to 550). So...

No but it's Nintendo so we have to keep digging until we can prove the good news to be false.

... pretty much this.
 
I'm not as technologically-inclined in this area, so can someone explain to me how two reliable insiders on GAF (Osiris and Matt) confirmed that PS4/XB1 releases could easily be ported over with little problem, but the Switch is much closer to a Wii U power?

Did Nintendo pull a Wii U and lower the actual specs compared to the initial ones given to devs?

The WiiU devkits downgrade is only a rumor that had never been confirmed from a reliable source. It's a myth. After all the rumors for Switch goes in the other direction, that Dev kits got more power with each iteration.

It's easy to port because Switch/PS4/XOne have basically the same feature set, PS4/XOne are faster. But it is no problem to scale down some features to port a game to Switch.
 
The architecture of the Wii-U was a bitch, the architecture of the Switch is supposed to be a lot better and require less effort, for porting purposes easy is good, lots of effort is bad. :P.

But you have to keep expectations in check, the power differences mean that 1:1 ports from XB1/PS4 are close to impossible, there will be cuts, games will be scaled back in some areas etc. as typical when porting to a technically inferior platform.

This is what I was looking for--thanks!

Nintendo never lowered the Wii U specs in the first place, they only bumped the CPU and GPU clocks (respectively from 1GHZ to 1.24 and from 400MHZ to 550). So...



... pretty much this.

The WiiU devkits downgrade is only a rumor that had never been confirmed from a reliable source. It's a myth. After all the rumors for Switch goes in the other direction, that Dev kits got more power with each iteration.

I am 100% certain there was a dev on GAF about 6 months ago that quietly confirmed it happened in a thread. It was out of nowhere. I wish I had saved the post link :/
 
Status
Not open for further replies.
Top Bottom