Support NeoGAF

[PdotMichael]

People will just laugh at the old barely slower than Xbox One comments. Like it happens with the 1TF WiiU machine claims. Life goes on.

Of course it's a little funny how low setting are now more than good enough.

 

[Hoo-doo]

Because that's not all the system is. It's a hybrid, and the console side is totally gimped. You're fine with a portable Wii U. Obviously a lot of othe people aren't.

I obviously get that, but the disappointment and perceived 'gimping' of the console is all relative and caused by the 'theoretical' power the unit was seemingly meant to have through all this tech nonsense and number fudging that's continuously happening in these threads.

What good does it do to keep doubling down on hopeful theoreticals and upward rounding of numbers when the device is likely to be rather modest in power?
If you're aiming so high and expecting so much, it's going to be increasingly hard to appreciate the device for what it actually will be.
It's going to be a souped up Wii-U that you can take on the go that'll have every single game Nintendo will produce for the foreseeable future. How does that not excite people enough?

 

[SirShandy]

Maybe because some people don't want to play games on portables anymore?

The Switch is a hybrid, it does not have to be used as a portable if you don't want to. You can sit down on a couch with a traditional controller and play it like a console.

 

[SeanIfOnly]

Does it play Mario, Splatoon, Zelda and Mario Kart... does it offer something new..

Ok sounds like a solid system to me. Perfect to pair with my Xbox One or PS4

 

[AlStrong]

11 mm x (11 mm + 3 mm) = 154 mm^2

Or perhaps

11 mm x (11 mm + 4 mm) = 165 mm^2

Not really outside the realm of possibility. But it would potentially alter yields.

You get roughly 30-40% more chips per wafer @ 121mm^2 compared to those two.

You'd think dead space would still be cheaper (and more energy efficient) than fabbing a whole extra module that wouldn't be used though, right? At the very least from a use of materials standpoint.

If you can shove an extra IP block on there and the block ends up being non-functional because of a wafer defect etc., then it doesn't matter if the area was already accounted for. Lithography is just exposure of the photomask in several stages, so it doesn't add extra if the area is fixed already. The main cost is enlarging the overall chip area.

If it does work and the same IP block on another location on the chip doesn't work, then you still get a functional chip. That's what redundancy is about, and why 2013 consoles have 2 extra CUs or why PS3's Cell was 8 SPEs with only 7 SPEs enabled. In the PS3 case, it wouldn't make sense to only expose 7 SPE's on the photomask leaving an 8 spot blank. For retail models, the 8th SPE is simply disabled, functional or not, so power consumption isn't an issue there.

i.e. if you're thinking of adding a 3rd IP block and the layout means you can add a 4th to maintain the rectangle, the 4th might as well be put in there instead of dead space.

If the layout means there's not enough space for a 4th, then the job is done or you space out the rest of the IP blocks to reduce thermal density.

 

[AlStrong]

double post.

 

[Schnozberry]

You get roughly 30-40% more chips per wafer @ 121mm^2 compared to those two.

Indeed. It's too bad 450mm wafers won't come at TSMC until 2018.

 

[TunaLover]

I can't imagine they're going to let the CPU be a huge bottleneck for this thing for no good reason, particularly since they should have the ability to upclock it just with an OS update.

I remember LCGeek and Nate stating that the CPU outperform the ones on PS4/XB1 by quite margin, I know that isn't a big feat but still. That change now that we know the clock speeds?

 

[EDarkness]

I obviously get that, but the disappointment and perceived 'gimping' of the console is all relative and caused by the 'theoretical' power the unit was seemingly meant to have through all this tech nonsense and number fudging that's continuously happening in these threads.

What good does it do to keep doubling down on hopeful theoreticals and upward rounding of numbers when the device is likely to be rather modest in power?
If you're aiming so high and expecting so much, it's going to be increasingly hard to appreciate the device for what it actually will be.
It's going to be a souped up Wii-U that you can take on the go that'll have every single game Nintendo will produce for the foreseeable future. How does that not excite people enough?

I don't really buy Nintendo games, so that aspect of it doesn't matter to me. I have a Wii U, so why would anyone want another Wii U? Modest specs out of what could have been is the frustrating part, and as I've said in the past the problem with hybrid systems is exactly this. The handheld side has to win out since that's the limiting factor. I can't speak for anyone else, but I don't care about the portable side at all, and I feel like console guys are getting the shaft. Of course it will all come down to games, but still, I don't think anyone would have expected Nintendo and Nvidia to hamstring the system so severely.

People are just wondering why they would do such a thing, I think.

 

[Skittzo0413]

I remember LCGeek and Nate stating that the CPU outperform the ones on PS4/XB1 by quite margin, I know that isn't a big feat but still. That change now that we know the clock speeds?

If the clock speeds and the specs which list 4 A57 cores are accurate then the CPU is probably marginally better than the Wii U's. Which would be weird, when they've specifically talked about that being a bottleneck.

 

[TunaLover]

If the clock speeds and the specs which list 4 A57 cores are accurate then the CPU is probably marginally better than the Wii U's. Which would be weird, when they've specifically talked about that being a bottleneck.

Wii U CPU was a big headache for unoptimized UE games, it didn't help it was Wii legacy architecture to keep BC.

 

[Xdrive05]

If the clock speeds and the specs which list 4 A57 cores are accurate then the CPU is probably marginally better than the Wii U's. Which would be weird, when they've specifically talked about that being a bottleneck.

Wow is that true? I thought it was a given that, even at half speed, the Switch CPU is still way better than Wii-U's, because that was a cobbled together three core implementation of the GC CPU?

 

[Donnie]

Nah 4x A57 at 1Ghz is still quite a bit better than Espresso for most modern code. If you're dealing with direct ports then there could be a problem as its then a three core split with code suited to Espresso.

Still don't think its just 4X A57 though.

 

[Skittzo0413]

Wow is that true? I thought it was a given that, even at half speed, the Switch CPU is still way better than Wii-U's, because that was a cobbled together three core implementation of the GC CPU?

That was what I gathered based on Blu's reaction, but that may have been more referring to direct ports from Wii U as Donnie says:

Nah 4x A57 at 1Ghz is still quite a bit better than Espresso for most modern code. If you're dealing with direct ports then there could be a problem as its then a three core split with code suited to Espresso.

Still don't think its just 4X A57 though.

That makes more sense. Either way, it's a bit surprising that, when Miyamoto even talks about about how bad the Wii U's CPU is, they would choose again to use a fairly low powered CPU. Hopefully it is more than 4x A57s, as DF has again explicitly said they can't confirm the final CPU or GPU configurations.

Matt also explicitly told us that the Switch would not be as CPU limited (likely relative to its own GPU) as PS4 and XB1 are. Is that something that would still be true for 4x A57s at 1GHz?

 

[z0m3le]

Nah 4x A57 at 1Ghz is still quite a bit better than Espresso for most modern code. If you're dealing with direct ports then there could be a problem as its then a three core split with code suited to Espresso.

Still don't think its just 4X A57 though.

The real issue is that espresso has an insanely short pipeline and a higher clockspeed, when it isn't behind, it will jump way ahead with smaller pieces of code, that is the issue.

Though there is an extra core here to help with that.

 

[Rodin]

I remember LCGeek and Nate stating that the CPU outperform the ones on PS4/XB1 by quite margin, I know that isn't a big feat but still. That change now that we know the clock speeds?

Yes, maybe they didn't know the clockspeed.

If the clock speeds and the specs which list 4 A57 cores are accurate then the CPU is probably marginally better than the Wii U's. Which would be weird, when they've specifically talked about that being a bottleneck.

Wait, if you're referring to blu's post i don't think that's what he said. "Possibly having troubles with Wii U ports" doesn't mean that it's marginally better, it's still a lot better but not to the point that some parts of the code heavily optimized for how Espresso worked would easily translate to the new architecture.

 

[z0m3le]

You guys just can't resist rounding the numbers upwards, upwards and upwards.

Go back and read the 'Nintendo Switch: Powered by Custom Nvidia Tegra Chip (Official)' thread again.
It's almost eerie to see people just take something and extrapolate and extrapolate upwards until it finally satisfies some vague subconscious demand for more processing power. It's really not worth the disappointment in the end.

An affordable and fully portable Wii-U with a semi-decent battery life is freaking great news. Why can't people be satisfied with that?

393gflops vs ~400gflops? We know what maxwell vs gcn is as well, and we know what gcn is vs vliw5... Wii u's 176gflops in vliw5 is = to about 128gflops gcn, which is = to about 92gflops maxwell. We aren't talking about new features like checkerboard rendering or fp16, as a handheld switch directly compares to wii u with over 60% better performance.

Docked it is over 4times better performance, it is just shy of half the performance of xb1 without new features. There is no smoke and mirrors about these numbers though I will point out that as developers push more for vulkan, unless Nvidia's drivers got better, that architecture advantage shrinks a lot. We don't have to worry about this being a wii u, it is what Wii u should have been from the start.

 

[TunaLover]

What do you think the RAM split will be for Switch? 1gb for OS still?

 

[TwilightPrincess]

Yes, maybe they didn't know the clockspeed.



Wait, if you're referring to blu's post i don't think that's what he said. "Possibly having troubles with Wii U ports" doesn't mean that it's marginally better, it's still a lot better but not to the point that some parts of the code heavily optimized for how Espresso worked would easily translate to the new architecture.

4xA57 even at the same spead shouldn't be better than 6-8 Xbox One jaguar cores in reality. If its 4x A57 LCGeeks rumour/leak was wrong, simple as that. 4xA57 are not going to be 20-30% stronger than XboxOne CPU even at full clockspeed, downclocked even less.

 

[z0m3le]

What do you think the RAM split will be for Switch? 1gb for OS still?

No reason to discount the rumor of 3.2gb for games. Every rumor has pretty much been right so far, only one that was wrong was the Foxconn one, though I'm surprised Foxconn didn't leak anything, very unlike them.

 

[blu]

Nah 4x A57 at 1Ghz is still quite a bit better than Espresso for most modern code. If you're dealing with direct ports then there could be a problem as its then a three core split with code suited to Espresso.

Still don't think its just 4X A57 though.

That was what I gathered based on Blu's reaction, but that may have been more referring to direct ports from Wii U as Donnie says:

That makes more sense. Either way, it's a bit surprising that, when Miyamoto even talks about about how bad the Wii U's CPU is, they would choose again to use a fairly low powered CPU. Hopefully it is more than 4x A57s, as DF has again explicitly said they can't confirm the final CPU or GPU configurations.

Matt also explicitly told us that the Switch would not be as CPU limited (likely relative to its own GPU) as PS4 and XB1 are. Is that something that would still be true for 4x A57s at 1GHz?

Donnie is right, I was referring to direct ports from wiiU, where the core count and IPC x clock (negative) difference would not leave much leeway in some situations. Of course, on the average the notably higher IPC of the speculated A57 cluster and absolutely superior SIMD would be a clear improvement over Espresso. But my reaction was not so much in relation to Espress as it was in relation to some other cores..

Anyhow, since we inevitably keep returning to metrics like CPU FLOPS where we have to rely mainly on Geekbench scores, I took the liberty to put together this small SGEMM test. It's not so much a Geekbench wanna-be, as a test that avoids one of Geekbenches' gotchas where some datasets run entirely off L2 caches. Also, it's open and runs on any platform which has a gcc-compliant toolchain. If would be curious to see the results if somebody came across a Jetson TX1 as nvidia claim some exceptional-for-the-uarch performance (and I'm without a JTX1 myself).

 

[Cerium]

I remember LCGeek and Nate stating that the CPU outperform the ones on PS4/XB1 by quite margin

Man I forgot that Nate said that on top of Pascal and everything else.

 

[KingSnake]

I remember LCGeek and Nate stating that the CPU outperform the ones on PS4/XB1 by quite margin, I know that isn't a big feat but still. That change now that we know the clock speeds?

LCGeek said in the DF thread that she didn't know about the speed.

No reason to discount the rumor of 3.2gb for games. Every rumor has pretty much been right so far, only one that was wrong was the Foxconn one, though I'm surprised Foxconn didn't leak anything, very unlike them.

Except for the fact that the rumour about 3.2GB for games doesn't come from a certified source who has been right before. It comes from Vern here on GAF.

 

[PdotMichael]

Man I forgot that Nate said that on top of Pascal and everything else.

So was he right with anything outside of "confirming" what other people leaked?

 

[ggx2ac]

You guys just can't resist rounding the numbers upwards, upwards and upwards.

Go back and read the 'Nintendo Switch: Powered by Custom Nvidia Tegra Chip (Official)' thread again.
It's almost eerie to see people just take something and extrapolate and extrapolate upwards until it finally satisfies some vague subconscious demand for more processing power. It's really not worth the disappointment in the end.

An affordable and fully portable Wii-U with a semi-decent battery life is freaking great news. Why can't people be satisfied with that?

393gflops vs ~400gflops? We know what maxwell vs gcn is as well, and we know what gcn is vs vliw5... Wii u's 176gflops in vliw5 is = to about 128gflops gcn, which is = to about 92gflops maxwell. We aren't talking about new features like checkerboard rendering or fp16, as a handheld switch directly compares to wii u with over 60% better performance.

Docked it is over 4times better performance, it is just shy of half the performance of xb1 without new features. There is no smoke and mirrors about these numbers though I will point out that as developers push more for vulkan, unless Nvidia's drivers got better, that architecture advantage shrinks a lot. We don't have to worry about this being a wii u, it is what Wii u should have been from the start.

I think what Hoo-doo meant is that some people are too optimistic thinking that the Switch will end up being better than what is currently reported.

It may help to look at how bad the specs could go and not just how good they can be, like with how I disagree that Nintendo would go for 16nm and instead go for 28nm considering the inconsistencies with the clock speeds.

That's not to say it's true, it's just I'm finding some blind optimism here and there like some people now thinking that the SCD is a guarantee and that it will be 4 TFLOPS and cost $200 and provide 4K gaming. What makes me question that for example is that Nintendo are in the game of making things affordable and peripherals do not sell that well to a console userbase. If an SCD did ever happen I'd expect it to use the same Tegra tech possibly at stock clock speeds so that it could be cheaper than $100 along with using features like Wi-fi to make the visuals in portable mode better as stated in the patent.

 

[Gamer79]

Outside of Zelda running on jimmy kimmel I don't know of any other real game footage. It will be interesting to see.

 

[atbigelow]

I think what Hoo-doo meant is that some people are too optimistic thinking that the Switch will end up being better than what is currently reported.

It may help to look at how bad the specs could go and not just how good they can be, like with how I disagree that Nintendo would go for 16nm and instead go for 28nm considering the inconsistencies with the clock speeds.

That's not to say it's true, it's just I'm finding some blind optimism here and there like some people now thinking that the SCD is a guarantee and that it will be 4 TFLOPS and cost $200 and provide 4K gaming. What makes me question that for example is that Nintendo are in the game of making things affordable and peripherals do not sell that well to a console userbase. If an SCD did ever happen I'd expect it to use the same Tegra tech possibly at stock clock speeds so that it could be cheaper than $100 along with using features like Wi-fi to make the visuals in portable mode better as stated in the patent.

It isn't really even being reported. Clock speeds are not the full picture for anything.

 

[ggx2ac]

It isn't really even being reported. Clock speeds are not the full picture for anything.

Of course, but it definitely constrains what the set up of the system will be.

 

[japtor]

I'd guess (hope?) that dynamic resolution will be pretty common, particularly since it seems like it'd be a key part of the basic switching functionality.

 

[orioto]

I'd guess (hope?) that dynamic resolution will be pretty common, particularly since it seems like it'd be a key part of the basic switching functionality.

What's sad is that it feels like we're not going to see a lot of native rez on this... I'd somewhat prefer games scaled or the console, even the portable, that look super good at 720p, rather than crappy iq "next gen" games

 

[Cerium]

So was he right with anything outside of "confirming" what other people leaked?

No he wasn't. Unintentionally or not, for months he sold the fantasy of a Switch that was much much much more powerful than what we ended up getting. And now on Twitter he says it's all about the software.

I do feel betrayed. I really believed in him.

 

[LCGeek]

4xA57 even at the same spead shouldn't be better than 6-8 Xbox One jaguar cores in reality. If its 4x A57 LCGeeks rumour/leak was wrong, simple as that. 4xA57 are not going to be 20-30% stronger than XboxOne CPU even at full clockspeed, downclocked even less.

Except someone did his homework and it is competitive.

A customized ARM CPU not on windows which reserves power could easily pull that, regardless of console/pc would allow for that number.. However that won't be happening and it's downclocked. I've already stated I didn't expect downclocks. Yet be like others and forget I did nail what type of CPU it was going to end up with, though to be fair thraktor/blu were on the money once hints came out.

 

[lwilliams3]

http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/4

I did the math. Again it is a rough estimation, but GCN further improved on the efficiency that VLIW4 offered.



This is mainly why VLIW4 replaced VLIW5, AMD could remove 1ALU for every 5 they had and have roughly the same performance, so 160 cores would work out to 128, of course Wii U has 176gflops, those extra 16gflops would be further covered by the move from VLIW4 to GCN, which moved from 4alu groups to 16alu groups, allowing the 128 cores (alus) to handle multiple thread waves at once, think of it like a catch all, threads would feed into a wider bandwidth of processing than before, without changing the available number of cores, and since you could now do multiple instructions at once, you don't have to worry about ALUs going unused. VLIW4 to GCN was from a processing standpoint smaller than VLIW5 to VLIW4, hope you are satisfied with that, you can read more about it and anandtech is a great deep dive into VLIW4/5 and GCN if you search their archives. This is what the WUST thread did to me btw.

As for maxwell over GCN, I used ~28% better performance, in reality it can be as much as 40% but like I said this is a rough estimation on the safe side, I don't even add in the added functionalities like just being able to do certain effects better and more efficiently than R700's 2008 VLIW5 engine was capable of.

To everyone being down right dense, yes Zelda on docked Switch could do 1080p 60fps, it is straight up 4times as powerful, before feature enhancements , that is enough to handle the game at the higher demand, the only issue is that wii u's architecture is very different, while PS4, XB1 and Switch roughly have the same structure. IE a bad Wii U port can still run badly on Switch, while a bad ps4 port should be handled much better since PS4 and Switch are much more alike.

Thanks to your replies. I admired your long number-crunching posts since the WUST days.

 

[Plinko]

What's sad is that it feels like we're not going to see a lot of native rez on this... I'd somewhat prefer games scaled or the console, even the portable, that look super good at 720p, rather than crappy iq "next gen" games

Where is this coming from?

 

[vern]

LCGeek said in the DF thread that she didn't know about the speed.



Except for the fact that the rumour about 3.2GB for games doesn't come from a certified source who has been right before. It comes from Vern here on GAF.

I would argue that I'm reliable. 😂

Wonder why no one else has leaked that yet though? Anyone got any thoughts? Seems weird with all these leaks I'm still the only one claiming to know the RAM split.

 

[orioto]

Where is this coming from?

Everyone is talking about dynamic resolution and that it's the new big fad and how it will allow PS4 ports on Switch..

 

[LordKano]

I would argue that I'm reliable. 😂

Wonder why no one else has leaked that yet though? Anyone got any thoughts? Seems weird with all these leaks I'm still the only one claiming to know the RAM split.

I have an idea

 

[Vena]

I would argue that I'm reliable. 😂

Wonder why no one else has leaked that yet though? Anyone got any thoughts? Seems weird with all these leaks I'm still the only one claiming to know the RAM split.

There's actually still very little leaked.

All we have are clockspeeds.

 

[magash]

There's actually still very little leaked.

All we have are clockspeeds.

Speaking about very little being leaked does no one find it a bit strange that exact frequencies for the CPU and GPU are leaked but nothing else? I mean nothing has been released about the configuration, number of CUDA cores, the exact customizations Nintendo has implemented for the GPU; and on the CPU side we only know of the implementation of an A57. Nintendo tends to almost always customize and configure the components that they use especially on the CPU and the GPU side of things. I find it a bit striking that we haven't heard anything about alterations or customizations to the A57 that is purportedly being used to power the Switch.

 

[NateDrake]

No he wasn't. Unintentionally or not, for months he sold the fantasy of a Switch that was much much much more powerful than what we ended up getting. And now on Twitter he says it's all about the software.

I do feel betrayed. I really believed in him.

I never spoke about the CPU. I said Switch would be 3x Wii U based on the information I was being told. I wouldn't say 2.5x is a major decrease from 3x. Pascal was the only miss my contacts have had, and the custom Maxwell is using aspects of Pascal as reported by EG this week. Like I've said, and will continually have to repeat, Nintendo looked into Pascal, and other outlets had heard the same - up to a 99% certainty it was being used.

And the RAM information going around for the OS is accurate as in 3GB of RAM for games.

 

[vern]

There's actually still very little leaked.

All we have are clockspeeds.

Others had the same total RAM as me though. Didn't Emily or LKD also say 4gb? I'm not the only one suggesting that number.

 

[Malakai]

Can someone explain why would Eurogamer write out an whole article on speculation when the only thing they have confirmed are the clockspeed for the CPU and GPU?

 

[Kingpin Rogers]

I'm still thinking we'll see a 'real' Switch console in 2018.
A device that is more powerful and focused on TV output. Will play the same games as the Switch too.

If they did that without announcing it before the Launch of the Switch then I would be finished with Nintendo. It would be incredibly bullshit and they'd deserve to lose a lot of customers.

 

[KingSnake]

I think now we reached the denial levels of WUST post Arkham.

 

[Malakai]

I'm still thinking we'll see a 'real' Switch console in 2018.
A device that is more powerful and focused on TV output. Will play the same games as the Switch too.

Well, Unreal 4 had the code names Wolf Air for handheld mode and Wolf Sea for docked mode. Maybe Wolf Land is home console only code. LOL...

 

[OG_Original Gamer]

I never spoke about the CPU. I said Switch would be 3x Wii U based on the information I was being told. I wouldn't say 2.5x is a major decrease from 3x. Pascal was the only miss my contacts have had, and the custom Maxwell is using aspects of Pascal as reported by EG this week. Like I've said, and will continually have to repeat, Nintendo looked into Pascal, and other outlets had heard the same - up to a 99% certainty it was being used.

And the RAM information going around for the OS is accurate as in 3GB of RAM for games.

I wonder if the 2SM configuration still stand.

 

[vern]

I never spoke about the CPU. I said Switch would be 3x Wii U based on the information I was being told. I wouldn't say 2.5x is a major decrease from 3x. Pascal was the only miss my contacts have had, and the custom Maxwell is using aspects of Pascal as reported by EG this week. Like I've said, and will continually have to repeat, Nintendo looked into Pascal, and other outlets had heard the same - up to a 99% certainty it was being used.

And the RAM information going around for the OS is accurate as in 3GB of RAM for games.

At least I got backup on the RAM from someone. 👌🏼

 

[Vena]

Others had the same total RAM as me though. Didn't Emily or LKD also say 4gb? I'm not the only one suggesting that number.

Sorry, I meant specific details. 4GB of RAM was known for a while indeed, but I was speaking of the fine minutia like how we now have clocks but little else.

 

[Schnozberry]

I wonder if the 2SM configuration still stand.

Seems most likely. Any more would be a sweet bonus.

 

[kIdMuScLe]

If they did that without announcing it before the Launch of the Switch then I would be finished with Nintendo. It would be incredibly bullshit and they'd deserve to lose a lot of customers.

Did you abandoned Sony when the PS4 pro came out?