Nintendo Switch Tech Speculation discussion

[Goodtwin]

So. . . You take the performance of the handheld for the user that doesn't care about the function rather than the docked performance? Just thought I'd catch that illogical argument for you.

The docked performance is somewhere around 40% of the xb1, with the cpu more or less matching 4 ps4 cpu cores. This isn't a Wii U. The interesting thing about the design, assuming 4 A57 cores and maxwell 2sm, it should have a ton of thermal overhead even docked, so if they did run into a cpu hurdle, there really is no reason I can think of that would lock them out of a higher clock on the cpu.

We've seen it with the psp cpu going from 2xxmhz to 3xxmhz when God of war was released, as for gpu, it should be able to handle any game running on xb1 with the appropriate reductions.

Even without western 3rd party support, it will see a far superior library than wii u simply because Nintendo won't be dividing their attention equally across 2 major platforms. If wii u had 3ds' entire library on top of what was there, it would have not suffered the droughts it had and sold far better.

The idea that it needs to "win" the console war is just school yard nonsense, Nintendo needs to have a platform that can make a profit. I can't tell you the future, but switch should be more successful than wii u and surpass it within the first two years regardless of 3rd party support.

That's exactly right. I know some people cant fathom a scenario where a consumer cares greatly about Nintendo's first party games, they want a system they can take on the go, and really just offers something unique from the other gaming platforms on the market. For many people, they have a very narrow perception of what makes a good gaming console. Basically, the blueprints are this (think Xbox and PlayStation), and deviating from it foolishness. As a standalone console that plays games on the TV, primarily for Madden, COD, Assassins Creed, Fallout, and Batman, yes, the Switch will likely be the worst gaming console on the market for this purpose. I honestly believe most people that support the direction Nintendo has taken with the Switch are in agreement with this, but that's not the appeal of the Switch now is it.

Unified support from Nintendo's first party developers is the most obvious positive with the Switch. No longer will Nintendo have to split resources between two platforms. This not only means a lot more games from Nintendo on the platform than ever before, but also less redundancy. Mario Kart 7 and Mario Kart 8 are now just Mario Kart Switch with a team freed up to create something else. Hopefully this means neglected IP's will be resurrected, and new IP's created. Being portable is big for a lot of gamers. You cant play your PS4 or X1 on go. I know some people think being able to fit in your pocket is a big deal, but the most popular 3DS models are the XL models, and they hardly fit comfortably in a pocket, and certainly do not fit an a childs pocket. The 3DS and Vita have been a saving grace for many Japanese developers, many of which do not support traditional consoles. With the Vita and 3DS on their way out, which platform do you think those third parties are most likely to support.

Natedrake over at Neogaf is saying he is hearing battery life isn't as terrible as early reports suggested. He is hearing 3DS battery life, so around 4-5 hours. That is respectable, and if it does use USC C then that should be enough to not be problem for most people.

 

[Rootax]

So. . . You take the performance of the handheld for the user that doesn't care about the function rather than the docked performance? Just thought I'd catch that illogical argument for you.

The docked performance is somewhere around 40% of the xb1, with the cpu more or less matching 4 ps4 cpu cores. This isn't a Wii U. The interesting thing about the design, assuming 4 A57 cores and maxwell 2sm, it should have a ton of thermal overhead even docked, so if they did run into a cpu hurdle, there really is no reason I can think of that would lock them out of a higher clock on the cpu.

...

You can take the performance of the docked too, it's not night and day imo. Yeah, more gpu power (but still weak), same cpu power.... Plus, some devs will probably run their game the same, dock or undock, so choose the undock speed if you believe Eurogamer. And where is a 1ghz A57 about the same perf as a 1.6ghz jaguar (even with just 4 cores) ?

I agree that the unification is a good thing for Nintendo's devs. But still I doubt it will be enough. Well, it's will be enough if you think (and a lot of ppl do) that Nintendo is on his own island, and don't need third party, and Nintendo consumer are ok with that. Some are, but I feel a lot can't stand this anymore.

Oh well, we'll see anyway, in any case it will be interesting.

 

[Shifty Geezer]

That's exactly right...

I think your being unfair to the counterargument. I agree mostly with your assessment, but my criticisms, as I believe the rest in my corner are, are focussed at the larger business prospects and future of Nintendo as a brand and console player. As a handheld, a 3DS replacement, Switch is fine. As a means to grow Nintendo into a significant third player for the conventional core gamer, it's not. And that means less unit sales for Nintendo. The view as to what a console should be, which you consider narrow minded, is based on what sells to this market, as proven over 40 years of natural selection. Dual-stick TV based gaming with good graphics for the time is what the majority of the market wants. Ergo, to sell to that market you need to offer that experience. Deviation from that plan generally results in lower interest unless you hit a new audience, which typically is short lived because they are outside the normal hobby an have no long term interest in the new experience.

How many Switch (or Switches?) will Nintendo need to sell to maintain their current size? Will it grow the brand so next hardware, they have a stronger position for a new console? Or are they out of the running altogether in that regard? Does that mean they'll be bringing their games to the other consoles? A Japanese 1-2, Sony console with Nintendo game exclusivity, would probably be great for both parties.

It's in the context of that possible future, versus one where Nintendo becomes the dominant player as they used to be, that Switch is being criticised as underwhelming.

 

[Deleted member 13524]

The docked performance is somewhere around 40% of the xb1,
(...)
as for gpu, it should be able to handle any game running on xb1 with the appropriate reductions.

No developer can ever make games for docked mode only, so that's not the mandatory minimum target for developers.
At 300MHz, a 2SM GPU would have around 11.7% the raw FP32 performance of the Xbone.

Good luck with handling "any game with the appropriate reductions" on a GPU that is 9x weaker than the currently weakest platform, which already renders most games at 900p or below.

 

[sebbbi]

The docked performance is somewhere around 40% of the xb1, with the cpu more or less matching 4 ps4 cpu cores.

If the Eurogamer article holds, it will be Tegra X1 with 768 MHz GPU when docked. CPU 4 cores always at 1.02 GHz.

768 MHz * 256 MAD/cycle * 2 flop/MAD = 393216 Mflop/s = 0.39 Tflop/s. Xbox One GPU = 1.31 Tflop/s. That is 3.3x difference of GPU flops. Maxwell architecture is newer than GCN2. It achieves better utilization, but async compute improves GCN2 utilization (many devs have claimed up to 30% gains).

The CPU performance difference is similar. Cortex A57 and AMD Jaguar have similar IPC. 8x Jaguar cores at 1.6 GHz vs 4x A57 cores at 1.02 GHz. That is 3.1x difference of CPU performance. Jaguar of course needs software that scales well to high core counts. We still don't know Nintendo's CPU reservation. If a single core is reserved, it will be 3 cores vs 7.

Bandwidth: TegraX1 = 25.6 GB/s vs Xbox One = 68 GB/s + 204 GB/s ESRAM. Hard to compare since ESRAM is situational and Maxwell has delta color compression and tiled rasterizer. Both save bandwidth, but only in graphics workloads. Compute shaders are not affected.

If we believe these Eurogamer specs, the difference could be as high as 3x (in docked mode) vs Xbox One. This ballpark estimate is assuming AAA games: full utilization of all Jaguar CPU cores and lots of async compute on GPU.

These specs would make Switch around 2x faster than last gen consoles when docked and around equal to last gen consoles when handheld. Don't get me wrong. These are fantastic specs for an handheld device. Carrying Xbox 360 in your pocket is great. 720p was the most common last gen resolution. And it is a nice bonus that you can connect the handheld to your HDTV and play at 1080p (2.25x pixel count). No other handheld has done this.

 

[sebbbi]

Power consumption of current gen consoles measured:

Xbox One S (slim) is ~60 watts in gameplay. A mobile device of this size has to consume less than 10 watts (including display). Even less, if we assume 4 hour battery life (100% gameplay). It would be physically impossible to design a device that has 6x+ higher perf/watt than a die shrinked 20 nm console AP. Thus nobody should expect Nintendo Switch to match Xbox One S in performance.

People seem to have way too high expectations towards mobile devices. Phone/tablet peak flop (race to sleep) marketing has confused everybody. Just hold your hand near the exhaust of a current gen console when you play, and you understand that no mobile device could ever generate that much heat and offer 4 hours of game play. And less heat means less performance, unless we have a reason to expect drastically improved perf/watt over AMDs 20 nm designs. 1/3 perf of 60 watt console on a mobile device should be considered a good achievement, not a letdown.

 

[function]

Around 50% of modern game engine frame time goes to running compute shaders (lighting, post processing, AA, AO, reflections, etc). Maxwell's tiled rasterizer has zero impact on compute shaders. 25.6 GB/s is pretty low as everybody knows that 68 GB/s of Xbox One isn't that great either. ESRAM is needed to reach good performance. But I am talking about the POV of down porting current gen games to Switch. Switch certainly fares well against last gen consoles, and Maxwell's tiled rasterizer would certainly help older pixel + vertex shader based renderers. Too bad last gen consoles already got their last big AAA releases year ago. Easy ports between Xbox 360 and Switch are not available anymore. Xbox One is a significantly faster hardware. Straightforward code port is not possible. Content also needs to be simplified.

Thanks, I didn't know compute took up such a high proportion of GPU time!

I find myself wondering if Nintendo have such a focus on compute yet though, and if this might be a factor into their specifications for NX. One would assume Nvidia would be in an ideal position to advise, but otoh Nintendo always seem to be focused on what they require for their own software for the next 4~5 years when they're looking at hardware (i.e. N64, Wii, WiiU, 3DS).

Bandwidth: TegraX1 = 25.6 GB/s vs Xbox One = 68 GB/s + 204 GB/s ESRAM. Hard to compare since ESRAM is situational and Maxwell has delta color compression and tiled rasterizer. Both save bandwidth, but only in graphics workloads. Compute shaders are not affected.

On X1 couldn't you do some compute tasks from the esram (e.g. post process filters, motion blur), potentially while still running graphics workloads (perhaps shadow rendering to main memory?)

Power consumption of current gen consoles measured:

Xbox One S (slim) is ~60 watts in gameplay. A mobile device of this size has to consume less than 10 watts (including display). Even less, if we assume 4 hour battery life (100% gameplay). It would be physically impossible to design a device that has 6x+ higher perf/watt than a die shrinked 20 nm console AP. Thus nobody should expect Nintendo Switch to match Xbox One S in performance.

Isn't X1S on TSMC 16 nm? I don't think there's even a hint of a node advantage for NX to benefit from vs 2016 systems. It's possible it might even be on an older node than the X1S...

People seem to have way too high expectations towards mobile devices. Phone/tablet peak flop (race to sleep) marketing has confused everybody.

Yeah, this.

 

[Entropy]

IF we are comparing chips built on comparable processes, similar die size and similar technology level, then we can estimate the sustainable performance difference between a mobile SoC (est. 4W) and a stationary chip (est. 100W). If we are on the quadratic part of the power vs. frequency curve, that would yield a factor of five. If the stationary device is better modelled by power of three then the difference would be a factor of three. A hypothetical docked mobile SoC fulfilling the same criteria above but at just over 10W at the quadratic part would be a factor of three weaker than the 100W stationary part.
(In reality of course, none of the three conditions above are necessarily true. Performance/W of the XBoxOne on 28nm was better than that of the XBox360 for instance. A mobile SoC would likely be smaller, and a smaller device can be produced on a more expensive process at the same cost, and so on. )
What we have observed for some time is that performance of mobile solutions have grown faster than that of stationary devices, and are now at a point where something like the Switch is possible at all. Referring to forty years where this hasn't been the case isn't really meaningful. The question is whether a device that offers these capabilities is a compelling option to consumers. I'd say that it is - undeniably the general trend in both computing and entertainment is towards favouring mobile consumption. The Switch may or may not deliver the desired compromise at a compelling price.
We'll see about that soon enough.

 

[sebbbi]

Thanks, I didn't know compute took up such a high proportion of GPU time!

Last gen made post processing, ambient occlusion, shadow maps and per pixel lighting must have features. Geometry rasterization (to g-buffer) generally takes only 30%-40% of the frame.

If you don't use compute for post processing, AO and lighting, you are instead rendering a fullscreen quads. Zero overdraw and no geometry bottleneck -> no advantage from Nvidia's tiled rasterizer or better geometry performance compared to AMD. However DCC helps a bit. Maxwell has very good DCC. GCN2 didn't have any DCC. But, no overdraw = less BW bound.

On X1 couldn't you do some compute tasks from the esram (e.g. post process filters, motion blur), potentially while still running graphics workloads (perhaps shadow rendering to main memory?)

Yes. Shadow map rendering to main memory while doing compute in ESRAM works. Depth compression reduces BW cost.

 

[Blazkowicz]

Having a screen you can take on the go is great, but the Switch offers nowhere near the same mobility as a 3DS or a Vita. The old handhelds could fit into a pocket but the tablet certainly won't.

Laptops are popular although they hardly ever leave home for most users. Similarly a handheld no matter how big affords you to choose any couch or piece of furniture to sit on. There are past non-pocketable designs as well that were a big success, like the original DS and the original Game Boy - the latter one could be carried on vacations, on the bus and so on. If you can somehow bring water or something to drink where you go, or underpants, or something to write on then you can technically bring a handheld. Else we would often be thirsty, have smelly underwear or feet and be unable to attend school etc.

 

[Rodéric]

Laptops are popular although they hardly ever leave home for most users. Similarly a handheld no matter how big affords you to choose any couch or piece of furniture to sit on. There are past non-pocketable designs as well that were a big success, like the original DS and the original Game Boy - the latter one could be carried on vacations, on the bus and so on. If you can somehow bring water or something to drink where you go, or underpants, or something to write on then you can technically bring a handheld. Else we would often be thirsty, have smelly underwear or feet and be unable to attend school etc.

Don't know for other people but my tablet also stays at home and can be carried around and used anywhere in the house, making it so valuable. (That and ease of use/no boot time.)

 

[function]

Last gen made post processing, ambient occlusion, shadow maps and per pixel lighting must have features. Geometry rasterization (to g-buffer) generally takes only 30%-40% of the frame.

If you don't use compute for post processing, AO and lighting, you are instead rendering a fullscreen quads. Zero overdraw and no geometry bottleneck -> no advantage from Nvidia's tiled rasterizer or better geometry performance compared to AMD. However DCC helps a bit. Maxwell has very good DCC. GCN2 didn't have any DCC. But, no overdraw = less BW bound.

Thanks again. I hadn't considered the number of jobs for which the tiled rasterizer would offer no benefit, even when using the graphics pipeline.

AMD list the latest (GCN 1.2) DCC as providing (up to) "40% higher memory bandwidth efficiency". Assuming Nvidia are in the same ballpark as AMDs latest DCC, then with 25.6 GB/s (and assuming no embedded memory or added last level cache) then I think it's questionable how much value there would even be in running highly tuned X1 software above the NXs leaked 768 mHz.

Shield would presumably benefit from boosting to high clocks to make poorly tuned (low utilisation of both CPU and GPU, low bandwith consumption) software run better in benchmarks. NX software probably wouldn't see such easy gains. Just ... throttling or chocking on memory bottlenecks?

Don't know for other people but my tablet also stays at home and can be carried around and used anywhere in the house, making it so valuable. (That and ease of use/no boot time.)

You're talking about the toilet, aren't you.

 

[Pixel]

Power consumption of current gen consoles measured:

Xbox One S (slim) is ~60 watts in gameplay. A mobile device of this size has to consume less than 10 watts (including display). Even less, if we assume 4 hour battery life (100% gameplay). It would be physically impossible to design a device that has 6x+ higher perf/watt than a die shrinked 20 nm console AP. Thus nobody should expect Nintendo Switch to match Xbox One S in performance.

People seem to have way too high expectations towards mobile devices. Phone/tablet peak flop (race to sleep) marketing has confused everybody. Just hold your hand near the exhaust of a current gen console when you play, and you understand that no mobile device could ever generate that much heat and offer 4 hours of game play. And less heat means less performance, unless we have a reason to expect drastically improved perf/watt over AMDs 20 nm designs.

If Nintendo were willing to create something that weighed as much as a 7-10lbs laptop (depending on inclusion of hdd and bdrom ), with a large 30000mAh lithium battery that added $60 to the bill of materials, then something with the performance of Xbox One S would be feasible. It would weigh even less and get away with a smaller battery if it didn't have an hdd/bdrom and you stored your massive 15GB day one patches, dlc and 29GB Doom patches on removable flash devices.

 

[orangpelupa]

Thanks again. I hadn't considered the number of jobs for which the tiled rasterizer would offer no benefit, even when using the graphics pipeline.

AMD list the latest (GCN 1.2) DCC as providing (up to) "40% higher memory bandwidth efficiency". Assuming Nvidia are in the same ballpark as AMDs latest DCC, then with 25.6 GB/s (and assuming no embedded memory or added last level cache) then I think it's questionable how much value there would even be in running highly tuned X1 software above the NXs leaked 768 mHz.

Shield would presumably benefit from boosting to high clocks to make poorly tuned (low utilisation of both CPU and GPU, low bandwith consumption) software run better in benchmarks. NX software probably wouldn't see such easy gains. Just ... throttling or chocking on memory bottlenecks?



You're talking about the toilet, aren't you.

Playing a game in toilet is fun.

By the time you save the game, you already finished your business.

It's also a really nice pain relief for those horrendous morning with unbelievable stomachache

 

[Goodtwin]

I think your being unfair to the counterargument. I agree mostly with your assessment, but my criticisms, as I believe the rest in my corner are, are focussed at the larger business prospects and future of Nintendo as a brand and console player. As a handheld, a 3DS replacement, Switch is fine. As a means to grow Nintendo into a significant third player for the conventional core gamer, it's not. And that means less unit sales for Nintendo. The view as to what a console should be, which you consider narrow minded, is based on what sells to this market, as proven over 40 years of natural selection. Dual-stick TV based gaming with good graphics for the time is what the majority of the market wants. Ergo, to sell to that market you need to offer that experience. Deviation from that plan generally results in lower interest unless you hit a new audience, which typically is short lived because they are outside the normal hobby an have no long term interest in the new experience.

How many Switch (or Switches?) will Nintendo need to sell to maintain their current size? Will it grow the brand so next hardware, they have a stronger position for a new console? Or are they out of the running altogether in that regard? Does that mean they'll be bringing their games to the other consoles? A Japanese 1-2, Sony console with Nintendo game exclusivity, would probably be great for both parties.

It's in the context of that possible future, versus one where Nintendo becomes the dominant player as they used to be, that Switch is being criticised as underwhelming.

What is unfair about having the opinion that Nintendo can have great success without following the blueprints Sony and Microsoft follow? I personally see a me too console from Nintendo as just another player fighting over a market that is pretty well taken care of with the current players. Microsoft offers nearly an identical product to Sony's, and yet Sony is outselling them 2 to 1. I just think it makes more sense for Nintendo to play to their strengths, and offer a product to consumers that offers something the traditional consoles do not.

Nintendo may indeed need to reduce overhead at their hardware division. I don't expect the Switch to sell more than 80 million units. However, what the NES Classis has shown us is that Nintendo may want to consider releasing more novelty based hardware, as it seems to strike a chord with your non traditional gamer. I could see them doing a classic Gameboy next year. Strategically releasing these retro gaming products every Christmas season could move some units, and at the same time engage or reengage consumers with Nintendo's IP's. Reggie Fisime admitted that their mobile offerings and the NES Classis are really designed to get consume interacting with their IP's, with the hopes of more easily transitioning them to their dedicated gaming hardware. I'm also not so sure that the Switch couldn't have a TV box model that compete with the products like the Apple TV. Iwata said in the past that the NX could be one device or many, so I am sure Nintendo is considering possible models that are not hybrid, but strictly portable and strictly console. I don't think this is their planned direction currently, but is probably on the drawing board as a backup plan.

There is zero chance Nintendo can maintain its current size if they go third party. The entire hardware division would be closed down, laying off countless workers. So going third party to somehow preserve their size is a fallacy, a lot of people work at Nintendo to make hardware.

Only time will tell if Nintendo is making a good strategic move with the Switch. We are all entitled to our opinions, but in the end we just have to wait and see how things turn out.

 

[Shifty Geezer]

What is unfair about having the opinion that Nintendo can have great success without following the blueprints Sony and Microsoft follow?

Nothing. It's the attribution of the counterargument to narrowmindness and not just a different interpretation that's unfair.

I personally see a me too console from Nintendo as just another player fighting over a market that is pretty well taken care of with the current players.

I agree in part, although it'd be differentiated with Nintendo's exclusives. Of the current conventional market, what proportion of PS and XB buyers would buy a console that had their 3rd party favourites plus Nintendo's library over the rivals?

The major argument from me is that Nintendo's differentiation in this case is the wrong one and isn't going to work. It's not enough of a console to satisfy that market, making it basically a handheld that'll sell to the handheld audience which gets weaker YoY.

Nintendo may indeed need to reduce overhead at their hardware division. I don't expect the Switch to sell more than 80 million units.

I consider that very optimistic. This is 3DS sales...

http://www.statista.com/graphic/5/262074/worldwide-sales-of-the-nintendo-3ds-since-2004.jpg

Switch is going to have to be more desirable than 3DS.

However, what the NES Classis has shown us is that Nintendo may want to consider releasing more novelty based hardware, as it seems to strike a chord with your non traditional gamer. I could see them doing a classic Gameboy next year. Strategically releasing these retro gaming products every Christmas season could move some units, and at the same time engage or reengage consumers with Nintendo's IP's. Reggie Fisime admitted that their mobile offerings and the NES Classis are really designed to get consume interacting with their IP's, with the hopes of more easily transitioning them to their dedicated gaming hardware. I'm also not so sure that the Switch couldn't have a TV box model that compete with the products like the Apple TV. Iwata said in the past that the NX could be one device or many, so I am sure Nintendo is considering possible models that are not hybrid, but strictly portable and strictly console. I don't think this is their planned direction currently, but is probably on the drawing board as a backup plan

So the plan isn't to have a device people intrinsically want when revealed and advertised, but to resell Nintendo IP (again! How recycle can they get?!) and try to sell hardware based on games appearing on other devices? Do you think the conversion of mobile gamers playing Mario on their phone with one touch to buying hardware using a console interface is going to be high? I suppose they might attract back former Nintendo fans with that tactic.

There is zero chance Nintendo can maintain its current size if they go third party. The entire hardware division would be closed down, laying off countless workers.

Isn't it 280 workers?
https://en.wikipedia.org/wiki/Nintendo_Integrated_Research_&_Development

So going third party to somehow preserve their size is a fallacy, a lot of people work at Nintendo to make hardware.

Making hardware that doesn't sell in enough quantities means the closure of the whole of Nintendo eventually.

Only time will tell if Nintendo is making a good strategic move with the Switch. We are all entitled to our opinions, but in the end we just have to wait and see how things turn out.

Of course - why does that even need mentioning when that's the raison d'être of the board? :?

 

[sebbbi]

Thanks again. I hadn't considered the number of jobs for which the tiled rasterizer would offer no benefit, even when using the graphics pipeline.

AMD list the latest (GCN 1.2) DCC as providing (up to) "40% higher memory bandwidth efficiency". Assuming Nvidia are in the same ballpark as AMDs latest DCC, then with 25.6 GB/s (and assuming no embedded memory or added last level cache) then I think it's questionable how much value there would even be in running highly tuned X1 software above the NXs leaked 768 mHz.

Maxwell's tiled rasterizer reduces ROP bandwidth (overdraw to render target). DCC (Nvidia + newer AMD cards) reduces bandwidth of ROP writes and later read bandwidth of written render targets. Xbox One has ESRAM for the same purpose. ESRAM reduces main memory bandwidth usage of ROP writes (and reads afterwards) to zero (assuming RT resides in ESRAM). No matter how fancy DCC technique or how much overdraw memory bandwidth saved by tiling, these techniques cannot come close to a solution that completely nullifies the main memory bandwidth cost of most render targets. ESRAM also helps compute shaders. Down side is that ESRAM requires lots of programmer effort. Maxwell's bandwidth saving techniques are completely automatic.

According to this review, a single A57 core can consume around 5GB/s bandwidth:
http://www.anandtech.com/show/8718/the-samsung-galaxy-note-4-exynos-review/5

Anand said that multithreading with two A57 cores was enough to saturate the SOC bandwidth (13.2 GB/s in this case). Quad cores would thus be able to saturate the 25.6 GB/s bandwidth when doing memory heavy SIMD (NEON) processing. I am just wondering how much bandwidth Tegra X1 has left for the GPU if the games tax those 4 CPU cores to the max. Mobile software tends to be either GPU bound or CPU bound, rarely both at the same time (and usually bursty race-to-sleep style code). I am just wondering whether 25.6 GB/s is enough for console games (all CPU cores + GPU taxed at the same time). Both last gen consoles had more than 25.6 GB/s. Xbox 360 had 22.4 GB/s unified memory bandwidth + EDRAM and PS3 had 25.6 GB/s main memory + 22.4 GB/s graphics memory bandwidth. Fortunately Maxwell has modern memory bandwidth saving techniques + tiled rasterizer, meaning that Tegra X1 certainly beats last gen consoles in expected bandwidth. But I would expect bandwidth usage to be one of the things developers need to carefully optimize.

This is all of course assuming that Eurogamer article holds (Tegra X1 with 1600 MHz RAM):
http://www.eurogamer.net/articles/digitalfoundry-2016-nintendo-switch-spec-analysis

 

[Goodtwin]

I agree in part, although it'd be differentiated with Nintendo's exclusives. Of the current conventional market, what proportion of PS and XB buyers would buy a console that had their 3rd party favourites plus Nintendo's library over the rivals?

Impossible to tell. We have seen that the Gamecube, which did indeed have most third party titles sold very poorly. Meanwhile the Gameboy Advance was doing pretty well during this time period. This tells me consumers are more likely to purchase Nintendo hardware to play Nintendo when its not a traditional console.

I consider that very optimistic. This is 3DS sales...

http://www.statista.com/graphic/5/262074/worldwide-sales-of-the-nintendo-3ds-since-2004.jpg

Switch is going to have to be more desirable than 3DS.

Switch does have a chance to be more desirable than the 3DS. The unified Nintendo library is a huge plus. The Switch is far more capable, and is able to connect to the TV to play games. Perhaps some people will primarily play their Switch on the go, but some games like Zelda BoTW for example they may prefer to play that on their TV at home. Speaking of BoTW, that game alone is going to move units early on. Even with the Wii, Zelda TP was a big reason core Nintendo fans were so excited to buy a Wii.

Your sales link isn't complete. Its missing half of fiscal 2016, which ends in March. I'm pretty sure 3DS will outsell 2015 by a decent margin. 3DS could come close to selling as many units as the Gameboy Advance, not to shabby in the era of smart devices.

So the plan isn't to have a device people intrinsically want when revealed and advertised, but to resell Nintendo IP (again! How recycle can they get?!) and try to sell hardware based on games appearing on other devices? Do you think the conversion of mobile gamers playing Mario on their phone with one touch to buying hardware using a console interface is going to be high? I suppose they might attract back former Nintendo fans with that tactic.

Did you want a Bluray player for any other reason than to watch movies? There is a reason Disney used their IP's when creating an amusement park. Yes, Nintendo's IP's are a valuable asset when creating a desirable product. If Nintendo made games exclusively for the PlayStation, would it help drive sales? If so, then why cant you understand how those IP's will drive sales for Switch?

Isn't it 280 workers?
https://en.wikipedia.org/wiki/Nintendo_Integrated_Research_&_Development

If accurate then that's less than I would have expected, but at the same time means there is less overhead involved in making the hardware.

Of course - why does that even need mentioning when that's the raison d'être of the board? :?

I only mention it because its important to keep in mind that as strongly as you believe in your opinion, the only resolution to the disagreements on the issue is the test of time. We could go back and forth endlessly with counter arguments, examples of successes and failures, but we aren't likely going to change our opinions. Its not that I don't understand skepticism with what Nintendo is doing, there are no guarantees for success.

Nintendo can sell a ton of software even to a smaller userbase. The Wii U, by far the biggest console failure Nintendo has had (unless we count Virtual Boy), still sold a lot of Nintendo software.

https://www.nintendo.co.jp/ir/en/sales/software/wiiu.html

I think of the Switch as a portable first with the added perk of being able to play on the TV. Equally as important as having a good product is having a strong message, and marketing the Switch will be very important. By the end of 2017 I think we will know if the Switch is destined for success or not. Even the reveal in a couple weeks will be very telling of just how much software support will be there.

 

[AlNom]

Maybe they could add more GPU L2 ?

 

[Pixel]

Anand said that multithreading with two A57 cores was enough to saturate the SOC bandwidth (13.2 GB/s in this case). Quad cores would thus be able to saturate the 25.6 GB/s bandwidth when doing memory heavy SIMD (NEON) processing. I am just wondering how much bandwidth Tegra X1 has left for the GPU if the games tax those 4 CPU cores to the max. Mobile software tends to be either GPU bound or CPU bound, rarely both at the same time (and usually bursty race-to-sleep style code). I am just wondering whether 25.6 GB/s is enough for console games (all CPU cores + GPU taxed at the same time). Both last gen consoles had more than 25.6 GB/s. Xbox 360 had 22.4 GB/s unified memory bandwidth + EDRAM and PS3 had 25.6 GB/s main memory + 22.4 GB/s graphics memory bandwidth. Fortunately Maxwell has modern memory bandwidth saving techniques + tiled rasterizer, meaning that Tegra X1 certainly beats last gen consoles in expected bandwidth. But I would expect bandwidth usage to be one of the things developers need to carefully optimize.

A57 has similar cache sizes to Jaguar quadcore clusters
A57:
L1 : 48kB instruction/32kB data per core
L2: 2MB

Jaguar:
L1 32kB i/32KB d
L2: 2MB per quadcore cluster
So wouldn't the need to access main memory be some what similar to Jaguar?

Capcom released Resident Evil 5 for Android TV which is essentially a Switch hardware but more powerful (you know this but for other readers: Fully clocked Tegra X1 (2ghz quad A57, 1ghz gpu, 25.6GB/s bw)), and they had to reduce texture quality on some objects compared to the 360/PS3.