• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Ventura Beat: Nintendo Switch graphics are based on Nvidia's Maxwell Architecture

I knew it, Switch will cost 249/299$ and will release in March, Tegra Pascal architechture is too recent to be in there, they wouldn't have had the time to make a customized Pascal chip for Switch, we already know Nvidia has been working for almost 2 years on Switch so it couldn't be Pascal since it's too recent, and also as i said implementing a newer technology would mean higher costs and it doesn't look like Switch will be expensive.
That said i'm fine with it, i never thought Switch could reach XB1 performances even when running at full clock speed in the dock, and i'm fine with that, i want it for Nintendo games and its exclusives, i have my PC and Xbox for those multiplatforms whoch won't be able to run on Switch

...except we've had several posts from a very reliable insider like Matt stating that yes, Switch can get ports from PS4 and Xbox One on a technical level? It seems this time the real matter will be if games generate enough ROI to be considered a profitable venue in the long run.
 
You guys were seriously expecting Pascal? On a Nintendo console?
Yeah, everybody here has been saying:"it has Pascal" "Pascal is cheaper than Maxwell at this point" etc.
I knee it was impossible, Pascal Tegra chip is so new, and when you plan a console you do that years before, you can't just make a chip in a few months, so Maxwell was the only way, also there where already rumors from Emily and other sources that development kits had the X1 chip, so i guess final Switch Tegra will be a custom X1, with the main customization being the ability to scale it's clock depending on the situation.
My guess is Switch will be about 500GF in handheld mode and 600GF docked
 
We already had an idea of what Switch can do, we have seen Zelda working in real time in TV and handheld mode at a really more stable framerate than the Wii U version, so I can't care less about what's the hardware, sure it does it's job.

Damn, even the few seconds of the new Mario in the reveal trailer looked awesome, can't wait for January event.
 
...except we've had several posts from a very reliable insider like Matt stating that yes, Switch can get ports from PS4 and Xbox One on a technical level? It seems this time the real matter will be if games generate enough ROI to be considered a profitable venue in the long run.

Not saying it won't have multiplatforms, it will and it can since the GPU are similar and they can downgrade the games enough to run on Switch, though besides the profits reason ther could be cases where Switch won't get some multiplatforms since it'd had to be scaled down to much to be worth it, for example a game like Battlefield 1 already run at 720p on XB1 and 900p on PS4, a Switch version wouldn't be possible without heavily downgrading it at a way sub HD resolution
 
No matter how Nintendo tries to market it, no matter how some other gamers may try to spin it, the Switch for me is primarily a 3DS successor -- not a Wii U successor -- even though it's essentially both. It's a super-powered handheld with great TV out functionality. In that context, it's a beast no matter how you slice it. So rather it uses Pascal or Maxwell, it's going to be one hell of a bump and I can't wait to see what devs pump out on this thing.
 
Yeah, everybody here has been saying:"it has Pascal" "Pascal is cheaper than Maxwell at this point" etc.
I knee it was impossible, Pascal Tegra chip is so new, and when you plan a console you do that years before, you can't just make a chip in a few months, so Maxwell was the only way, also there where already rumors from Emily and other sources that development kits had the X1 chip, so i guess final Switch Tegra will be a custom X1, with the main customization being the ability to scale it's clock depending on the situation.
My guess is Switch will be about 500GF in handheld mode and 600GF docked

The thing is, Pascal isn't a significantly different architecture to Maxwell. The original roadmap was Maxwell and then Volta but because the Maxwell line of GPUs did so well in sales, they suddenly added Pascal into the roadmap which added some new features to the microarchitecture and had a 16nm node.

So it was reasonable to think it would use Pascal because it wasn't something that required years to develop the microarchitecture and that they haven't die shrunk old architectures and re-released them for a while.

Then there's the fact that 20nm nodes are a fab process that is rarely used for TSMC in manufacturing and will cost more for Nintendo to manufacture an SoC out of compared to 28nm or 16nmFF.
 
Isn't the biggest argument for pascal that it's smaller - while maxwell is 20/28nm - which will become more costly down the road as the current chipsets are all moving to 16nm?

And afaik the devkits which are desktop machines could still be maxwell based while the switch is pascal since they are essentially the same, but smaller? As long as they use the same instructionset, functions and use the same clockspead?

Pascal isn't faster as much as it is smaller, cheaper (especially down the road) and gives better battery life
 
Even if this were true, isn't desktop Maxwell better than desktop Pascal in overclocking? If that translates to the Switch, then docked performance may be higher.

Not that it's likely to be true; the original article was so poorly written and misinformed that I didn't bother to keep the tab open. I regret that now that they've stealth edited it. Anyone have the original embarrassment saved?
 
here it is. We can't take the site at face value now.

You can use that to discredit this author but at the time all the people blasting this article were running around NeoGAF talking about Polaris being in the switch. Seems to me most just believe whatever rumor is best at he moment.

Also wasn't the original Nvidia rumor that there were tons of unsold shield chips and Nvidia gave Nintendo a good deal. How or why Nvidia has a ton of unsold Pascal based chips lying around?

But it is nice that now that this rumor is out there some are acknowledging that Maxwell to Pascal is simply an efficiency gain.
 
Gonna quote this for the new page. I need to get to bed.

I have a few immediate thoughts after reading through the article:


  • Firstly, it's worth noting the difference between Maxwell and Pascal is almost entirely down to the manufacturing process. Maxwell was made on 28nm (and in the case of the TX1, 20nm) whereas Pascal is made on 16nm. The actual architectural difference between the two is minimal, and aside from improved color buffer compression, largely irrelevant for a device like the Switch.
  • Despite that, the article never makes any mention of the manufacturing process. I find that extremely strange, as it's obviously the defining difference between the two sets of GPUs.
  • In fact, the article gets the difference between the two completely the wrong way around, saying "Nintendo’s box is relatively small, and so it has to fit into the heat profile of a portable device, rather than a set-top box. That’s another reason that explains the older Maxwell technology, as opposed to the Pascal’s state-of-the-art tech." Pascal is literally a more power efficient version of Maxwell, so the incentive would be the other way around.
  • The author says "we expect the Nintendo Switch to be more than 1 teraflop in performance", which is notably higher than even those of us who were expecting Pascal were considering (I literally posted earlier today with a 500-750 Gflop estimate). If this is a Maxwell chip, then that would mean at least 4 SMs (512 "CUDA cores") at 1GHz, as they're not going to be able to push much past that on 28/20nm. This is a much larger GPU than most people would have been expecting.
I see a few different scenarios here:


  1. The Switch SoC uses Maxwell at 20nm, and simply has a much larger GPU than anticipated to account for the performance.
  2. Nintendo looked at the feature-set planned for Pascal when design started, realised that the new features were largely irrelevant, and decided that they would save time and just use a straight-forward die shrink of Maxwell to 16nm instead. That would technically be a Maxwell GPU, but would be almost completely indistinguishable from Pascal in terms of performance.
  3. The sources are wrong about Maxwell, the 1 Tflop performance, or both.
Basically, if you're to take the article as being accurate, then the only worthwhile takeaway is this quote:



A Maxwell Tflop is identical to a Pascal Tflop, and it's largely irrelevant to us whether they achieved that by using a larger Maxwell GPU on 20nm/28nm at a lower clock or a smaller Pascal GPU on 16nm at a higher clock.
 
so are we approaching semantics now? Whats the practical difference between a maxwell @16nm and a pascal @16nm?

I suppose what matters is... *is* it at 16nm, or is it a fat old 28nm chip? Thats the big question for me. I thought there were issues getting older processes die shrunk and it'd actually be cheaper to take the newer chip - aren't Xbox one S and PS4 Pro using a later architecture for their shrinks?

X1 is 20nm, so at worst it is that, but because of the manufacturing of that process winding down so drastically, and being replaced by 16/14nm process while next year will see 10nm product lines, maxwell would end up being more expensive, hotter, bigger, and slower. However the difference to performance is at the high end, 100gflops and we can still expect 500gflops+ which is where it has been all along. They treat this news like we were expecting xaiver, which is a new architecture based on Volta, that can produce 2TFLOPs but that is probably for "NEW" switch in 2018 or whatever they end up doing.
 
Everyone who has searched their heart while contemplating the idea of Nintendo doing a pascal handheld will already know the truth
 
This would be really disappointing, hope its a false report. Hypelevel sinks



The difference would be marginal. I'm always baffled on how people think, for better or worse, that Pascal would be a game changer. It would make a more efficient machine, not a different one.
 
X1 is 20nm, so at worst it is that, but because of the manufacturing of that process winding down so drastically, and being replaced by 16/14nm process while next year will see 10nm product lines, maxwell would end up being more expensive, hotter, bigger, and slower.

Because of this (I still can't believe they will use 20nm) I think a Switch revision will pop up pretty quickly down the line.
 
Completely expected, not at all suprised if true. This is Nintendo we are talking about here.

Once again people convinced themselves to be way too optimistic.
 
The difference would be marginal. I'm always baffled on how people think, for better or worse, that Pascal would be a game changer. It would make a more efficient machine, not a different one.

Well less heat and more battery life at the same power level or more power with the same battery life, is not "marginal" imo. I like new tech :)
Still going to buy the switch^^
 
Well less heat and more battery life at the same power level or more power with the same battery life, is not "marginal" imo. I like new tech :)
Still going to buy the switch^^


That's the same kind of improvement from Xbox One to Xbox One S. This won't make it a different machine is what I mean.


Not really unrelevant for a handheld.


True, but my point is you'd make it a better machine, not a different one.
 
I honestly am surprised anyone thought that Nintendo would not cut costs anywhere they can.
Of course, they are cost cutting. Every dollar they can save on one part, they can either buy better parts for an different element or make a cheaper product. What a strange criticism ...
 
After two generations of console from Nintendo that just weren't from me I didn't like what I saw of this in the announcement video as I really wanted them to go back to the less idiosyncratic console design of when tNintendo were at their best.

I know the setup isn't the same thing but having a screen in the default controller of the Wii U ruined it for me so I found a similar setup in the Switch to be pretty off-putting, partly because doing that increased the cost by adding something that was absolutely of no use to me, again.

Although to a certain extent pure processing power in a console is an enabler in relation to games, if this rumour is true, I don't necessarily see it as a bad thing for me due to the fact that it pretty much confirms that it's going to be a sub £200 console which really is the maximum I would have paid for this, even if it was Pascal based.

I'm actual more positive about the console now although the downside for me was that when they showed the Switch version of Zelda the image quality was lacking somewhat in relation to aliasing.

That's far more understandable now and although the image quality did suffer, this has gone from a £250-£300 console that I really had no interest in to a hopefully much cheaper console that I'm actually looking forward to. I'm sure that I'm in the minority here but if it;s true, the price will be a massive plus for me.

Make it £200 or £220 with one of these controllers and I'll be at the front of the queue on launch day.
nintendo-switch-controller-100688607-large.png
 
I honestly am surprised anyone thought that Nintendo would not cut costs anywhere they can.

The cost cutting is debatable in the long run, unless Nvidia offered them a really great deal.

And even the article states that it wasn't practically Nintendo's choice, but the fact that Pascal Tegra wasn't available on time.
 
After two generations of console from Nintendo that just weren't from me I didn't like what I saw of this in the announcement video as I really wanted them to go back to the less idiosyncratic console design of when tNintendo were at their best.

I know the setup isn't the same thing but having a screen in the default controller of the Wii U ruined it for me so I found a similar setup in the Switch to be pretty off-putting, partly because doing that increased the cost by adding something that was absolutely of no use to me, again.

Although to a certain extent pure processing power in a console is an enabler in relation to games, if this rumour is true, I don't necessarily see it as a bad thing for me due to the fact that it pretty much confirms that it's going to be a sub £200 console which really is the maximum I would have paid for this, even if it was Pascal based.

I'm actual more positive about the console now although the downside for me was that when they showed the Switch version of Zelda the image quality was lacking somewhat in relation to aliasing.

That's far more understandable now and although the image quality did suffer, this has gone from a £250-£300 console that I really had no interest in to a hopefully much cheaper console that I'm actually looking forward to. I'm sure that I'm in the minority here but if it;s true, the price will be a massive plus for me.

Make it £200 or £220 with one of these controllers and I'll be at the front of the queue on launch day.
nintendo-switch-controller-100688607-large.png

The leaks were always 200-250$, not 200-250£
 
This article makes no sense.

It is not like Tegra Pascal was already launched and it is smaller, cooler and draw less power than Tegra Maxwell.

Gonna quote this for the new page. I need to get to bed.
I'm sure these 1 TFLOPS is FP16 half-precision... so about 500GFLOPS in FP32.
 
It'd be pretty funny if whatever the next Shield TV they reveal at CES in January ends up being more powerful than the Switch.

(This is regarding the claims from the article in the OP that Pascal architecture for Tegra wasn't ready but we will most likely have Nintendo Switch and the next Shield TV launching in 2017.)
 
I have a few immediate thoughts after reading through the article:


  • Firstly, it's worth noting the difference between Maxwell and Pascal is almost entirely down to the manufacturing process. Maxwell was made on 28nm (and in the case of the TX1, 20nm) whereas Pascal is made on 16nm. The actual architectural difference between the two is minimal, and aside from improved color buffer compression, largely irrelevant for a device like the Switch.
  • Despite that, the article never makes any mention of the manufacturing process. I find that extremely strange, as it's obviously the defining difference between the two sets of GPUs.
  • In fact, the article gets the difference between the two completely the wrong way around, saying "Nintendo’s box is relatively small, and so it has to fit into the heat profile of a portable device, rather than a set-top box. That’s another reason that explains the older Maxwell technology, as opposed to the Pascal’s state-of-the-art tech." Pascal is literally a more power efficient version of Maxwell, so the incentive would be the other way around.
  • The author says "we expect the Nintendo Switch to be more than 1 teraflop in performance", which is notably higher than even those of us who were expecting Pascal were considering (I literally posted earlier today with a 500-750 Gflop estimate). If this is a Maxwell chip, then that would mean at least 4 SMs (512 "CUDA cores") at 1GHz, as they're not going to be able to push much past that on 28/20nm. This is a much larger GPU than most people would have been expecting.
I see a few different scenarios here:


  1. The Switch SoC uses Maxwell at 20nm, and simply has a much larger GPU than anticipated to account for the performance.
  2. Nintendo looked at the feature-set planned for Pascal when design started, realised that the new features were largely irrelevant, and decided that they would save time and just use a straight-forward die shrink of Maxwell to 16nm instead. That would technically be a Maxwell GPU, but would be almost completely indistinguishable from Pascal in terms of performance.
  3. The sources are wrong about Maxwell, the 1 Tflop performance, or both.
Basically, if you're to take the article as being accurate, then the only worthwhile takeaway is this quote:



A Maxwell Tflop is identical to a Pascal Tflop, and it's largely irrelevant to us whether they achieved that by using a larger Maxwell GPU on 20nm/28nm at a lower clock or a smaller Pascal GPU on 16nm at a higher clock.

This right here should end the thread

Maxwell? So $199 (without power cord) confirmed?

I see this thing as a massive upgrade to the 3DS, if the price is right, this could be a hit.

Not surprising Nintendo isn't pushing for power any more, that much is obvious if you look at the last 3 generations of their hardware.

No one expected them to push power. And again, Pascal is not a huge jump over Maxwell in terms of power, but in terms of efficiency. Going with Pascal over Maxwell would mean more battery and less heat for the same power as Maxwell. Think it would be cheaper for Nintendo to go Pascal as well, which is why this article makes no sense.
 
Maxwell? So $199 (without power cord) confirmed?

I see this thing as a massive upgrade to the 3DS, if the price is right, this could be a hit.

Not surprising Nintendo isn't pushing for power any more, that much is obvious if you look at the last 3 generations of their hardware.
 
Pascal Tegra would never be ready for a mass production device launched in march 2017.

BTW, NVidia was said to be planning to release a Tegra X1 Shield Tablet for late 2016 but canned the plans because that's basically what the Switch is.

The portable X1 was only used in one device, the Pixel C.
 
Must have been said already, but still :

NVIDIA said:
Nintendo Switch is powered by the performance of the custom Tegra processor. The high-efficiency scalable processor includes an NVIDIA GPU based on the same architecture as the world’s top-performing GeForce gaming graphics cards.

Still sounds like Pascal to me.
 
You guys were seriously expecting Pascal? On a Nintendo console?

Months ago NateDrake told us it was so we believed it was. I'll still go with that over VentureBeat. Besides, less than a month left so we'll know very soon.
 
I actuall think a design based off Tegra X1 makes more sense with its CPU configuration (4xA57 + 4xA53 rather than the Denver cores), a die shrink to 16nm would make a lot of sense nonetheless.
 
Damn, some posters are talking about Maxwell as if it's the GeForce FX architecture or something. I wonder if they even know what it is.

(Answer: it's the same arch as the GF 960/970/980 line).
 
Damn, some posters are talking about Maxwell as if it's the GeForce FX architecture or something. I wonder if they even know what it is.

(Answer: it's the same arch as the GF 960/970/980 line).

Masses commenting on/judging stuff they have the faintest idea about? It's the norm.
 
Must have been said already, but still :



Still sounds like Pascal to me.

Does sound like it, but people in here already said Maxwell and Pascal are not that different architecture wise. You could also question if it is Pascal, why not just state it? (because Nintendo ninjas, but still)
 
Wow, this thread is a cesspool.

Just to purposely make a dig into many of the posters here.

Welcome to the world of video games where hardcore enthusiast gamers never get in touch with the tech side of the gaming industry they're so interested in and constantly spout off what's best without an inkling of knowledge or understanding and unwilling to ask questions.

Anyway, I'm off to bed. Night. Peace!

Note for others: Maxwell and Pascal are not all that different from each other. Pascal is more of a performance model of Maxwell, and yes, it clearly would benefit Nintendo, but Maxwell at 16nm would also do a good job as well since it's just a step away from Pascal. And it being a custom chip, I full expect it to be this weird in-between like some custom chips typically are.
 
Top Bottom