icecold1983
Member
running on unreal 4 isnt a guarantee of anything. they will have low quality art that can be cheaply produced and has to run on far less powerful hardware than an ipad pro
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Confirmed: The Nintendo Switch is powered by an Nvidia Tegra X1
- Thread starter FZZ
- Start date
- Status
running on unreal 4 isnt a guarantee of anything. they will have low quality art that can be cheaply produced and has to run on far less powerful hardware than an ipad pro
Thats not terribly great results over only a half hour of usage according to the tests. The larger form factor may help a little but its still passive cooling in a device mostly made up of battery with the same size SoCs as their phone siblings. It would be really interesting to see how the GPU handles being under load for 4+ hours.
icecold1983
Member
doesnt switch battery only last for 2.5 hours while playing zelda? not terribly great? the iphone 6s plus, comparable size to switch, barely throttles at all. it would run circles around switch. its not even a great comparison since iphone 7 is now out which only widens the gap
At max brightness and speakers turned up. Realisticly 3 hours.
I don't own one of these phones but maybe a GPU benchmark app could be setup to run on a loop to stress GPU performance.
Tegra X1 performs really well and I don't consider it a low end part just because it's not the newest SOC. Comparing to Apples modern SOCs is fun but not fair, Apple has never sold their designs to another company and Nvidia would have to use ARM A72 or A73 CPUs to approach Apples performance realm.
I have no clue if Nvidia has managed to improve Denver CPU performance in a competitive way, it has to go through an ISA translation to ARM that can have some performance penalties, so X2 will probably have a bigger benefit from higher clocks from the die shrink rather than architecture improvements.
The Switch is going to get games that will be more in depth than mobile phones simply because it's a gaming platform, it doesn't have to be the fastest device ever made; but it is competitive now and hopefully Nintendo updates with whatever Nvidia develops in the future.
EhrgeizVII
Member
An iPhone 6+ would run circles around Switch doing what? Making phone calls? Hmm you might have a point...
icecold1983
Member
doing anything that requires mathematical calculations. keep fighting that good fight tho
EhrgeizVII
Member
Well, good luck running Zelda on your phone dude. These comparisons are silly.
Chatting with your friends and setting up lobbies.
EhrgeizVII
Member
icecold1983
Member
comparing hardware is silly in a thread about the switch hardware? whats silly is your arbitrary, last defense zelda garbage
Lol we're already back to this shit.
Pixel C runs on Android, has shitty libraries and throttles to sub docked Switch (the Shield TV already does and there's no reason for a tablet sized device that starts with a lower clock not to do the same or worst). These benchmarks are useless in regards to gaming performances because they don't tax the hardware long enough to tell you how much the device would throttle after 1-2 hours of running an extremely taxing game, where the ideal scenario is to have clocks cranked up to the max to run a complex title for a long period. Sustained performances are crucial for gaming, that's why Switch has fixed clocks and active cooling. Running Angry Birds 2 and running Super Mario Odyssey isn't the same thing.
Finally, the A9X runs on a much more complex and taxing OS than Switch, but iOS+Metal is still a much better environment than Android+OpenGL or whatever Pixel was using for this benchmark so the TX1 chip started at a disadvantage. And by the way, good luck running Breath of the Wild with 15GB/s of bandwidth.
Tl;dr stop spreading nonsense based on bullshit non-evidence with this passion.
icecold1983
Member
wow you managed to be incorrect in every single thing you said wrt hardware with this post. well done
why would you compare to docked switch clocks btw?
And how am i going to answer to this incredible argument you made here lmao
EhrgeizVII
Member
What's silly is comparing an iPhone to a Switch because they do completely different things. An iPhone can't run the stuff a Switch does and vice-versa because phone/console, so how can it "run circles" around it?
icecold1983
Member
both have cpus and gpus that process mathematical calculations. one do so much faster than the other
blu
Wants the largest console games publisher to avoid Nintendo's platforms.
Just curious, do you consider throttling in your calculations?
GhostTrick
Banned
That's not what benchmark claims though. The iPhone 6s plus would be comparable to Switch, even slower considering it'd have to drive an 1080p screen.
Barely throttle at all ? Then what are the clocks ? What are the sustained clocks ? Do you have data to make that point. Running Zelda itself sounds like a silly argument, but behind that is a real argument: What is an iPhone 6s able to run without throttling ? Cause phone games sure arent a benchmark for that.
Hmmm, maybe, maybe not.
BotW is currently the Switch's most intense game. It runs on the Switch in portable mode for 3 to 3 and a half hours before the battery drains to empty from full charge. Of course the battery is simultaneously powering the fan to cool the Switch down.
Now the iPad Pro definitely has a more powerful chipset but may be subject to throttling so as not to get too hot. Would the iPad Pro be able to run BotW with sustained performance (like the Switch), MFi controller connected, only passive cooling, for 3 to 3 and a half hours or more?
TLZ
Banned
But they're both ARM tablets.
icecold1983
Member
did you read the link provided above? recent i phones dont throttle nearly enough to bring it close to switch performance
its rated at 1.84 ghz, which is damn close to what it sustains almost the entirety of that 30 minute stress test
also, keep in mind the same game will always have a lower utilization rate on a more powerful gpu, thereby making throttling much less likely to happen regardless
GhostTrick
Banned
They are indeed but one is passively cooled and the other is actively cooled. Which raise a question regarding GPU clocks and their sustainability.
The link above is about CPU clockspeed. What about the GPU clockspeed ?
xochipiltontli
Member
Because comparing the low but stable undocked Switch clocks to the scores of an iPad running a benchmark where it actually gets to use the full force of its GPU for 5 whole minutes makes just as much sense when trying to arbitrarily decide what can run what decently when taking into consideration battery life and heat dissipation.
GhostTrick
Banned
did you read the link provided above? recent i phones dont throttle nearly enough to bring it close to switch performance
its rated at 1.84 ghz, which is damn close to what it sustains almost the entirety of that 30 minute stress test
also, keep in mind the same game will always have a lower utilization rate on a more powerful gpu, thereby making throttling much less likely to happen regardless
I'll repeat myself: You're providing CPU clocks. The GPU on iPhone 6s must be like a lot bigger than the CPU. What makes more heat is the GPU. Which is more likely to throttle than your CPU, especizlly when running demanding titles.
Let's take iPhone 7's GPU for example:
http://www.pcworld.com/article/3146...-the-gpu-in-apples-iphone-7-are-unlocked.html
http://www.pcworld.com/article/3146...-the-gpu-in-apples-iphone-7-are-unlocked.html
Eh..yes iPhone 6+ runs circles around a Nintendo Switch and also does more. It just doesn't have Nintendo games and the JoyCons. Developers just don't make big AAA games for tablets and phones, because people wouldn't buy them for a high price.
Yes it would. The iPad Pro is a friggn beast and people really underestimate it. If we take it to 720p too. The native solution of the iPad Pro wouldn't be possible.
That doesn't explain why developers don't make decent looking $10-$15 games though like on PC and consoles.
Even Super Mario Run for example is nothing to write home about.
And then you wake up.
Try to read some of the posts here before doubling down on the nonsense, they don't write themselves.
The comparison between gaming computers and dedicated consoles didn't make much sense back in the 80s and 90s given the widely different input methods, software libraries and prices. It didn't make sense to compare them on the basis of "look, my computer can do all these things, what can your console do?" either.
Now, with the Switch vs. the iPad, it makes even less sense. Sure, on the surface, they're both tablets with mobile hardware in it, and they're both able to push 3D games. But that's like comparing a helicopter and a plane because they're both flying machines designed for transportation. Yeah, that's true, but you don't use them in the same situations. Helicopters aren't competing with planes because their respective capabilities make them suitable for very different jobs. Planes aren't competing with helicopters.
Now, with the Switch vs. the iPad, it makes even less sense. Sure, on the surface, they're both tablets with mobile hardware in it, and they're both able to push 3D games. But that's like comparing a helicopter and a plane because they're both flying machines designed for transportation. Yeah, that's true, but you don't use them in the same situations. Helicopters aren't competing with planes because their respective capabilities make them suitable for very different jobs. Planes aren't competing with helicopters.
Cue someone linking to a study that shows helicopter sales are down due to planes or something :lol
Very good analogy. Case closed.
Does anyone think the SOC will get a midlife die shrink? Does Nintendo do that? I know other consoles have.
blu
Wants the largest console games publisher to avoid Nintendo's platforms.
"..Nearly enough to bring it close to Switch performance.."?
Any throttling can be detrimental to latency-critical tasks. Guess the category high-fps games belong to. That plot of yours is an example what a gamedev does not want to see on a clock chart.
Skittzo0413
Member
did you read the link provided above? recent i phones dont throttle nearly enough to bring it close to switch performance
its rated at 1.84 ghz, which is damn close to what it sustains almost the entirety of that 30 minute stress test
also, keep in mind the same game will always have a lower utilization rate on a more powerful gpu, thereby making throttling much less likely to happen regardless
First off, that's the CPU frequency. And second, dips like that -even for a few seconds- cannot happen in a game console. That would completely ruin your FPS among other logic related things. A game console cannot throttle at all- the processor speeds must remain consistent, at least throughout the same areas.
Here is the AMD A8 7600 (550gflops) playing battlefield 1 in DX11 at 720p and medium settings. https://www.youtube.com/watch?v=hdZN-JuUbbs
I don't think it's unreasonable to see how Switch can handle these games given better APIs, dialing in graphics a bit more, and lower CPU overhead (lower performance as well obviously) thanks to those APIs. Since this APU is limited in memory bandwidth, running on DDR3 it makes a good candidate to learn how the game would perform for Switch, especially if the game is running DX11 where Nvidia does enjoy a nice performance per flop advantage over AMD.
I am not trying to say that it would be 1:1 with this APU, but it would be pretty close, as for the handheld, you'd have to reduce resolution, but on a screen that small, I don't think there would be much issue with the image quality, and there is no resolution scaling in the video which could be another place where Switch plays catch up. My point was to show that BF1 is something that could have been ported to the Switch if the timing was right and EA thought it would sell.
A8 7600 other games from this user (Again, this isn't a 1:1 but it is a lot closer than blindly saying "Switch can't run this" the real issue here isn't the gpu side but the cpu side, dx11 has a lot of cpu overhead that I think will help but this APU is much faster on the cpu side, so while the games can look like this, some games might need reduction to game logic, though it is unknown how much, single threaded performance is good enough to meet ps4 and the api should free up cpu cycles):
Overwatch: https://youtu.be/8gITummnxuM?t=10m49s 720p 26-38 higher settings
The Division: https://www.youtube.com/watch?v=P8yQPNsDFGs 720p 29-43 low settings
Titanfall 2: https://www.youtube.com/watch?v=hnq8qrOVEvQ 720p 28-35 mixed low/high settings
For Honor: https://www.youtube.com/watch?v=JvRFpHi7moY 720p 32-38 mixed low/medium settings
There are other games on his channel you can check out, this was just a quick run down of what a gpu similar to Switch's (when docked) is capable of, obviously when docked, lower settings and resolution/scaling will need to occur, but it's an interesting look, especially the memory bandwidth here.
I don't think it's unreasonable to see how Switch can handle these games given better APIs, dialing in graphics a bit more, and lower CPU overhead (lower performance as well obviously) thanks to those APIs. Since this APU is limited in memory bandwidth, running on DDR3 it makes a good candidate to learn how the game would perform for Switch, especially if the game is running DX11 where Nvidia does enjoy a nice performance per flop advantage over AMD.
I am not trying to say that it would be 1:1 with this APU, but it would be pretty close, as for the handheld, you'd have to reduce resolution, but on a screen that small, I don't think there would be much issue with the image quality, and there is no resolution scaling in the video which could be another place where Switch plays catch up. My point was to show that BF1 is something that could have been ported to the Switch if the timing was right and EA thought it would sell.
A8 7600 other games from this user (Again, this isn't a 1:1 but it is a lot closer than blindly saying "Switch can't run this" the real issue here isn't the gpu side but the cpu side, dx11 has a lot of cpu overhead that I think will help but this APU is much faster on the cpu side, so while the games can look like this, some games might need reduction to game logic, though it is unknown how much, single threaded performance is good enough to meet ps4 and the api should free up cpu cycles):
Overwatch: https://youtu.be/8gITummnxuM?t=10m49s 720p 26-38 higher settings
The Division: https://www.youtube.com/watch?v=P8yQPNsDFGs 720p 29-43 low settings
Titanfall 2: https://www.youtube.com/watch?v=hnq8qrOVEvQ 720p 28-35 mixed low/high settings
For Honor: https://www.youtube.com/watch?v=JvRFpHi7moY 720p 32-38 mixed low/medium settings
There are other games on his channel you can check out, this was just a quick run down of what a gpu similar to Switch's (when docked) is capable of, obviously when docked, lower settings and resolution/scaling will need to occur, but it's an interesting look, especially the memory bandwidth here.
cleveridea
Member
Ouch if double performance and half the power draw if true. I'd rather have waited till later in the year and bought a switch with x2.
http://seekingalpha.com/article/4057051-nvidia-happened-san-francisco-knew
http://seekingalpha.com/article/4057051-nvidia-happened-san-francisco-knew
It's not https://developer.nvidia.com/embedded/buy/jetson-tx2
The GPU runs at 1122mhz or 575gflops (15watts) vs X1's 1ghz or 512gflops, now there is a max clock of 1.3ghz for the GPU, which would push it to 665gflops but this is well beyond that power consumption.
Nintendo would have obviously benefited from moving X1 to 16nm too, but it doesn't make it a new device, it would be a small bump that would likely still need to be taken back to 1ghz to get it not to throttle in the Switch form factor.
http://www.anandtech.com/show/11185/nvidia-announces-jetson-tx2-parker
cleveridea
Member
Interesting, though it does say on anandtech "up to double the performance" but who knows what regular situation
I can imagine Nintendo balking at the price though, $399 in the article
From DF:
Undocked
cpu: 1020MHz
GPU: 307.2MHz/384MHz
Memory: 1331.2MHz
Docked
CPU: 1020MHz
GPU: 307.2MHz/384MHz/768MHz
Memory: 1331.2MHz/1600MHz
There could have been minor tweaking after this leak as we dont know exactly when its from.
The day after their leak of the X1 devkit stuff they did a sit down and discussed what an x2 chip could mean for switch but never released it because their developer contacts started to say it wasn't x2 and that is likely when they got their clocks, so September or October, whichever that video was published.
They heard about the boost to the handheld's gpu and we should all go by the current clocks we have from DF, but it isn't confirmed that the other clocks haven't changed as well, though I think it's 100% safe to assume that the Foxconn clocks are out of the question with a 20nm chip unchanged from Tegra x1 at least visually.
E-Cat
Member
Its based on disproven information.
E-Cat
Member
I see.
Zelda looks amazing and that's a Wii U title ported over. You can bet their next Zelda will look better too. Plus look at previous gen games like uc3, last of us, Forza, halo 4, etc and the switch can handle those games better. It's enough to make some great looking games especially Nintendo ips
E-Cat
Member
I agree that it's incredible what they've accomplished with such little power.
My guess is that in a couple of years, a revision will appear at the same price point $300 with a Tx2. And that is when Nvidia announces the Tx3
- Status