IronicSonic
Member
That would be an incredible good scenario. Skyrim in the debut trailer could point in that direction
-
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.
Nintendo Switch: Powered by Custom Nvidia Tegra Chip (Official)
- Thread starter ekim
- Start date
That would be an incredible good scenario. Skyrim in the debut trailer could point in that direction
Calm Mind
Member
Don't mock Ace! He meant well.
Candescence
Member
Around 75-80% is probably a fairly reasonable estimate. Any estimate that allows current-gen third party games to be playable at decent framerates at 720p is the reasonable ballpark, really, that's obviously the objective with this thing.
Sometimes I can still hear his voice... RIP Ace.
Ok, i was just asking to get a general idea about how big the boost/advantage could be by having double precision in a gaming context. If Blu says anywhere between 25 and 50% of computations could benefit from it, let's just say, that the 500 GF chip, might punch above its weight and deliver between 600 and 750 GF depending on circumstances.
I'm also assuming the advantage of going with the Pascal successor, will be entirely focused on power efficiency and not performance.
I think providing a single number is really hard. E.g. (and these are just examples to illustrate the point!) it could feasibly provide 90% or more of the CPU performance, but only 50% of the memory bandwidth (but also require less bandwidth in some scenarios due to the different rasterization scheme employed by NV since Maxwell). That would make some games straightforward to port and some really hard to port based on their individual workload profiles / major bottlenecks.
Overall, I think a limiter for some next-gen ports could be that they need to run in both docked and portable mode, and I'm still uncertain about the frequencies/performance we'll see in the latter.
Of course, the biggest hurdle to widespread porting of third party titles might not be technical at all.
Skittzo0413
Member
There is also that Nvidia Flops vs AMD Flops advantage (on PCs at least) which could wind up making an exact Flop comparison kind of cloudy. I think ~500 GFlops is a good estimate at this point considering we know the July devkits were using TX1s with active cooling, and the final device has multiple vents and reports of active cooling.
We actually have a pretty clear picture of what's going to be in this thing. Obviously the final SoC will be custom (as per the OP) but having the devkit use an off the shelf TX1 does tell us quite a bit.
Schnozberry
Member
Pascal will allow higher clocks at lower power draw than Maxwell, so you can get some of the best of both worlds, especially in docked mode. It would be even better for efficiency if they went with ARM A72 cores, but that seems pretty pie in the sky. Nvidia stuck with A57 for their Parker Design, choosing to focus on adding their Denver Cores instead.
The fact that different games will have different considerations to downport to the NX because of certain differences in architecture like you mentioned is already concerning for third party support. If the system has a beast CPU and GPU but only 4 gigs of ram then no matter what you do it won't be able to run big open world titles. And third parties might just go "not worth it" and not optimize their games for this architecture. So they need to match current gen as closely as possible with their hardware profile.
The thing is there are two different ways of approaching that question.
The first, and the one that I was answering with my original post, is "If you take an existing game and change 1/3rd of the shader workload to FP16 on Switch, what performance increase will you get?" The answer to that is a 20% increase, not a 33% increase, as I explained.
The second possible way to approach the question is "If a developer chooses to heavily optimise a game for Switch, where 1/3rd of the existing workload could run at FP16, what performance increase would you get?" This is a much more difficult question to answer, as a developer in this instance would favour graphical techniques and effects which can be largely or wholly performed in FP16 over those which require FP32. So, if Ubisoft is porting, say, the next Assassin's Creed to Switch, and they have to decide which effects to disable to get the game running smoothly on the hardware, they're going to be more likely to disable the effects which have a large FP32 workload than the effects which can be run at FP16. So the Switch version of the game is going to have a workload that has a larger proportion of FP16-safe code than the PS4 and XBO versions, simply because it allows them to get the best balance of performance and visuals on the platform.
blu
Wants the largest console games publisher to avoid Nintendo's platforms.
Right, my bad - I inverted the problem. A 660GF worth of fp32 would take a 1:1 split to get the same time in fp32 + twice-faster fp16.
Schnozberry
Member
Of the versions of the Tegra X1 that made it into production, only the Shield Android TV device had an active cooler, and it consumed up to 20W of power. The thing is, the cooler on the Shield TV was basically silent, even under full load. The device is also extremely small, not much thicker than than the Nintendo Switch is pictured to be on the end of the device where the vents are.
So for the Switch dev kits to be running noisy active coolers, it would either need to be a really shitty thermal setup, or the Tegra X1 was being pushed past it's normal limits to approximate the final hardware.
Skittzo0413
Member
Yeah, I should have said ~500 GFlops is a good low end estimate. As in, even in portable mode (if such a thing exists) 500 is likely going to be the floor. Especially considering the mounting evidence that the final chip will be Pascal based.
This is something that really doesn't have any solid sources behind, 512 gflops is X1 at 1Ghz in Shield TV which does have really high TDP, I hardly doubt Switch can work at that speed as a handheld device.
Now, of course there are some things in the air like the possibility of a more efficient fab process (since we all know 20mm was quite bad and would be quite dumb and ankward to use it) and logic suggest us that at the very least Switch Tegra chip should be made with a more modern 16nm finfet which has shown amazing results in the GeForce Pascal desktop lineup.
But aside from rumours telling us that Switch Tegra chip is based in Pascal architecture, we can't conclude that is going to work at such speeds, we even don't have a really solid source assuring as that "Docked mode" is a thing, just a rumour that "it may be" (and while I would hope so, since it seems like a such a simple thing to do we really don't know for sure).
So I would advise against hoping for 500-700 docked Gflops for the time being, I for one I'm putting it in the 350ish handheld mode with no docked mode for the time being. That would be around 3 times more powerful than wiiu (not in raw Gflops, but with a much better cpu, and an much much modern gpu architecture should put 350 Nvidia Gflops in that ballpark)
To expand further what I expect from what we know, I expect something very close to X1 at 16nm finfet with moderate clocks to enable a reasonable battery life and with some extra work done in the memory subsystem, if Nintendo has done something right since Gamecube times with their hardware is taking care of avoiding memory bottlenecks in this regard.
Xdrive05
Member
So then based on what we know right now, is it reasonable to guess that Switch performance will be comfortably ahead of 360/PS3/Wii-U, but clearly not in the XBO/PS4 camp? Meaning it can handle perfect ports of last gen, but don't expect very good ports of current gen games? Just based on what we know right now. Is that a fair and informed speculation?
I think this is a sane way to look at this thing, yes. Wii-U games on the go is going to be already such a ridiculous step up from what current handhelds can do.
TheDinoman
Member
Pretty sure Emily Rogers and the like have already confirmed that it's stronger than the Wii U but weaker than the PS4/Xbone.
I don't think A72 cores are that unlikely. They're smaller (i.e. cheaper) than A57s, they consume less power and give higher performance. The only reason it would be using A57s is if the chip design was finalised so early that switching to A72s would be impossible, and given the fact that devices with A72 cores were in customers hands about a year and a half before the Switch launch, I would find that very unusual.
Regarding Parker, I strongly suspect that it was in fact taped out very early, and they were simply waiting for 16FF to become viable. They had been publicly talking about the chip's configuration for a long time (before the A72 was even announced, afaik), and as soon as its availability is officially announced we find out that they're well under way with design work on its successor, Xavier.
Yeah, it's a bit counter-intuitive. Of course this all assumes that you're simply taking an existing game and changing a certain proportion of the workload to FP16, whereas in the process of optimising a game for the platform the percentage of the workload which can be run at FP16 is more likely to grow than to shrink (as in my reply to ozfunghi above).
This wouldn't explain the active cooling.
Skittzo0413
Member
As I understand it, this will be a lot closer to XB1 in real world performance than Wii U, while supporting a lot of middleware and development tools. The CPU and some of the other features will likely be ahead of XB1/PS4, but the raw GPU power will be below.
So it should be able to get XB1/PS4 ports fairly easily, though expect a slightly lower resolution (especially for handheld mode as it has a 720p screen) and likely fewer/less intensive effects, lower texture resolution, etc.
Gladly taking the risk.
DragonKnight
Member
Agreed.
But do we know for sure it has active cooling? I mean, we know that development kits have active cooling, but we really don't know if the Switch has it. Yes, we can see vents in the unit, that may be an indicator of active cooling, but we really don't know.
It may just be vents to allow for passive cooling for all we know, so till we get more official/hard info we really don't know, and active cooling / docked mode is just speculation at this point.
Hey, I'm the first one to acknowledge that "docked mode" makes so much sense, but I think that is more reasonable to not expect it to avoid any kind of disappointment.
Living the Dream
Member
Would be neat if there was a "switch" on the device to run it in docked or handheld mode when undocked. That way if you didn't care about battery power it could run at full potention when undocked if you want it to.
Skittzo0413
Member
Emily Rogers (who got essentially everything about the Switch right so far) reported that the final unit will include active cooling.
Huh? 4gb of RAM is definitely enough to run open world games.
Nightbringer
Don´t hit me for my bad english plase
The key is simple:
More Cores but less clockspeed is better in a handheld device than less cores and more clockspeed.
More Cores but less clockspeed is better in a handheld device than less cores and more clockspeed.
Being the most powered handheld doesnt mean it can compete with One/PS4.
Schnozberry
Member
If it can get close enough to be on people's radar for ports, that would be all that is required.
GTA5 ran on 512MB of RAM.
NOLA_Gaffer
Banned
Multiple open-world titles ran on the Wii U's 2GB of RAM as well.
Doczu
Member
I'm pretty sure he meant that the upcoming wave of open world games would have to make bigger sacrifices to run on 4 gigs. I'd say RDR2 would probably have to be cut and stichted like Frankenstein to be animated.
sir_bumble_bee
Member
That's funny. Xenoblade X's open world was larger than anything seen in an RPG and it was on Wii U
Sorceror12
Member
Wii U had 2 gig of RAM but not all of it was accessible to games.. I think 1 gig or 1.5 gigs?
Doczu
Member
People need to stop dreaming
The Switch will not get multiplatform games that are AAA. It will only get port for less demanding games like fighters, indie games, HD remasters and etc. There is a reason Nintendo decided to show Skyrim remaster instead of The Witcher 3 or Fallout 4.
How would those vents work for passive cooling? Do you blow in them? You shake the tablet to cool it? There's no reason to make them so thick without active cooling. Dissipate heat through metal surfaces, no vents like those.
Schnozberry
Member
I think a lot of that will depend on the success of the Switch and the kind of audience it draws. There was a much larger chasm of performance between the 360/PS3 and the Wii, yet versions of major AAA titles made it over. The Wii U was abandoned for a lot of reasons, chief of which was it's failure to garner a wide base of support to make ports worth publishing.
Father_Brain
Banned
Yep. There may be a few exceptions here and there, but that's it.
Notwithstanding uncertainty about RAM, and I will assume the chips will run clocked to their standard clock which is perhaps possible when it is docked, the power question boils down to three uncertainties:
1. What Tegra chip will be used (Tegra X1 or Tegra Parker)?
2. How does the power measurement method (known as FLOPS) translate between AMD and NVIDIA cards (NVIDIA's FLOPS tend to be 'stronger', i.e. the communication allows for the FLOPS to be realised more efficiently, giving, and I found this ratio in another Neogaf thread, a ratio of approximately 4:3 in favour of NVIDIA, and this translates approximately one on one to an increase in FLOPS as compared to AMD. We want the AMD FLOPS, because we talk about FLOPS for Xbox One and PS4 as well, being 1.31(?) TFLOPS and 1.84 TFLOPS, but they use AMD rather than NVIDIA).
3. There are two ways (actually, there are more, but without loss of much accuracy we can say there are two) to do computations: FP32 and FP16. The first is slower but more accurate and the second is faster (roughly twice, in fact) but less accurate. Finding a balance between accuracy and speed in this method can increase the FLOPS rate (using only FP16 would give twice the number as compared to using only FP32, for example).
Tegra X1 has 512 GFLOPS for FP32, and Tegra Parker has roughly 750 GFLOPS FP32. If we could, for example, use FP16 for 1/3 of all computations, then we gain a 20% increase in FLOPS.* So you see there can be a significant increase. Xbox One and PS4 cannot use this so-called mixed-precision computation (PS4 Pro can, that's where the rumours about PS4 Pro doing 8.4 TFLOPS come from), so the Switch has a potential advantage in power in this regard.
Let's do a calculation: assuming the Tegra X1, we have 512 GFLOPS of NVIDIA FP32. Assuming (with no particular reason for assuming this, but some more technically-schooled GAFfers called it plausible, but it differs on a game-by-game basis) that 1/3 of the computations can be done in FP16 (which, as I mentioned, results in a 20% gain), we can compute the comparison between the two as follows:
Switch power = 512 * 1.20 * 4/3 = 819 GFLOPS. (the 4/3 is that NVIDIA to AMD ratio I mentioned before) This resultant number is how the Switch actually compares in power to PS4 and Xbox One (which are, respectively, 1.84 TFLOPS and 1.31(?) TFLOPS, remember 1 TFLOPS = 1000 GFLOPS).
If, on the other hand, the Switch uses a Tegra Parker, then the power will be (with the usual caveats that we are guessing a lot of numbers):
Switch power = 750 * 1.2 * 4/3 = 1200 GFLOPS = 1.20 TFLOPS. So, you see that using the latter setup, the Switch could be very close in power to the Xbox One. Remember, though, that the gain from FP16 computations is just a guess, as well as the ratio between NVIDIA and AMD FLOPS (the effect, though, is very real, just not numerically determined).
About the clock I mentioned: there is a standard clock value (Tegra X1 has it at 1 GHz), and the FLOPS rate scales linearly with this clock value (so, halving the clock value will half the FLOPS rate). In handheld mode, the clock value will go down as this saves heat production and ergo battery life. In dock mode, however, active cooling could possibly allow the chip to run at full clock speed and therefore allow the power values I determined above.
Disclaimer: The info I produce here is produced by someone who is not a computer engineer (yet), so there might be something wrong in my explanation. If someone spots an error in my explanation (remember, though, that this is purely a FLOPS determination: we simply don't know how RAM and other things will play into the equation), please correct me.
TL;DR/Conclusion: Depending on many factors, the Switch can possibly be very close to the Xbox One for at least a number of games, but that does assume lot of things we simply do not know, and things that often depend on a game-by-game basis. On the other end of the spectrum, though, the power could possibly be roughly half of the Xbox One, so you see there is a lot we do not know and a large margin for errors.
*: See Thraktor's post (#1551) to see how this gain can be calculated.
Edit: For those interested, I converted Thraktor's system of equations into a nice and simple calculation. Take a specific ratio (I will showcase the ratio FP32 : FP16 = 2 : 3). Do the following:
total = 2 * FP32 + FP16 = 2*2 + 3 = 7.
Now divide FP16 by the total:
gain = FP16 / total = 3/7 ~ 0.43. You can check by solving Thraktor's system that this result pans out every time. Here is the proof:
Assume FP32 : FP16 = n : m.
We must prove that FP32 + FP16 = FLOPS * (1 + m/(2n + m) ) (i.e. the new power is the original plus the fractional gain m/(2n +m) which is the fraction I described above).
Thraktor's system says:
FP32 + (FP16 / 2) = FLOPS
FP32 / FP16 = n/m. => FP16 = FP32 * m/n
FP32 + (FP32 * m/n) / 2 = FLOPS
FP32 (1 + m/(2n) ) = FLOPS
FP32 * ( (2n + m) / (2n) ) = FLOPS
FP32 =FLOPS * (2n) / (2n + m).
FP16 + FP32 = FLOPS * ((2n) / (2n + m)) * (1 + m/n). = FLOPS * ((2n) / (2n + m)) * ( (m + n) / n) = FLOPS * ( (2n * (m + n) / (n * (2n + m)) = FLOPS * (2*(m + n) / (2n + m)) = FLOPS * ( (2n + m) / (2n + m) + m / (2n + m) ) = FLOPS * ( 1 + m / (2n + m) ).
And that is what I had to prove (QED, as they say).
Candescence
Member
That's my take on it as well. If the Switch suddenly starts selling gangbusters for whatever reason, publishers and devs won't be able to ignore the potential revenue from catering to that brand new install base and will likely bend over backwards to get games working on the device even if it's to the detriment of the PS4 and Xbone versions.
JoshuaJSlone
Member
Tablet share seems to be over 10% and growing, though. And of course it would make them accessible on most home PCs.
Which part do you mean? The Shields I've seen look more like a DS without the bottom screen, or GBASP.sir_bumble_bee said:
Herb Alpert
Banned
i think you're right. I would be OK with this...
Not THAT much. UE4 is scale-able to pretty much everything.
The Switch not supporting it would have been a much bigger, much more worrisome news.
Grampa Simpson
Banned
I think that in handheld mode most games will run at 720p or sub 720p resolutions with a variable (gsync style) framerate. I think in docked mode, it will lose the variable framerate, and most games will run at 720p or 1080p with a locked framerate at either 30 or 60 depending on what the developers squeeze out of the system. 3rd party multi platform stuff will most likely all run at 720p 30fps locked framerate while docked.
My phone has 38gflops and supports UE4.
I think it speaks more about its architecture rather than power.
I thing all Nintendo need to do is aim for the most power they can get out of a $275 Switch with decent battery life and be able to have slightly more horsepower when docked.
If I had to guess it will be
Undocked - slightly better than Wii U power
Docked (fans and cooling system turn on) - moderately more powerful than Wii U
People need to stop thinking this will have AAA multiplatform games. Has no one not learnt anything from Wii and Wii U. Also if this was so powerful and Xbone One level graphics Nintendo woukd have shown off The Witcher 3 and Infinite Warfare over Splatoon and Skyrim which both don't look much different to what we can have on Wii U by the way.
If I had to guess it will be
Undocked - slightly better than Wii U power
Docked (fans and cooling system turn on) - moderately more powerful than Wii U
People need to stop thinking this will have AAA multiplatform games. Has no one not learnt anything from Wii and Wii U. Also if this was so powerful and Xbone One level graphics Nintendo woukd have shown off The Witcher 3 and Infinite Warfare over Splatoon and Skyrim which both don't look much different to what we can have on Wii U by the way.
Similar threads
- Locked