Support NeoGAF

[KingSnake]

People on GAF treat FLOPs like they are the only thing that matter, but FLOPs are no different from Bits back in the day. They give a basic measurement, but miss a lot more detail that only the architecture can give you.

The modern architecture makes a GPU more efficient, but doesn't invent new GFLOPS. You can gain maybe 10%-25% due to efficiency and that under an ideal scenario.

 

[Rodin]

If the clock is 384Mhz then my bad.

Switch Portable run a bit better then Wii U... probably ~50% better.

It's more than that. Fast runs at 640x720 on Wii U and 1280x720 with better frame rate and graphics on Switch. That suggests it's at least 2x more powerful when undocked and we don't even know if shinen used fp16. This is also still a port, not a game made for Switch from the scratch.

The modern architecture makes a GPU more efficient, but doesn't invent new GFLOPS. You can gain maybe 10%-25% due to efficiency and that under an ideal scenario.

Actual results point to the gap being larger than what you're suggesting though. 10-25% doesn't mean anything, especially when there's a 8 years gap and a change in vendor between the two.

This is just about the GPU. Gap in CPU and RAM is probably even larger.

 

[ethomaz]

Like I just said, FLoPs are not the best way to tell what a system is capable of anymore than Bits were back in the day. Architecture is far more important than FLoPs are and Switches Architecture is well above what the Wii U has. I laugh at anyone saying that a VIWL5 based GPU is as powerful or more powerful than a Maxwell based GPU.

I just get the clock wrong...

Even accounting for Maxwell vs GCN differences... it will give a ~50% more GPU power for Switch in Portable.

It is not anywhere close to XB1 like some here are saying... it is close to Wii U.

 

[borghe]

I don't 'equate' the two. I'm looking at what we're seeing with the games available right now.
The jump from 720p native to 1080p native is going to eat up quite a lot of that extra docked performance no matter which way you slice it.

But please, share your thoughts on the matter, seeing as you apparently seem to know better.

There is an excellent post in the old "mobile CPUs more powerful than 360" thread that lays it out pretty well. Basically rendering fp32 a docked switch is ~3 times more powerful than 360. Rendering to fp16 its about 5 times more powerful. An XBONE is about 10 times more powerful.

The point being that, while we weaker than XBONE, Switch is a dramatically more powerful and efficient console than 360.

For reference the undocked numbers are about 1.5/3 times for fp32/16 rendering compared to 360. So while that is much closer, I don't think any sane person would expect much more today.

X1 @ 384Mhz = ~190 GFLOPS.

That is about 50% over Wii U power accounting the nVidia vs AMD flops performance.

Wii U has weaker CPU and less memory (I guess) but the GPU power is about these 50% better in Protable mode.

Edit - Seems like my clock for Swtich was wrong... fixing.

Your numbers are incredibly off. Like insanely off. I'll look for the thread but

Effective power is somewhere around
360 ~130GFlops
PS3 ~150Gflops
XBONE ~1.2Tflops
Switch fp32 ~390Gflops
Switch fp16 ~600Gflops

Scaling is something like that. I'll look for the thread.

 

[AmyS]

Julian Eggebrecht on the two Star Wars demos (teaser & gameplay) real-time on GameCube in 2000.

- Both videos run in realtime and use 50% of the hardware

- The X-wing was taken from the original- Modell of ILM from the star wars SpEd with ILMs Textures and Shaders. The x-wing alone has 30'000 poly. and the pilot in the x-wing 4000.

- Both demos run constantly with 60fps double buffered true-color and full screen anti aliased and deflickered.
There were no slow downs, like i said, we still had a lot of puffer last week when the stuff had to be finished.
(i hope you guys understand what the hell he's talking about.)

- We have rebuild the surface 1:1 of the second, original deathstar from the film.
(okay now it comes the bitchy part..)
The simpel shapes are "pro kachel" (can't translate that one..) [it means per tile - Rage] up to 300 polygons. Every element has 512x512 a True color texture.
You can see 25 of it in the demo. There are 70 ties and wingmen onscreen. So, there are Together like 200'000 poly. on 60 fps with up to 8 light sources and gloss, dirt and bump maps.

====If the X-box can handle this i eat a broom (german proverb )====

- We run the exactly same demo on the current X-Box prototype and it can't handle more than 12 fps. But let's wait for the final hardware...

Now seriously: on a 640x480 screen you can't really see the difference from both demos with the original film.
We had last week some guys from ILM here that worked on the SpEd here and coulnd't believe it!

https://www.youtube.com/watch?v=cTvYFFN42R0
https://www.youtube.com/watch?v=iVEVJDmIlxU

 

[KingSnake]

X1 @ 384Mhz = ~190 GFLOPS.

That is about 50% over Wii U power accounting the nVidia vs AMD flops performance.

Wii U has weaker CPU and less memory (I guess) but the GPU power is about these 50% better in Protable mode.

Edit - Seems like my clock for Swtich was wrong... fixing.

You missed this:

http://www.neogaf.com/forum/showthread.php?t=1347552

Edit: you changed it meanwhile.

 

[frankie_baby]

The modern architecture makes a GPU more efficient, but doesn't invent new GFLOPS. You can gain maybe 10%-25% due to efficiency and that under an ideal scenario.

Add efficiency and fp16 together though and you could get a fairly chunky improvement

 

[mavericktopgun]

It is probably not.

It is.

 

[Spiral Insanity]

I hope Turrican will be 2D like Sonic Mania.
Does this mean we could see Turrican games on Virtual Console?

 

[Mokujin]

X1 @ 307Mhz = ~160 GFLOPS.

That is about Wii U power sorry.

Wii U has weaker CPU and less memory (I guess) but the GPU power is about the same in Protable mode.

Comparing ATI R700 flops to Maxwell flops is a no-no, Nvidia flops are 20-25% more efficient than todays AMD cards. [Always messy to raise this point but that's how things work] Boost mode should be accounted also which brings portable to 200 Gflops

Plus 4GB goes a long way over one of the main bottlenecks of that gen, it's 8 times as memory as PS360 had.

But even if it were only WiiU-PS3-X360 on portable mode, well, that's the beauty of it, it's an absolute joy to play that calibre on your hands anywhere, docked mode it's a great bonus on top of that.

 

[ethomaz]

It is.

It is about 50%... I just used old clocks.

 

[KingSnake]

Add efficiency and fp16 together though and you could get a fairly chunky improvement

Let's wait first for FP16 to be used in a significant way before counting on it. Right now only UE4 seems to support it properly but we don't know if it's properly used though.

 

[LordOfChaos]

Kind of a 'water is wet' statement. Any reasonable expectation was between that wide range.

 

[Skittzo0413]

I just get the clock wrong...

Even accounting for Maxwell vs GCN differences... it will give a ~50% more GPU power for Switch in Portable.

It is not anywhere close to XB1 like some here are saying... it is close to Wii U.

I think people are mainly responding to your claim that the Switch isn't more powerful than the Wii U in portable mode from the last page.

We don't know for sure how it compares to the XB1 but we do know it's more powerful than the Wii U in all aspects, even when in portable mode. Even at the old clock speeds this is still true.

 

[Jumeira]

what an utterly useless statement

Seriously.

 

[Jonneh3003]

I'm still astonished at how much performance a machine of this size can deliver.

A lot of it's taken up by a battery too.

 

[KingSnake]

Actual results point to the gap being larger than what you're suggesting though. 10-25% doesn't mean anything, especially when there's a 8 years gap and a change in vendor between the two.

What actual results are you talking about?

 

[ethomaz]

I think people are mainly responding to your claim that the Switch isn't more powerful than the Wii U in portable mode from the last page.

We don't know for sure how it compares to the XB1 but we do know it's more powerful than the Wii U in all aspects, even when in portable mode. Even at the old clock speeds this is still true.

I will fix my first post because the increase in clock that I didn't know.

Docked: Close to Wii U than XB1
Portable: Close to Wii U

 

[Spiral Insanity]

I think people are mainly responding to your claim that the Switch isn't more powerful than the Wii U in portable mode from the last page.

We don't know for sure how it compares to the XB1 but we do know it's more powerful than the Wii U in all aspects, even when in portable mode.

Correction, especially when it is in portable mode.

 

[ArchedThunder]

It is about 50%... I just used old clocks.

In portable it's at least 2X Wii U and in docked it sits somewhere in the middle between Wii U and Xbox One. Not only are the flops not the most important part, the architectural difference between Wii U and Switch is massive and then there is also the RAM and CPU.

 

[Kurt]

I think we will see games at E3 so we have some decent (compare) of what's possible on the switch. I think that the leap between them won't be huge.
The same minor differences of what we already see with ps4 pro vs xbox one.

 

[RoadHazard]

Actually, it sits between the Game Boy and the PS4 Pro.

 

[Fredrik]

I just listened to the part about Factor 5. So in a nutshell:

- they solves all outstanding legal issues with all Factor 5 entities, also the US one.
- Factor 5 GmbH, the German company, is a legally registered company again since a few weeks.
- they've bought back all their right to their old own IPs
- this also includes all IPs from former German publishers Rainbow Arts and Softgold, which were with THQ prior to their bankruptcy.
- He won't confirm any development starting but it is their idea to bring back old IPs and games and maybe develop further sequels "at some point".

So all in all, there is a prospect of re-releases of old F5 games such as Turrican and maybe even a new Turrican, but this does not change anything on the Roque Squadron games for now.

This is fantastic! I would buy the Amiga Turrican trilogy day 1 for Switch if it would come there, Turrican 1 alone could hold it's own against any retro rerelease we've seen so far.

 

[VariantX]

Yup, for sure.

People think simply because Zelda is stylized and runs at 900p on TV mode that the Switch can't handle even basic PS3/360 ports. I'm not sure if they are just trying to temper their expectations or trolling.

Which is a ridiculous standpoint to have because neither of those consoles could run Zelda at 900p let alone 720p. People quickly have forgotten how much harder it got to have games running at 720p torwards the second half of that generation because of developers having to push visual effects while making trade offs here and there.

 

[frankie_baby]

Let's wait first for FP16 to be used in a significant way before counting on it. Right now only UE4 seems to support it properly but we don't know if it's properly used though.

Who says it isn't being used already

 

[Garibaldi]

But I thought The Last of Us couldn't run on the switch!

Spoiler

some people on this board have actually argued that...

It can't though.

Spoiler

The executables wouldn't run on the Switch *chortle*

 

[miller.skippins]

Actually, it sits between the Game Boy and the PS4 Pro.

Liar!!

I did my calculations based on the specs and it's between the ENIAC and IBM Watson

 

[Skittzo0413]

I will fix my first post because the increase in clock that I didn't know.

Docked: Close to Wii U than XB1
Portable: Close to Wii U

Yeah that's a more fair way of putting it but it still may not be true. We'll have to wait to see how multiplatform ports like Steep stack up or see how much of the FP16 double speed processing can actually be used. If it's actually 70% for some games like that Ubisoft dev said then that would put the docked Switch closer to the XB1 easily.

Correction, especially when it is in portable mode.

Er, sure? It's obviously more powerful docked than portable though so I'm not sure what this means haha

 

[wildfire]

Which is what is expected no?

There actually were like 5 weirdos who kept claiming the Switch is weaker than the Wii U because it used mobile hardware.


The real debate is that you either think the Switch is closer to the Wii U or the Xbox One and most people thought it was closer to the Wii U in capabilities when this asserts it's closer to the Xbox One.

 

[tkscz]

I just get the clock wrong...

Even accounting for Maxwell vs GCN differences... it will give a ~50% more GPU power for Switch in Portable.

It is not anywhere close to XB1 like some here are saying... it is close to Wii U.

I'm saying clocks are one thing, but different tech allows the Switch to produce aspects and effects that the Wii U can't. Not saying it will change how many GFLoPs the Switch has over the Wii U, only that Switch games graphically will be closer to what XBO games look like than Wii U due to the tech. An Nvidia Tigra TX1 is closer to an AMD HD Radeon 7750 than an AMD HD Radeon 4750 is.

 

[KingSnake]

Who says it isn't being used already

The resolution and performance of the games that we know about until now would point that it's probably not used in a meaningful way. Maybe Snake Pass uses it, though.

 

[Gotdatmoney]

Proof needed.

Just research the hardware? The GPU alone is far far better than the PS3 or 360's.

You can argue that NVIDIA GFLOPS is stronger than AMD GFLOPS and I agree.

Even so Switch GPU power is a bit better than Wii U... like 20-30%.

No where close to XB1 sorry.

I wasn't contesting that the Switch was close to the X1. I was contesting that calculating a flops number and dictating strength is not how hardware stremgth works.

I wasn't even going to make the Nvidia flops vs AMD flops argument becausenO dont know how true that is in a fixed hardware console configuration.

But being able to run modern toolsets, utilizing a vastly more efficient architecture and in general just running on much newer hardware gives advantages beyond 20-30% WiiU. And this is just the GPU. CPU wise there is no contest. Same with ram size. Bandwidth is.much more arguable though.

And even if you look at the raw raw numbers alone you are going to see better results in terms of looks on handheld Switch games than WiiU games.

 

[Spiral Insanity]

Yeah that's a more fair way of putting it but it still may not be true. We'll have to wait to see how multiplatform ports like Steep stack up or see how much of the FP16 double speed processing can actually be used. If it's actually 70% for some games like that Ubisoft dev said then that would put the docked Switch closer to the XB1 easily.



Er, sure? It's obviously more powerful docked than portable though so I'm not sure what this means haha

There hasn't been a game that runs better docked, AFAIK. Even if it is more powerful docked it seems the hardware shines at 720p, making up for the difference.

Even the Neo Geo games have some issues docked.

 

[Sora the Cat]

Actually, it sits between the Game Boy and the PS4 Pro.

I dont think you guys realize how lame these jokes really are.

 

[dragonyeuw]

Switch has:

-a much better/more modern cpu/gpu than wiiU
-more RAM
-Developing​ is MUCH easier on Switch
-openGL/Vulkan and Nvidia experience

I guess only time will tell.

If anything, closer to the XBO than the Wii U.

The GPU is a modern asset Maxwell based GPU, while the Wii U has a 2010 based VIWL5 AMD CPU (something around the HD Radeon 4000 series) and the XBO has an GCN based mordern GPU (based on the HD Radeon 7000 series). To give a better idea, API wise, the Switch can do things on the level of DX11.1 (maybe 12, we don't know what the costume API can fully do yet), while the Wii U tops out at DX10 with some assets from DX11 (of course Nintendo can't use DX as that's a Microsoft copyrighted API).

The CPU is an ARM A57, which is much better than the GameCube based (but upgraded) IBM Power CPU the Wii U has, and even has advantages over the Jaguar CPU the XBO has.

The Switch has 4GBs of LPDDR4 RAM with 3.2GBs of it usable for games. LPDDR4 is faster and more efficient than the 2GBs (only 1GB usable) GDDR3 in the Wii U but not as fast as the 8GBs (5GB usable) GDDR5 in the XBO, but closer to XBO than Wii U.

People on GAF treat FLOPs like they are the only thing that matter, but FLOPs are no different from Bits back in the day. They give a basic measurement, but miss a lot more detail that only the architecture can give you.

Yeah I get how the specs line up, I should have said I was asking more rhetorically because the statement ' it's between the Wii U and Xbox one' on its own merits doesn't say much of anything.

I imagine with time, given the ease of development, the Switch will be punching above its weight so to speak, and pull a few surprises in spite of where it sits spec-wise to the PS4 and XB1.

 

[AmyS]

Xbox 360/Wii U....|.Switch..|...............................Xbone...........PS4

 

[efyu_lemonardo]

The resolution and performance of the games that we know about until now would point that it's probably not used in a meaningful way. Maybe Snake Pass uses it, though.

Is it technically possible to render different parts of the same frame using different shader accuracy?

 

[KingSnake]

Nvidia flops vs AMD flops argument is relevant only in the PC world. In the console world is a bit more refined than that. What Nvidia brings on the table for sure as big advantage over AMD is the tile rendering and a more efficient use of the memory bandwidth.

 

[tzare]

I find interesting this power debate since many have said that this generation consoles are really low-mid range powered pcs with shitty cpus. And now, just after the also 'average improvement ' of the recently relesaseed PS4pro , and the upcoming Scorpio, we have another 'contender' that isn't even a match for vanilla XBO.
So despite being a relatively powerful home console compared to Ps360 and WiiU, it is not that close to XBO which usually gets the worst multi platform games. Are people really expecting that developers will care and tailor their multi platforms to the switch and squeeze its flops?

 

[xenorevlis]

The Switch has 4GBs of LPDDR4 RAM with 3.2GBs of it usable for games. LPDDR4 is faster and more efficient than the 2GBs (only 1GB usable) GDDR3 in the Wii U but not as fast as the 8GBs (5GB usable) GDDR5 in the XBO, but closer to XBO than Wii U.

People on GAF treat FLOPs like they are the only thing that matter, but FLOPs are no different from Bits back in the day. They give a basic measurement, but miss a lot more detail that only the architecture can give you.

This is incorrect. XBO uses 8GB of DDR3. It's PS4 that uses GDDR5. Also operation count may differ in terms of what hardware efficiency can produce in the real world but with all of them on modern architectures they are an easy way to assess power and fillrate.

There is no realistic scenario in which Switch is at 700 Gflops with FP32. Come on.

Of course there is. The custom X1 can definitely hit close to that level and apparently there is a firmware update on the horizon to clock up the docked mode slightly. In mobile mode I also believe it is close to double the power of Wii U.

Julian Eggebrecht said (and I believe him) that developing is at least as easy as on PS4. Even easier as he hinted.

It's probably just as easy, but when you compare to PS4 you actually have to consider additional aspects like Pro features, Spotify (which can trigger music to mute in-game), party chat that can override in-game chat, trophies, etc.

 

[Zombegoast]

Well certainly the Switch got a more graphically intensive racing game than the Xbox One at launch.

But for real,

Comparing architecture is apples to orange. If people look just numbers on GPU, people will buy older GPU expecting them to perform better than a 1060.

 

[frankie_baby]

The resolution and performance of the games that we know about until now would point that it's probably not used in a meaningful way. Maybe Snake Pass uses it, though.

I'd imagine Fast rmx probably uses it

 

[Caio]

We know Software is what really matters; but come on Nintendo, you can give us some more power, to allow development of multiplatform games which do not need to be tuned down to hell.

 

[frankie_baby]

Is it technically possible to render different parts of the same frame using different shader accuracy?

Yes

 

[Jonneh3003]

Proof needed.

I mean judging off handheld performance in Zelda alone we know the Switch can out do Wii U which had games like:

There shouldn't really be any question if it can out perform PS3 games.

 

[OryoN]

Rogue Leader was such a beast game. Did they ever finally gotta it running smoothly on dolphin emulator?

 

[wildfire]

I find interesting this power debate since many have said that this generation consoles are really low-mid range powered pcs with shitty cpus. And now, just after the also 'average improvement ' of the recently relesaseed PS4pro , and the upcoming Scorpio, we have another 'contender' that isn't even a match for vanilla XBO.
So despite being a relatively powerful home console compared to Ps360 and WiiU, it is not that close to XBO which usually gets the worst multi platform games. Are people really expecting that developers will care and tailor their multi platforms to the switch and squeeze its flops?

If it really is easier to develop for than the PS4 then that should mean it is cheaper than porting to the XB1.

As long as the Switch has healthy sales similar to the post price cut 3DS then they'll get ports.

 

[efyu_lemonardo]

Yes

So you can switch between fp16 and fp32 mid frame? That makes fp16 performance extremely relevant.

 

[tkscz]

Xbox 360/Wii U....|.Switch..|...............................Xbone...........PS4

Even that's hard to say for sure. Wii U had a much better GPU than the Xbox 360 (Radeon X1900 with upgraded shader tech vs a Radeon HD 4750 with upgraded shader tech) and had double the usable RAM than the 360. However it had a slower, harder to use CPU and crap bandwidth, making development for 3rd parties a chore.

 

[KingSnake]

Of course there is. The custom X1 can definitely hit close to that level and apparently there is a firmware update on the horizon to clock up the docked mode slightly. In mobile mode I also believe it is close to double the power of Wii U.

http://www.anandtech.com/show/8811/nvidia-tegra-x1-preview/2

The theoretic maximum for X1 is 512 GLFOPS at FP32. We already know that the X1 in Nvidia Shield TV 2015 throttles a lot at its maximum clocks. We have the clocks from Eurogamer which make for maximum 393 GFLOPS while docked.

Wii U had 176 GFLOPS.

So tell me by what magic do you reach 700 GFLOPS at FP32?