Support NeoGAF

[StuBurns]

Either way that's no bigger difference than you see in some PS3/360 cross platform games.

Isn't there like a three fold difference in resolution? That's a bigger technical gulf than any PS3/360 multiplatform game.

 

[Datschge]

And what do you mean console?

They also use SoC designs. If not from the start then later on as optimization. SoCs are nothing special, so an ARM combined with an PICA200 on a SoC is nothing earth shattering either.

 

[Lonely1]

Er, what are you guys talking about in those MHTri images? The overall IQ of the Wii version was crap. Also, is this using an 4GB cart? if not then downgrades might be required by the storage medium alone.

 

[Datschge]

The three images with similar area, the same resolution and aspect ratio:

The discussion about differences is as esoteric as those of PS360 games.

 

[Mr_Brit]

Er, what are you guys talking about in those MHTri images? The overall IQ of the Wii version was crap. Also, is this using an 4GB cart? if not then downgrades might be required by the storage medium alone.

If the Wii IQ is crap then what is the 3DS IQ?

 

[Lonely1]

If the Wii IQ is crap then what is the 3DS IQ?

In the context of Wii games, MHTri had remarkable bad IQ... Even the plain text was hard to read on a good SDTV set. No need for such kind of reply.

The discussion about differences is as esoteric as those of PS360 games.

Well, is easy to see that the 3DS characters clearly have shadows.

 

[StuBurns]

In the context of Wii games, MHTri had remarkable bad IQ...

While true, if you're comparing the technical performance of the machines, it was still drawing a lot more.

 

[Luigiv]

Interesting, that image shows 2xAA.
Perhaps they have slightly jittered samples in 2D mode?

Why would they jitter sample when they could implement real SSAA for the same price?

 

[beril]

While true, if you're comparing the technical performance of the machines, it was still drawing a lot more.

In 3d-mode and including the second screen the 3DS renders 87.5% as many pixels as the Wii does in progressive scan. Granted there's probably not alot going on the second screen; but then again it has to render twice as much geometry on the upper screen.

Either way the screen resolution isn't really much to argue about. A 3DS game will obviously never look exactly like a Wii game because of the lower resolution, but because of the resolution it can use reduce the texture sizes and possibly reduce some geometry detail as well without being visually noticable

 

[Mr_Brit]

Why would they jitter sample when they could implement real SSAA for the same price?

Isn't the performance penalty for 2x2 supersampling greater than the 3D mode?

 

[kinggroin]

I'm not denying that fact.

However, the difference between launch PS2 games and some of the games released later in its life are SO significant that it may well have been a different platform all together. Take a look at something like Burnout 1 vs Burnout Revenge. The difference is ENORMOUS.

We went from a launch title like this...

to this...

You really think we're going to see leaps of that magnitude on 3DS?

Maybe not, but the best of the current 3ds titles are starting much higher thanks to a simplified and familiar game development language and tools. I'd argue that street fighter IV at 60fps is every bit as good looking as the very best fighter on ps2.

Monster Hunter looks absolutely amazing as well, surpassing the ps2 iterations on texture resolution, lighting, and texture effects (shaders).

There's no denying that the 3ds is a very capable handheld when handled right, but its a but disengenious to make the comparison you're making in regards to leap in visuals from early to late Gen stuff. The ps2 was a bitch to learn initially, and had so much higher to go considering how low things started.

 

[M3d10n]

Why would they jitter sample when they could implement real SSAA for the same price?

Because it's easier: the code for it is already there (the same one that renders 3D). It also produces better results, because you blend two distinct samples together.

For 2X SSAA they'd need to render the game to a 565*340 buffer then downscale it via bilinear filtering, which wouldn't look as good.

 

[BDGAME]

The three images with similar area, the same resolution and aspect ratio:

The discussion about differences is as esoteric as those of PS360 games.

So the city dock look the same in both games. The Wii game has a better texture in the baskets and 3DS version has better light and shadows.

Now put 3D and 3DS version is the superior.

 

[StuBurns]

In 3d-mode and including the second screen the 3DS renders 87.5% as many pixels as the Wii does in progressive scan. Granted there's probably not alot going on the second screen; but then again it has to render twice as much geometry on the upper screen.

Either way the screen resolution isn't really much to argue about. A 3DS game will obviously never look exactly like a Wii game because of the lower resolution, but because of the resolution it can use reduce the texture sizes and possibly reduce some geometry detail as well without being visually noticable

Well we're looking at the 2D screens, so what it's doing in 3D we can't really talk about yet.

And maybe I'm wrong, but surely it's technically possible to have 3DS IQ best normal 480p no AA?

 

[M3d10n]

So the city dock look the same in both games. The Wii game has a better texture in the baskets and 3DS version has better light and shadows.

Now put 3D and 3DS version is the superior.

The Wii screenshot is dowscaled. If it rendered natively at the 3DS resolution textures would probably look the same due to mip-mapping.

At lower resolutions, it's not possible to show high resolution textures unless they were either up-close or don't use mip-mapping at all (which would cause shimmering when zoomed out).

Just for fun, someone should set Dolphin to 400x240 and take screenshots of that. Or maybe do it with some random PC games.

 

[PdotMichael]

So the city dock look the same in both games. The Wii game has a better texture in the baskets and 3DS version has better light and shadows.

Now put 3D and 3DS version is the superior.

logic?

The texture quality is overall lower and has a more visible LOD-system (look at the right and left corner) and we don't know about framerate or polygoncount (we need better pictures).

Just for fun, someone should set Dolphin to 400x240 and take screenshots of that. Or maybe do it with some random PC games.

or scales the 3DS pictures up to 480p

 

[BDGAME]

logic?

The texture quality is overall lower and has a more visible LOD-system (look at the right and left corner) and we don't know about framerate or polygoncount (we need better pictures).



or scales the 3DS pictures up to 480p

3D is not a free process. it requires more power from machine to do it and the visuals look better in 3D.

 

[BDGAME]

The Wii screenshot is dowscaled. If it rendered natively at the 3DS resolution textures would probably look the same due to mip-mapping.

At lower resolutions, it's not possible to show high resolution textures unless they were either up-close or don't use mip-mapping at all (which would cause shimmering when zoomed out).

Just for fun, someone should set Dolphin to 400x240 and take screenshots of that. Or maybe do it with some random PC games.

So the mip-maping problem is the responsible to that bad wall texture in MGS3D?

 

[Fafalada]

I can't stop laughing at the people who say a 3DS game is visually enhanced over a Wii game.
That's like saying a Vita game is visually enhanced over a PS3 game.

Afaik, 3Ds GPU stacks to Wii's a fair bit more favorably then Vita's does to PS3's - "on-paper".
Granted I'm basing this on the supposed Vita's spec = "Iphone4 x 8.0934535 OMFG" or whatever the marketing claims are.

 

[PdotMichael]

3D is not a free process. it requires more power from machine to do it and the visuals look better in 3D.

3ds: 2 x 400x240 = 96000

Wii: 720 * 480 = 346500

 

[StuBurns]

3ds: 2 x 400x240 = 96000

Wii: 720 * 480 = 346500

It's not that simple, 3D isn't the same as doubling the resolution. And the bottom screen is still there.

It's not the same either, but for all intents and purposes, this does appear to be basically par between the two if you ignore the inherent resolution limiations of the 3DS, which is very impressive.

 

[M3d10n]

So the mip-maping problem is the responsible to that bad wall texture in MGS3D?

Not sure if serious.

 

[BDGAME]

Not sure if serious.

Ok, maybe I don't understand your post.

 

[kinggroin]

3ds: 2 x 400x240 = 96000

Wii: 720 * 480 = 346500

Wii is 640x480 with rectangle pixels no?

 

[beril]

Well we're looking at the 2D screens, so what it's doing in 3D we can't really talk about yet.

Well I think most people are assuming that it'll look the same in 3D sans AA, and possibly lower framerate. It's not very likely that they'll be using lower geometry meshes once you slide up the 3D.

And maybe I'm wrong, but surely it's technically possible to have 3DS IQ best normal 480p no AA?

Depends on how you define IQ. It obviously can't output more pixels than there are physically on the screen. Even it if would render natively 1080p and downscale it to 240p most people would probably say an 480p image has better image quality. And I'm assuming there's a framebuffer size cap, or just not enough VRAM to do something that ridiculous, even though it would look very nice

 

[StuBurns]

Depends on how you define IQ. It obviously can't output more pixels than there are physically on the screen. Even it if would render natively 1080p and downscale it to 240p most people would probably say an 480p image has better image quality. And I'm assuming there's a framebuffer size cap, or just not enough VRAM to do something that ridiculous, even though it would look very nice

I didn't mean that extreme, but let's say for example, a 720p image with no AA, is going to look worse than a 700p image with good AA, is it not just a question of adding more and more AA and AF to compensate?

 

[dark10x]

I believe it outputs 720x480, but the actual area being rendered is less than that. It's for this reason why so many Wii games have large overscan regions. So, really, the resolution between different Wii games tends to vary a bit. Most of Nintendo's titles fill more of the screen while many other are stuck with massive overscan borders.

 

[beril]

I didn't mean that extreme, but let's say for example, a 720p image with no AA, is going to look worse than a 700p image with good AA, is it not just a question of adding more and more AA and AF to compensate?

Sure but it still only has half the vertical resolution, so would need a LOT of AA to make up for it. And I'm guessing the reason that no games use AA in 3D mode is that there's not enough VRAM for it.

 

[Mr_Brit]

Sure but it still only has half the vertical resolution, so would need a LOT of AA to make up for it. And I'm guessing the reason that no games use AA in 3D mode is that there's not enough VRAM for it.

I'm pretty sure no 3DS games use AA in 3D mode is because 3DS only supports supersampling AA.

 

[StevieP]

Although this is true, it should be pointed out that the ARM11 is two processor generations (within the same architecture family) behind the state of the art (Cortex-A9)...it's pretty much old news

There is nothing "state of the art" about the A9, either. It's just not as far behind.
Cutting edge is Cortex A15. Both Apple and Sony are both still using 2009-era Power VR 543's for their GPU as well, which are easily outclassed by the cutting edge Rogue (6) Series.

 

[Celine]

I'm pretty sure no 3DS games use AA in 3D mode is because 3DS only supports supersampling AA.

Not much a game but Pokedex seems to employ some sort of AA even in 3D.

 

[kinggroin]

I believe it outputs 720x480, but the actual area being rendered is less than that. It's for this reason why so many Wii games have large overscan regions. So, really, the resolution between different Wii games tends to vary a bit. Most of Nintendo's titles fill more of the screen while many other are stuck with massive overscan borders.

According to these guys at beyond3d, it is an anamorphic 640x480 image which your TV them has to stretch (which explains why 4:3 has a sharper image).

http://forum.beyond3d.com/showthread.php?t=46169

 

[EvilMario]

That's a pretty crap picture to pull from Tri on the Wii. It's not even that blurry on my 46" screen.

 

[StuBurns]

There is nothing "state of the art" about the A9, either. It's just not as far behind.
Cutting edge is Cortex A15. Both Apple and Sony are both still using 2009-era Power VR 543's for their GPU as well, which are easily outclassed by the cutting edge Rogue (6) Series.

For the laymen and women here such as myself, what kind of tangible difference is that?

 

[Fernando Rocker]

So... is the rumor true about the first dev kits having less RAM than the actual RAM amount?

 

[firelink]

Regarding the 3DS GPU vs the Wii GPU:

I am going to pursue this topic as cautiously as I can, making sure I do not make any factual statements - because there are none to be made.

Wii's GPU is fixed function only, and, although the 3DS GPU is also, it contains enhancements that make it easier to program for as opposed to the Wii, which has to be hard coded. I am not sure if the PICA200 inside of the 3DS came with these enhancements. It looks like it, but there is no way for me to tell. Based on that, the 3DS GPU should either be capable of the same shaders as the Wii, or much better ones, closer to the shaders possible on PS360 and Vita.

No one is really quite sure what clockspeed the 3DS is clocked at, however, an analysis of the hardware and how much power it uses adds up to the 3DS using around 400Mhz. The analysis came to this conclusion because the PICA200 uses 1mWh per MHz. The 3DS consumes 1300 mAh. Apparently if you do the math, and add up the power consumption of all the components, it comes out exactly right if you use the 400MHz figure. But no one knows. It was often rumored that the 3DS had dual ARM11 processors at 233MHz and a 133MHz DMP PICA200, however, there were also rumors that it would contain 64MB of a different type of RAM, no VRAM, and would even have a nVidia Tegra chip instead of the DMP one, so no one can really know who/what to trust. And considering the 1048 0H ARM core in the 3DS is a custom model number, we might never know.

For the sake of being fair, I'll use 100MHz, 133MHz and 400MHz numbers.

According to DMP, the PICA200 is capable of:

400 Mpixels/second @ 100MHz
532 Mpixels/second @ 133MHz (I got this via ratios, check my math, make sure I made no errors)
1600 Mpixels/second @ 400MHz

The Wii, in comparison, is capable of:
~970 Mpixels/second.

3DS polygon performance:

40M polys/second @ 100MHz
~50M polys/second @ 133MHz (40*1.33)
160M polys/second @ 400MHz

Wii polygon performance:

30 million polys/second.

I am guessing here with the Wii, but it has been said that the Wii has an overclocked Flipper inside of it. Flipper had a peak polygon performance of 20M polys/second. The Wii is rumored to have a clock speed of 1.5x the GCN, making the polygon performance 30M polys/second

Given these details, the amount of RAM (96MB FCRAM vs 24MB 1T-SRAM and 64MB GDDR3 RAM) and VRAM (3DS: 4MB vs Wii: 3MB), one can conclude that the 3DS GPU is overall more powerful given the resolution.

Unless the 3DS is clocked higher than 133MHz, the fillrate for it is about 400 million lower than the Wii., or almost half. The Wii at 480p widescreen has a pixel count of 409,920. The 3DS in 3D mode, has a pixel count of 192,000. As you can see from this, the 3DS is using a resolution less than half of that of the Wii. If a game was made purely in 2D mode, it would be a 4th.

Given this, a Wii game downsized to the 3DS would have extra headroom for things like shaders and other functions, if the 3DS is clocked at 133MHz. If it is clocked higher like the power analysis suggests, at 400MHz, the 3DS could actually run a Wii game at the Wii's resolution better than the Wii can.

No matter what the clockspeed though, the 3DS's polygon performance is better or much better.

 

[beril]

I'm pretty sure no 3DS games use AA in 3D mode is because 3DS only supports supersampling AA.

What I meant was that if it was a performance issue some less demanding games might still use AA even in 3D, but if it's a framebuffer memory issue it would explain why no one does it because it wouldn't be possible. But apparently the Pokedex proves me wrong

 

[Durante]

There is nothing "state of the art" about the A9, either. It's just not as far behind.

A quad-core A9 absolutely is state of the art. There isn't a single product shipping with it yet. And there won't be a shipping product with A15 for at least a year.

 

[Mr_Brit]

For the laymen and women here such as myself, what kind of tangible difference is that?

Well PowerVR series 6 is meant to be nearly twice as fast as the 5 series in the Vita so if Sony were to go for the series 6 instead graphics performance should be nearly doubled.

Edit: Also, Series 6 has newer features as well.

 

[StevieP]

For the laymen and women here such as myself, what kind of tangible difference is that?

CortexA9 vs A15, 1ghz vs 2ghz and a deeper pipeline (basically 3x the processing power at a lower power consumption)

PowerVR 543: DX9/OGL2.1, 1000mpixels fillrate.
PowerVR 6: DX10.1/OGL3.2, 6000mpixels fillrate.

 

[BDGAME]

Regarding the 3DS GPU vs the Wii GPU:

I am going to pursue this topic as cautiously as I can, making sure I do not make any factual statements - because there are none to be made.

Wii's GPU is fixed function only, and, although the 3DS GPU is also, it contains enhancements that make it easier to program for as opposed to the Wii, which has to be hard coded. I am not sure if the PICA200 inside of the 3DS came with these enhancements. It looks like it, but there is no way for me to tell. Based on that, the 3DS GPU should either be capable of the same shaders as the Wii, or much better ones, closer to the shaders possible on PS360 and Vita.

No one is really quite sure what clockspeed the 3DS is clocked at, however, an analysis of the hardware and how much power it uses adds up to the 3DS using around 400Mhz. The analysis came to this conclusion because the PICA200 uses 1mWh per MHz. The 3DS consumes 1300 mAh. Apparently if you do the math, and add up the power consumption of all the components, it comes out exactly right if you use the 400MHz figure. But no one knows. It was often rumored that the 3DS had dual ARM11 processors at 233MHz and a 133MHz DMP PICA200, however, there were also rumors that it would contain 64MB of a different type of RAM, no VRAM, and would even have a nVidia Tegra chip instead of the DMP one, so no one can really know who/what to trust. And considering the 1048 0H ARM core in the 3DS is a custom model number, we might never know.

For the sake of being fair, I'll use 100MHz, 133MHz and 400MHz numbers.

According to DMP, the PICA200 is capable of:

400 Mpixels/second @ 100MHz
532 Mpixels/second @ 133MHz (I got this via ratios, check my math, make sure I made no errors)
1600 Mpixels/second @ 400MHz

The Wii, in comparison, is capable of:
~970 Mpixels/second.

3DS polygon performance:

40M polys/second @ 100MHz
~50M polys/second @ 133MHz (40*1.33)
160M polys/second @ 400MHz

Wii polygon performance:

30 million polys/second.

I am guessing here with the Wii, but it has been said that the Wii has an overclocked Flipper inside of it. Flipper had a peak polygon performance of 20M polys/second. The Wii is rumored to have a clock speed of 1.5x the GCN, making the polygon performance 30M polys/second

Given these details, the amount of RAM (96MB FCRAM vs 24MB 1T-SRAM and 64MB GDDR3 RAM) and VRAM (3DS: 4MB vs Wii: 3MB), one can conclude that the 3DS GPU is overall more powerful given the resolution.

Unless the 3DS is clocked higher than 133MHz, the fillrate for it is about 400 million lower than the Wii., or almost half. The Wii at 480p widescreen has a pixel count of 409,920. The 3DS in 3D mode, has a pixel count of 192,000. As you can see from this, the 3DS is using a resolution less than half of that of the Wii. If a game was made purely in 2D mode, it would be a 4th.

Given this, a Wii game downsized to the 3DS would have extra headroom for things like shaders and other functions, if the 3DS is clocked at 133MHz. If it is clocked higher like the power analysis suggests, at 400MHz, the 3DS could actually run a Wii game at the Wii's resolution better than the Wii can.

No matter what the clockspeed though, the 3DS's polygon performance is better or much better.

If your analise is wright, the first gen games of 3DS are really underperformed, like the first gen of wii.

 

[Datschge]

...like most GC/Wii games period.

Fixed

 

[Nuclear Muffin]

If your analise is wright, the first gen games of 3DS are really underperformed, like the first gen of wii.

Well, that is pretty normal. Most 1st gen games barely scratch the surface of their respective hardware.

Also (unlike with previous Nintendo hardware), Nintendo will not be the developer that pushes out the best graphics on their own console - since they have little experience with modern shader hardware.

 

[PdotMichael]

You can't compare marketing data. That's bullshit - like the 66 millions polygons @PS2 for example.

If the 3DS is easily faster than Wii, Nintendo games like Luigi's Mansion 2, Super Mario Land 2, Mario Kart would look better than the Wii/GCN versions. And Luigi's Mansions was a launch game for the GCN.

A little reality check, please.

 

[BDGAME]

You can't compare marketing data. That's bullshit - like the 66 millions polygons @PS2 for example.

If the 3DS is easily faster than Wii, Nintendo games like Luigi's Mansion 2, Super Mario Land 2, Mario Kart would look better than the Wii/GCN versions. And Luigi's Mansions was a launch game for the GCN.

A little reality check, please.

These games don't launch yet to know if they will or not look better than the Wii/NGC versions.

 

[Boney]

You can't compare marketing data. That's bullshit - like the 66 millions polygons @PS2 for example.

If the 3DS is easily faster than Wii, Nintendo games like Luigi's Mansion 2, Super Mario Land 2, Mario Kart would look better than the Wii/GCN versions. And Luigi's Mansions was a launch game for the GCN.

A little reality check, please.

Well Super Mario Land 2 does look dated.

 

[nincompoop]

Well, that is pretty normal. Most 1st gen games barely scratch the surface of their respective hardware.

Also (unlike with previous Nintendo hardware), Nintendo will not be the developer that pushes out the best graphics on their own console - since they have little experience with modern shader hardware.

And yet Super Mario Galaxy and Metroid Other M have more impressive shaders than anything we've seen on 3DS from any developer so far.

Well Super Mario Land 2 does look dated.

I bet if you lowered a SMG image to the original GB's resolution and greyscaled it, you'd barely be able to tell the difference!

 

[PdotMichael]

These games don't launch yet to know if they will or not look better than the Wii/NGC versions.

they will not looks better than the Gamescom demos

 

[Brad Grenz]

The Wii screenshot is dowscaled. If it rendered natively at the 3DS resolution textures would probably look the same due to mip-mapping.

At lower resolutions, it's not possible to show high resolution textures unless they were either up-close or don't use mip-mapping at all (which would cause shimmering when zoomed out).

Just for fun, someone should set Dolphin to 400x240 and take screenshots of that. Or maybe do it with some random PC games.

Done! Actually, these came out as 400x224 in 16x9 mode.

First is one done in DX9 with no AA and no anisotropic filtering:

And here is one in DX11 with 4x MSAA and 16x Aniso:

As you can see, it's a bit clearer but the low resolution destroys most of the texture detail in both cases. The aliasing is also still pretty bad in motion with lots of pixel crawling even with MSAA, again because there are so few pixels to work with.

 

[Lazy8s]

The polygon performance of a hardware can be measured in an innumerable number of ways, and the marketing departments of different companies share no standard way of doing it.

Couple that with the fact that triangle rates are rarely the limiting factor in a scene, and quoting polygon figures is just noise.