Support NeoGAF

[Shogmaster]

Your point? The resolution of those pics is not a nice multiple of 480x272, so a crap load of obvious artifacts where introduced.

The large marketing screeny in the beginning of this thread was also not a multiple of 480x272. It was probably an acculative buffer render shot from dev station @ 960x544, then the web site we linked from resized it down to 700x394 via bicubic resampling, adding a shit load of AA into the images.

Your pic does not have the image quality of a native 480x272 rendering.

Sure it does. It's been resized to 480x272 with nearest neighbor resampling. It probably won't match how that exact scene would look on a PSP since we don't have such an image from the marketing monkeys, but it's the closets we have in this thread for approximating aliasing at that resolution.

Shog do you have a PSP? Your resized screen doesn't look nearly as clean as what the PSP outputs.

I don't own my own PSP, but I play my buddy's all the time. And playing through Ridge Racers more than he did, I can tell you that 480x272 without aliasing looks pretty much jaggy city.

Then why is there zero dithering? You can't have current gen games with modern lighting techniques and have no dithering in 16-bit color (or very serious banding issues if the ps2 can't do dithering), it just ain't possible. That's why the games on the psp that do run in 16-bit are very apparently they are indeed 16-bit color because of the dithering. You mentioned the ps2 gtas were in 16-bit color. If that were true, the colored lighting plus full screen filter effects would look absolutely horrendous in 16-bit color. Night time scenes alone would be a horrible dithered mess.

I thought they were a dithered mess? I do admit that since I don't have a PS2 in the house anymore, I can't be too sure.

 

[Pimpbaa]

I thought they were a dithered mess? I do admit that since I don't have a PS2 in the house anymore, I can't be too sure.

No, they were not! Jaggy mess maybe, but no dithering in sight.

 

[CamHostage]

The resolution of those pics is not a nice multiple of 480x272, so a crap load of obvious artifacts where introduced. Your pic does not have the image quality of a native 480x272 rendering.

Shog do you have a PSP? Your resized screen doesn't look nearly as clean as what the PSP outputs.

Again, PSP dev shots off the screen buffer are bigger than native res. The poster (or Ubisoft) did not go into PhotoShop and make the screens stupidly big on purpose. Unless Ubisoft went and did that, this is what the devs handed off (not sure why they're this size -- they're usually 1200x900 or something to that effect, but maybe in current dev kits, you can target render?) You could size them down and repost, but you'd be introducing PhotoShop artifacts into the shot, so it wouldn't really help it be more accurate.

Then why is there zero dithering? You can't have current gen games with modern lighting techniques and have no dithering in 16-bit color (or very serious banding issues if the ps2 can't do dithering), it just ain't possible.

Buffer grabs from the PSP usually still show dithering. Oversized shots of Ridge Racer PSP showed the rear window dithering. So, not sure.

 

[Lord Error]

Sure it does. It's been resized to 480x272 with nearest neighbor resampling. It probably won't match how that exact scene would look on a PSP since we don't have such an image from the marketing monkeys, but it's the closets we have in this thread for approximating aliasing at that resolution

I guess I could agree with that, but it really looks worse that the actual game will look for two reasons:

- Using your resampling method, you get jaggier and more inconsistent aliasing than the actual game will have. I don't have to explain you why, I'm sure you know.

- Once compressed JPEG picture is again recompressed by you. Actual game will have no JPEG artifacting.

Actual game will also be moving, which adds to the effect that you don't pay much attention to aliasing details, and the screen refresh bluring can actually help with this a bit.

The fact of the matter is that the truth is somewhere in between your dreadful looking picture, and those nice supersampled pictures. I clearly remember this being the case with the Ridge Racers - we got both direct frame buffer screens full of JPEG artifacting, and some smooth looking supersampled screens, and the game in motion clearly looked much better than those direct buffer screens.

I thought they were a dithered mess? I do admit that since I don't have a PS2 in the house anymore, I can't be too sure.

many PS2 games run in 24bit color, but some run in 16bit and have dithering. I don't rememeber seeing a 16 bit game with no dithering.

On Gamecube, and on PSP, the 16 bit color buffer is extremely common though.

 

[Raistlin]

The large marketing screeny in the beginning of this thread was also not a multiple of 480x272. It was probably an acculative buffer render shot from dev station @ 960x544, then the web site we linked from resized it down to 700x394 via bicubic resampling, adding a shit load of AA into the images.

That is entirely possible ... that does not however make my point any less valid. You cannot accurately resize, add possible AA, then reduce the size without knowing exact details of the resizing algorithm (even then, some information could still be lost). It's not like they just multiplied by some constant, and you divided by the same constant to return it to it's exact original state

My point still stands, your pic is not entirely accurate - and has some obvious artifacts in it. It is more jagged than native res pics, and has some obvious artifacts such as Moire. While it's possible we could see some Moire effect in motion - the initial still obviously did not have it, and it isn't because of AA.

Sure it does. It's been resized to 480x272 with nearest neighbor resampling. It probably won't match how that exact scene would look on a PSP since we don't have such an image from the marketing monkeys, but it's the closets we have in this thread for approximating aliasing at that resolution.

Agreed, it is probably close - but it is not exact and has some obvious issues. I do not agree that the aliasing is normally that noticeable in PSP games - nor should there be other obvious scaling errors.

Also, It's entirely possible that the original pics are native dev pics. During development, it is not unusual to work at different resolutions than the end-product will use.

 

[Raistlin]

Shog,

I would like to point out a bit of hypocrisy in your postings.

In past debates regarding the 1080p TV’s for gaming, you have been quick to point out that one should purchase a 720p TV since the majority of games this gen will be in that resolution. Your reasoning has been that the upconversion to 1080p is not an even multiple, and that such a conversion will yield terrible image quality.

While it is true that the conversion is not an even multiple, I have argued that if properly handled it can look great. The multiple is 1.5, which when used with the proper algorithm can be averaged to produce a very clean representation of the original matieral. My evidence has been in viewing 720p material on native 1080p TV’s - I have viewed this on 2 of Sony’s LCoS sets, and a couple of wobbulated DLP sets – and the results where quite attractive.

In this thread however, you are claiming your pics are an accurate representation of the image on a PSP. The original resolution is 700x394, while your are 480x272. In this case, the multiples are 1.4583333333333333333333333333333 x 1.4485294117647058823529411764706 (Note: these are truncated values, taken as far as the windows calculator will take them).

While one may represent upconversion, and the other downconversion – please explain how 1.5 is terrible but the others are fine? The aspect ratios aren’t even exactly the same.

 

[Shogmaster]

Shog,

I would like to point out a bit of hypocrisy in your postings.

In past debates regarding the 1080p TV’s for gaming, you have been quick to point out that one should purchase a 720p TV since the majority of games this gen will be in that resolution. Your reasoning has been that the upconversion to 1080p is not an even multiple, and that such a conversion will yield terrible image quality.

While it is true that the conversion is not an even multiple, I have argued that if properly handled it can look great. The multiple is 1.5, which when used with the proper algorithm can be averaged to produce a very clean representation of the original matieral. My evidence has been in viewing 720p material on native 1080p TV’s - I have viewed this on 2 of Sony’s LCoS sets, and a couple of wobbulated DLP sets – and the results where quite attractive.

In this thread however, you are claiming your pics are an accurate representation of the image on a PSP. The original resolution is 700x394, while your are 480x272. In this case, the multiples are 1.4583333333333333333333333333333 x 1.4485294117647058823529411764706 (Note: these are truncated values, taken as far as the windows calculator will take them).

While one may represent upconversion, and the other downconversion – please explain how 1.5 is terrible but the others are fine? The aspect ratios aren’t even exactly the same.

I think you've overthought this one. All I'm saying is that the original pics in the thread @ 700x394 which will be nothing like the native 480x272 output of the PSP. The resizing I did was best way to emulate this.

What you are saying about our past discussion about downscaling 1080p to 720p would be relevant if PSP internally renders at 700x394 then downscales for the LCD, but that's not how it works. The PSP's GPU will natively render for 480x272. My resized pic with nearest neighbor downsample is the best representation of that short of us being supplied by the dev with native 480x272 frame buffer pics.

 

[Raistlin]

I know ...

I'm just busting your chops after the denials your downsample introduced any errors.

 

[Shogmaster]

I know ...

I'm just busting your chops after the denials your downsample introduced any errors.

Wha?!? When did I ever say downsampling never introduces errors? It's far less likely than upsampling is what I've always said busta! *grumble grumble*

 

[Raistlin]

I'm referring to the Splinter Cell pic you made

 

[Shogmaster]

I'm referring to the Splinter Cell pic you made

Hey, I never said that my pic was "perfect".... I just said it's far more accurate than the 700x394 pic proabably bicubicly downsampled from even higher res press pics.

 

[Fusebox]

Its like, Splinter Cell PSP is my donut, and there's these two retards fighting over the colour of the icing, but it's mine and you guys are actually drooling and spitting all over it as you argue and I'm seriously considering not wanting my donut anymore.

 

[Shogmaster]

Its like, Splinter Cell PSP is my donut, and there's these two retards fighting over the colour of the icing, but it's mine and you guys are actually drooling and spitting all over it as you argue and I'm seriously considering not wanting my donut anymore.

MISSION ACOMPLISHED! *place FCK pic here*

 

[Fafalada]

If PS2 version of GTAs ran @ 16 bit color with 4MB of eDRAM on the GPU, then the PSP version cannot escape the same fate with 2MB of eDRAM on it's GPU.

It doesn't work that way. PSP memory usage patterns are 'very' different from PS2s - and once you get past the rendering buffers, eDram usage is nothing alike on the two platforms at all.
Also don't forget PSP has 1/2 the res of PS2 (and only 16bit Z) , so it doesn't need as much memory for video buffers to start with.

I was told that most PS2 games don't run in full 24bit color because that will not leave enough room in the eDRAM for the devs to put the amount of textures that they desire in games.

That's wrong on both accounts - any PS2 game with half decent texturing stores most texture data in main memory. And secondly, majority of PS2 games run 24bit color (at least for backbuffers).

Front buffers depend - some games will run 16bitZ, 24bit Front, some will give Z the precision instead and go with 16bit front buffer - downsampling from 24bit backbuffer in the final step after all the rendering is finished gives mostly artifact free results (DC games were great example of this approach).

 

[Shogmaster]

It doesn't work that way. PSP memory usage patterns are 'very' different from PS2s - and once you get past the rendering buffers, eDram usage is nothing alike on the two platforms at all.

Interesting. Is that due to how different the GPUs are? I just thought the parallels of the memory structures made it for similar uses, but I guess GS is nothing like PSP's GPU.

Also don't forget PSP has 1/2 the res of PS2 (and only 16bit Z) , so it doesn't need as much memory for video buffers to start with.

Why only 16bit Z?

That's wrong on both accounts - any PS2 game with half decent texturing stores most texture data in main memory. And secondly, majority of PS2 games run 24bit color (at least for backbuffers).

Front buffers depend - some games will run 16bitZ, 24bit Front, some will give Z the precision instead and go with 16bit front buffer - downsampling from 24bit backbuffer in the final step after all the rendering is finished gives mostly artifact free results (DC games were great example of this approach).

Huh. So would you say majority of the PS2 games are generally "24bit" in output? And how does the GC and XBox compare?

 

[Fafalada]

Interesting. Is that due to how different the GPUs are? I just thought the parallels of the memory structures made it for similar uses, but I guess GS is nothing like PSP's GPU.

Well PSP memory subsystem is an interesting evolution of PS2 approach - I wish I could say more but you know - public board
But suffice to say PSP is less restricted in how various units access memory then PS2 was - and eDram is more general purpose then on any other console arch I've encountered (PS2 is fairly specialized, GC / 360 even more so then PS2).

Why only 16bit Z?

Who knows, probably a design choice. Given the target resolution/screen size, it's not that big of a deal.

Huh. So would you say majority of the PS2 games are generally "24bit" in output? And how does the GC and XBox compare?

Definately yes on PS2. XBox titles afaik are pretty PC in nature - usually uniform bit-depth across all buffers, since memory is more abundant.
GC is a bit of a mystery(we had this debate on B3D recently) - given how many games exhibit blending/dither artifacts, preferred color buffer choice must either be 16bit, or 24bit color with 6:6:6:6 configuration(effectively, 18bit for color + 6bit alpha).
I would be surprised if second choice is the popular one though - destination alpha is not all that important IMO, so personally I would not sacrifice color depth for it without some damn good reason.

 

[Juice]

Its like, Splinter Cell PSP is my donut, and there's these two retards fighting over the colour of the icing, but it's mine and you guys are actually drooling and spitting all over it as you argue and I'm seriously considering not wanting my donut anymore.

Bravo.

You're sassy. I like that.

 

[Shogmaster]

Well PSP memory subsystem is an interesting evolution of PS2 approach - I wish I could say more but you know - public board
But suffice to say PSP is less restricted in how various units access memory then PS2 was - and eDram is more general purpose then on any other console arch I've encountered (PS2 is fairly specialized, GC / 360 even more so then PS2).


Who knows, probably a design choice. Given the target resolution/screen size, it's not that big of a deal.


Definately yes on PS2. XBox titles afaik are pretty PC in nature - usually uniform bit-depth across all buffers, since memory is more abundant.
GC is a bit of a mystery(we had this debate on B3D recently) - given how many games exhibit blending/dither artifacts, preferred color buffer choice must either be 16bit, or 24bit color with 6:6:6:6 configuration(effectively, 18bit for color + 6bit alpha).
I would be surprised if second choice is the popular one though - destination alpha is not all that important IMO, so personally I would not sacrifice color depth for it without some damn good reason.

Oh man, very intersting stuff! Is this basically because GC's eDRAM is only 3MB to PS2's 4MB?

Anyways, thanks for always willing to take time to provide tons of good info!

 

[Raistlin]

Its like, Splinter Cell PSP is my donut, and there's these two retards fighting over the colour of the icing, but it's mine and you guys are actually drooling and spitting all over it as you argue and I'm seriously considering not wanting my donut anymore.

Sweet!!!