Image Quality and Framebuffer Speculations for WIP/alpha/beta/E3 games *Read the first post*

It's a pretty generic term to describe any sort of screen-space analysis and performing another arbitrary form of anti-aliasing. It'll mostly be bound by math power vs quality.
 
I'd still like more steps to confirm 1152, but it's definitely not 1280.

If they are running a deferred renderer with 12 byte G-buffer (4b Z, 8b data), the maximum resolution that would fit in edram without tiling is 1200x720. Since render targets need to be in units of 80 pixels horizontally, with a fixed 720p there are only a couple options really: 1120 and 1200.
 
It's pretty significant. It's about as bad as the Ghostbusters port on PS3 early on.

clarity wise those 360 shots were saved with a low compression level. the xbox 360 doesn't really blur like that at sub hd. the Hana scalar chip inside of the 360 helps on providing clarity on it's output. jagged levels is what you look at, and things in the distance would be pretty hard to see with a lot of jaggedness going on, things like thin objects.

360
http://img717.imageshack.us/img717/9669/1111360.png
ps3
http://img27.imageshack.us/img27/2109/1111ps3.png

it's pretty hard to see a lot of jaggedness in the closeups.
http://img85.imageshack.us/img85/3084/finalfantasyxiii360015.jpg
http://img3.imageshack.us/img3/5857/finalfantasyxiii360021.jpg

edit, each of the shots are saved at 1267x713 so it's obvious the images were down scaled.
 
Last edited by a moderator:
If they are running a deferred renderer with 12 byte G-buffer (4b Z, 8b data), the maximum resolution that would fit in edram without tiling is 1200x720. Since render targets need to be in units of 80 pixels horizontally, with a fixed 720p there are only a couple options really: 1120 and 1200.

Yeah, there's the 80x16 tile size sans AA*... but even Halo 3 was, by their admission, 1152x640. With 1120 being 7/8 and 1200 being 15/16, I just didn't have enough steps to do a successive check to make sure the ratio holds (only had about 20-22). I'd prefer at least three successive checks to make sure. e.g. 21/24 vs 30/32 vs 27/30 vs 36/40 and so on. Clearly, one can discount certain ratios between those.

*
MSDN said:
Render targets must be padded to certain dimensions based on the multisample mode. The padding required for 32-bits-per-pixel formats is 80×16 for 1× antialiasing, 80×8 for 2× antialiasing, and 40×8 for 4× antialiasing.
http://msdn.microsoft.com/en-us/library/bb447675.aspx (for anyone else who didn't know).

So yeah, padding to 1200 would seem likely if they stick with 1152.:) Wasted area in the last tile then.



On a side note, I'm almost curious enough about NG2's 585 vertical res. Alhough one would need 73 steps per 90 pixels to confirm the difference with 13/16 ratio (if it's actually 584 instead).
 
clarity wise those 360 shots were saved with a low compression level. the xbox 360 doesn't really blur like that at sub hd. the Hana scalar chip inside of the 360 helps on providing clarity on it's output. jagged levels is what you look at, and things in the distance would be pretty hard to see with a lot of jaggedness going on, things like thin objects.

HANA isn't a scaler chip. It converts the framebuffer to HDMI, component, transcodes 1080p into 1080i and handles standard def formats. In fact, there is no scaler chip at all. Xenos does it.

As the HUD is native 720p (hence a final "720p" image in the framebuffer), there's no evidence at all that the hardware scaler is being used for the scaling.

We aren't going to know how good or bad it is until we see it motion really or see some much better shots.

A bit surprised to see these shots lurking about as there's a worldwide embargo in place.
 
HANA isn't a scaler chip. It converts the framebuffer to HDMI, component, transcodes 1080p into 1080i and handles standard def formats. In fact, there is no scaler chip at all. Xenos does it.

As the HUD is native 720p (hence a final "720p" image in the framebuffer), there's no evidence at all that the hardware scaler is being used for the scaling.

hmmmm, interesting bit of info there. it seems somewhat impressive, considering that every resolution that 360 supports the game that it loads up never changes in clarity or native pixels when lowered or raised through all the resolutions. (which is different with ps3 where it varies from game to game.)

We aren't going to know how good or bad it is until we see it motion really or see some much better shots.

A bit surprised to see these shots lurking about as there's a worldwide embargo in place.

i actually didn't want to get into the details of ff13, but it was the fact that a bunch of shots of 360's copy of ff13 turned up. some of these shots are blatantly sub hd while others are above that and are cleaner. some of these shots have very poor ground textures while others shots are just perfectly fine.

a couple people made blogs proclaiming they've played both versions and are reporting huge chunks of both detail and geometry missing in the 360 version, while others made complete coverages on their experience saying it was very good and even put in a couple words for it saying the in-game graphics were very crisp.

(last paragraph)
http://www.thelostgamer.com/2010/02/12/final-fantasy-xiii-xbox-360-hands-on-preview/

my guess is, there are some who played the beta (and yes there was a playable beta of 360's copy, IGN covered it.) while some have played a newer build. (possibly a review copy.)
 
Last edited by a moderator:
hmmmm, interesting bit of info there. it seems somewhat impressive, considering that every resolution that 360 supports the game that it loads up never changes in clarity or native pixels when lowered or raised through all the resolutions. (which is different with ps3 where it varies from game to game.)

PS3 employs software solutions to scale from start to finish. Xbox 360 uses Xenos to scale the finished 720p frame up to whatever resolution the dashboard is set to. However, you cannot say for sure that the process of scaling the sub-HD buffer up to 720p (before the HUD is overlaid) is done using the same hardware process.

I don't quite understand this idea of "sub HD" and "above that". If the framebuffer is scaling up *any* resolution up to 1280x720, then it is sub HD. Are you suggesting that some of the shots you've seen of the 360 version are full 720p like the PS3 game?

BTW all the major sites have final retail code.
 
@Copsandrappers:

The only issue about which there is any reasonable question is that one ground texture which looked worse in the one screen, but for all we know it was taken before the full res texture could finish loading and "pop in". The rendering resolution looks pretty cut and tried, though. 720p hud on a 576p scaled image does not sound like another SE PR mixup. That's just what the game looks like on the 360. On the one hand we have MazingerDUDE on NeoGAF counting pixels and literally showing his math and on the other hand you have an odd, isolated report from GameSpot or some blog where someone looking at god knows what display setup and saying, "oh I think the 360 version seems sharper." I know what seems more credible to me.
 
Last edited by a moderator:
It's a pretty generic term to describe any sort of screen-space analysis and performing another arbitrary form of anti-aliasing. It'll mostly be bound by math power vs quality.

Actually, analytical anti-aliasing usually means that for each pixel a primitive touches, you analytically determine the the color contribution, i.e. integrate color over the area of the pixel(*). At least that's how I've always seen it used.

Now I'm pretty sure that's not something you can just implement on a rasterizer, for a couple of reasons.

(* and not by using a cheap 1-16 sample pseudo Monte Carlo integration, i.e. SSAA ;))
 
It's pretty significant. It's about as bad as the Ghostbusters port on PS3 early on.

i have no clue how you came to that conclusion. look at the near bottom of the screen both ground texture are pixelated. ghostbusters had textures significantly lower resolution these are probably the same exact assets... the only difference you are seeing is the resolution at which the game is being rendered.
 
It's pretty significant. It's about as bad as the Ghostbusters port on PS3 early on.

You must be joking right? That ground texture doesn't look significantly different between the two versions. The leg on the other hand shows what might be a downgrade although it's hard to tell with the lighting and angle being different.

Regards,
SB
 
Yeah... none of those videos are 720p, unless I'm missing something... Anyways, there's little point trying to prove reality again.
 
Back
Top