LAIR Thread - * Rules: post #469

Status
Not open for further replies.
As long as nAo doesn't confirm and nobody knows, its all speculation what "true" res it runs at. That means as far as anyone can tell, its truly 1080p

Right I know, the devs have said this game will run in native 1080p so I will definitely give them the benefit of the doubt, even if these images may seem to indicate otherwise.
 
These new shots are 720p, with gobs of AA (probably rendered at 5120x2880 and downsampled). They look good though, so I'd personally take 720p with 4xAA instead of the previous shots. There's no pixel doubling thing going on here.

As for the shots themselves, the artwork is very nice. Looks great!
 
It does look decent. But at the same time, it still looks like garbage. The soldier's better be smarter this time because last time they looked like a bunch of idiots.

I think that technically, with 1080p and all the effects they're putting in, its impressive, but how they put it together is going to be a real problem. I wonder how it'll look in motion.
 
It looks impressive. I'm not really excited about the gameplay. But might buy it anyway just to check out the graphics.

Julian said in an interview with the 1UP show that the game was at 90%. We should probably see it by summer.
 
If this is the case, then is it truly 1080p? Or as long as horizontal resolution is 1080p (ignoring vertical res) we can call it 1080p?
After resolving the data polygon edges still have horizontal definition of 1920 pixels, in theory only internal pixels of polygons would be in 960 format as MSAA doesn't take extra samples within polygon.
In most cases we do not see single samples in low contras areas anyway, so this might be very nice way to gain some extra out of the machine.
 
Wow, that makes an awful lot of sense and explains why only the near horizontal edges look the way they do. And for completeness, here's a zoomed in (600%) pic showing the pixel pairs:
lair5.png

I don't think that particular shot adequatly confirms Mintmaters theory. There is still too much unique data there. And actually very few pixels with actual pixel doubling. If you look closely even the pixels that do look doubled are actually slightly different.

If their "post processing filter" was able to hide the majority of doubled pixels even at 600% magnification thats a pretty damn micraculous filter! But I think it's a simpler conclusion that they must be doing something a little more clever than rendering at half resolution.

There is something odd about the way their edges are rendered though. I can't quite place it. But they must be empolying some kind of trick with their MSAA buffers. The edges almost seem like they quite "don't belong" with the non edge pixels. There is a dithering. Is it possible they are rendering just the edges in a render target that is 16bit or less?
 
I don't think that particular shot adequatly confirms Mintmaters theory. There is still too much unique data there. And actually very few pixels with actual pixel doubling. If you look closely even the pixels that do look doubled are actually slightly different.

If their "post processing filter" was able to hide the majority of doubled pixels even at 600% magnification thats a pretty damn micraculous filter! But I think it's a simpler conclusion that they must be doing something a little more clever than rendering at half resolution.

There is something odd about the way their edges are rendered though. I can't quite place it. But they must be empolying some kind of trick with their MSAA buffers. The edges almost seem like they quite "don't belong" with the non edge pixels. There is a dithering. Is it possible they are rendering just the edges in a render target that is 16bit or less?

Interesting, I may need to fix the contrast on my monitor(!), but still - interesting. What if they were using an unresolved rotated grid 2xSSAA method? That way there would be additional texel samples limiting, but not eliminating, the pixel doubling. I'm just thinking out loud here, so what would be the cost of that kind of AA?
 
These new shots are 720p, with gobs of AA (probably rendered at 5120x2880 and downsampled). They look good though, so I'd personally take 720p with 4xAA instead of the previous shots. There's no pixel doubling thing going on here.


Actually, if you take a look at shot one (not only in this one), the wing on the left upper side looks just about 2xAA (at most 4xAA), so something else might be going on here...

These screens look a little bit washed out to me, maybe some post processing filter was applied to the afterwards?
 
I don't think that particular shot adequatly confirms Mintmaters theory. There is still too much unique data there. And actually very few pixels with actual pixel doubling. If you look closely even the pixels that do look doubled are actually slightly different.

If their "post processing filter" was able to hide the majority of doubled pixels even at 600% magnification thats a pretty damn micraculous filter! But I think it's a simpler conclusion that they must be doing something a little more clever than rendering at half resolution.
Nothing miraculous needed - A simple two pixel moving average on the 1920x1080 image will get rid of the pixel doubling. But I don't think that's it, because we wouldn't see any pixel pairs then. I think it's alpha blending (which is per sample), geometry edges (of the non-silhoutte variety) with centroid sampling, and subtle HDR post-processing (brighter spots lose the doubling). Nowhere in the picture will you find single pixel wide texture detail, but single pixel tall details are there.

Mmmkay didn't pick the best spot. In the second pic look at the wing near location (1180, 800) and also the architecture below it.

There is something odd about the way their edges are rendered though. I can't quite place it. But they must be empolying some kind of trick with their MSAA buffers. The edges almost seem like they quite "don't belong" with the non edge pixels. There is a dithering. Is it possible they are rendering just the edges in a render target that is 16bit or less?
The 2xMSAA sample pattern Nvidia uses (diagonal) will give you precisely the effect we're seeing when following the procedure I mentioned. It also explains why it only affects horizontal edges. I'm pretty sure I'm right on this one, and I can draw you a diagram to explain what causes the edge pattern if you want.
 
Actually, if you take a look at shot one (not only in this one), the wing on the left upper side looks just about 2xAA (at most 4xAA), so something else might be going on here...
Looking closer, it's probably 3x3 supersampling instead of the 4x4 I suggested earlier, but it's not 2xAA or 4xAA. Here can you see how 2x2 ordered grid gives you a single pixel inbetween the longer colour steps on near horizontal/vertical surfaces? That's what we have, except the grid is bigger.

IMO the rough-looking edges on the wing in some other shots are due to alpha testing. We're limited by texture resolution there as opposed to screen resolution.

Except for the dragon wings, every other edge is extremely clean. I don't think you can get this with only 4xAA (though I'm generally satisfied with 4xAA).
 
http://media.ps3.ign.com/media/761/761161/vid_1913754.html

Interview with President of Factor 5 on LAIR.

Nearly 15 minutes long.

It sounds better than I expected. I guess because they keep showing that bridge level over and over and over it seems like the game is pretty shallow. But he makes it sound like it has some depth to it.

The game still looks like it still has some frame rate issues though. Hopefully they can lock that up for the retail version. I guess even Gears had frame rate issues until almost right before it went retail. So there is still hope left.
 
Will they ever show anything new?

I am waiting to see improvements on framerate, AI and colors. They hurt my eyes and AI on the ground soldiers looked stupid. Fireballs also needed lots of work back then
 
I use to be excited about this game. Now I am not. The game looks like crap.

Could I be disappointed with Killzone 2 as well?
 
Meh... Where is "Oprah" the Giant Sea Serpent ? :p

Found these shots in GAF. They are... em... different ?

Close-up of Oprah finally: http://www.ultimagame.com/play_station_3/foto_f154241.jpg

Others:
http://www.ultimagame.com/play_station_3/foto_f154250.jpg
http://www.ultimagame.com/play_station_3/foto_f154242.jpg
http://www.ultimagame.com/play_station_3/foto_f154253.jpg
http://www.ultimagame.com/play_station_3/foto_f154252.jpg
http://www.ultimagame.com/play_station_3/foto_f154251.jpg

This game starts to remind me of old Japanese monster flicks (Go-ji-la !!). :D

P.S. The pictures are 1280 x 720... in case you're on dial up.

EDIT: More here: http://www.neogaf.com/forum/showthread.php?p=5647428#post5647428
 
Status
Not open for further replies.
Back
Top