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

They definately can and they should - 1ms of GPU time is really cheap, its only 0.5fps loss in 30hz game, its really nothing. I would probably accept game that runs in constant 29fps and has than 30fps and aliased.

It has to be said, looking at the results from FXAAv2 and FXAAv3, and assuming the performance figures we've seen are representative of those you'd get in a real game, it's beginning to look like a no-brainer to use it in any new development projects. I'd be very disappointed to see any of the AAA titles due out at the end of year with weak or no AA.
 
They definitely can and they should - 1ms of GPU time is really cheap, its only 0.5fps loss in 30hz game, its really nothing. I would probably accept game that runs in constant 29fps with AA than 30fps and aliased.

BTW its actually interesting that Sony's GPU MLAA implementation takes 16ms of RSX time, when FXAA only 1ms.

Thanks for clarifying. :smile:

As I understand it for a game to run at a locked 30fps it has to run almost always above that..so a 0.5fps loss is even more less significant isn't it?

It has to be said, looking at the results from FXAAv2 and FXAAv3, and assuming the performance figures we've seen are representative of those you'd get in a real game, it's beginning to look like a no-brainer to use it in any new development projects. I'd be very disappointed to see any of the AAA titles due out at the end of year with weak or no AA.

FXAA sounds like a good and cheap solution that will definitely make any game that has no AA at all look even better and theoritically with such a small penalty there will be small or maybe even no loss at all for the refresh-rate.

It would've been cool to see games like Gears 3, Batman: AC and Uncharted 3 for example having a smoother IQ with FXAA this fall but I wouldn't bet on it. :(
 
Do we have any idea if they plan to use MLAA or FXAA?

AFAIK, they haven't mentioned a thing about it. It's just people on forums saying what they want or "believe" or jumping to conclusions from the tweet from the Sony dev. Judging by the screenshots, I'm sticking with my findings from E3 - edge post-process or "analytical AA". One isn't really going to be able to tell which post-process the dev is using unless they mention it specifically. There are just too many variations to really pinpoint anything.

Edge detection is really going to be hit or miss depending on the colour palette of the game anyway, especially for the cheaper ED methods that don't use depth information. That would explain why some maps look better filtered than others from what I've seen.
 
AFAIK, they haven't mentioned a thing about it. It's just people on forums saying what they want or "believe" or jumping to conclusions from the tweet from the Sony dev. Judging by the screenshots, I'm sticking with my findings from E3 - edge post-process or "analytical AA". One isn't really going to be able to tell which post-process the dev is using unless they mention it specifically. There are just too many variations to really pinpoint anything.

Edge detection is really going to be hit or miss depending on the colour palette of the game anyway, especially for the cheaper ED methods that don't use depth information. That would explain why some maps look better filtered than others from what I've seen.

Really interesting... I only hope the bad implementation will be improved in some area before the release.
 
Resistance 3 appears to be 960x704. There are spots that exhibit 2xAA (& quincunx blur pattern), but it's rather broken, most likely because they're using a deferred renderer. R2 used a variant of light pre-pass along with quincunx IIRC.

2xAA can be seen near the center-top of the boat's windows as well as the top edge of the ammo box). http://www.abload.de/img/20110614164042bsl0.png

Some of the edges have quite a bit of blurring... especially around the view weapon. *shrug*

And no, it doesn't seem to be the 2xMSAA remapping trick either.
 
pretty low res but they keep it at the 16:9 aspect ratio and slightly higher res than 640X720 which was what KZ3 use? Any other downgrade in graphics fidelidy?
 
There's quite a bit changed with the LOD and shadow res is lower. There does seem to be some bloom artefacting.

It's basically 49% of the pixels, so it's just under 640x720 in terms of raw pixels. I would guess they kept the aspect ratio for scaling purposes - it's much more balanced than double-pixel wide steps.

I wonder if they might be able to re-use some of the results from the edge filtering they employ across both eye-views.
 
Al what is the 2xMSAA remapping trick?

The IQ in that R3 screenshot is terrible... sorry that's not a technical remark but my goodness
 
i tried out sonic generations demo and looks like dropping it down to 30hz allowed them to do some really nice stuff, like global illumination and hdr bloom. tho there seems to be noticeable input latency which is really bad for a game like this. 2xmsaa

edit: ps3 version
 
Last edited by a moderator:
Al what is the 2xMSAA remapping trick?

Devs have access to the actual multisamples on RSX, so they can remap the samples as a pixel and write it out to a larger framebuffer than was rendered. It can either be used to properly shade the multisamples ala Killzone 2 or it can be used to fake a higher resolution than was rendered.
 
Back
Top