Neverending Upscaling/Resolutions/AA etc Thread #2 *Rules: post: #616 *

@liolio: memory footprint has to be different: fp16 requires 64bit buffer (16bits per channel) while fp10 uses 32bit one in 10-10-10-2 schema (10 bits per color channel, 2 bits for alpha).
References:
http://www.beyond3d.com/content/articles/4/4
http://forum.beyond3d.com/showthread.php?t=21553

 

I'm running both versions of GTA here side by side. The texture artifacting on 360 is off-putting, but its impact seems to be limited to faraway items (for the most part). The lack of AA and the lower resolution is immediately apparent on PS3.

There have been reports of higher frame rates on PS3 but I'm not seeing it at all to be honest - both versions seem to be as good/bad as each other based on my initial observations.

 

thanks guy my memory was wrong so fp10 saves space but have the same computational as fp16 (will ry to remeber properly next time).

 

Has anyone noticed the similarities between GTA4 and the game Wreckless: The Yakuza Missions for the original XBOX. GTA 4 really doesn't appear to be much of a leap graphically from what was capable on the original xbox console with this game. Both had similar looks, heavy filters and fx, a ton going on onscreen at once, but also had blurry low-rez looking graphics. GTA 4 is impressive when you step back and see how realistic it looks, (although blurry) but it still amazes me that this was the best they can do for the current gen consoles. When you see what they're able to achieve with other games: COD4, Drake's Fortune, GOW etc. It just really dissapoints me to think that 640p and 720p is really pushing the limits of the consoles.

 

How would expanding FP10 to FP16 affect performance in a significant way? You expand mantissa with 0s and add/sub something to/from the exponent. That's as cheap as conversion can get. Obviously The other way round it's a little bit trickier as you have to deal with denormalized FP16 values (as long as internal FP16 math is IEEE 754 compliant) and values out of FP10 range. But I think that the overal overhead of the conversion process is close to the gain from smaller set to read/write. Calculations will obviously take as much as in FP16 case, so it seems that there is memory gain, quality loss and no perf impact really. Or am I wrong and packing/unpacking data is THAT expensive?

 

Has the lighting been confirmed to be different on PS3? I found these 360 shots which look to be the same time of day as that PS3 screen.

 

They look similar in these screens (although the PS3 one looks brighter and warmer). IMHO, the city looks better on the PS3 one. Then again, both versions show jaggies here, but the 360 one should have less.

 

I wouldn't say "slightly higher range". It's 8192:1 (with a sign bit too) compared to 256:1 ro RGBA8. Considering that most quality cameras have 8-9 stops of dynamic range, FP10 is plenty for photorealism plus lots of room for overbright if the rest of your renderer is up to the task.

The speed isn't really a downside, either. FP10 renders at full speed for opaque pixels and half speed for blending. FP16 renders at half speed (no blending available). FP10 on Xenos is as fast as RGBA8 on RSX. It's only when you use hardware filtering in a post-process pass (and that isn't always useful anyway) that texturing speed drops to half for that particular part of the renderer.

 

Can you tell what Fable 2 is running at?

http://kotaku.com/assets/images/kotaku/2008/05/F2_house_1280.jpg
http://community.lionhead.com/photos/fable2_photos/images/2907781/original.aspx

 

Just throwing this out here again. If anyone has perfect shots of the PS3 version and they want me to get identical Xbox 360 shots to compare, I can try to capture them.

 

If you play a native 720p game on the PS3 but you are using a 1080p HDMI output to a 1080p TV, is the PS3 scaling or is it the TV? I'm thinking about getting a PS3 soon and want to cover all my bases.

Sorry if this question has been answered before.

 

To be honest, it's a pointless discussion. Looking at the two versions, lighting seems to be VERY similar so long as the in-game clock is the same. Those shots could've been taken at any time during the day cycle. Regardless, it has nothing to do with upscaling, resolution or AA.

The upcoming Eurogamer shots should resolve everything.

 

So long as the game doesn't software-upscale to 1080p, the PS3 leaves it to your display to do the scaling.

Soldier of Fortune Payback is one example of a game that will scale to 1080p if the options is ticked on your XMB, even though it's native 720p. Conflict: Denied Ops is another.

 

I agree. Earlier in the video, I thought that the picture looked brighter and the lighting more natural in the PS3 version. Then the video got to the grenade launcher section and that got reversed. The 360 became more brighter and lighting looked more natural. It basically came down to the time of day of the scene.

 

MERCS 2?
http://www.worthplaying.com/press/281/Mercenaries2_0113.jpg
http://www.worthplaying.com/press/281/Mercenaries2_0112.jpg

 

If you run with automatic setup on the HDMI part, the PS3 will detect all the resolutions your display device supports. It will default to the highest resolution for the XMB.

From there it´s upto the games themselves. For example, GTA4 will use 720p eventhough 1080p is possible, but GTP5 will use 1080p if possible. How ever there is workarounds, for example with GTA4, you setup the HDMI manually and select 1080p as the only possible resolution and GTA4 will scale to 1080p and in GTP5 case select 720p and force it to render to 720p

 

Guys i'm new here, but i'm not new to the tech field, so i'm not a newbie.

I've been comparing both versions of GTA:IV on a XBR960(34" Sony) via a HDMI switch box, and iv'e been doing so for about 20 hrs total.

Iv'e read in this thread that the PS3 version is rendered at 630p, with no FSAA, but with 8XAF and FP16.

Iv'e read the 360 version is rendered in 720p, with 2XFSAA, with 2XAF, and FP10.

Are both of these statements correct?

About the dithering on the PS3 version, i realize this gives more precise color accuracy, but why does the IQ of the PS3 version seem more blurry(less cleaner) then the 360 version?

For instance look at these pics:

http://static.videogamer.com/videogamer/images/pub/misc/vsapril29large3.jpg

http://static.videogamer.com/videogamer/images/pub/misc/vsapril29large4.jpg

In the first pic, Niko looks more detailed on his face(360 pic) but appears to have more color definition in both his face and hand.

Is the color difference because of the use of FP16?

Also, in the 2nd pic, the forground on the 360 version is more crisp, but the IQ as a whole is not well filtered but it's less blurry then the PS3.

I agree the Ps3's IQ looks more uniform(foreground to background), but why must this come at a cost to the IQ looking more blurry in comparsion to the 360's version?

Is this a case where the PS3 is using too much dithering, so that the IQ gets too blurry?

What are you guys thought on the PS3's smoother..yet more blurry looking IQ compared to the 360's?

Also, would you prefer a more uniform IQ(PS3) and give up the cleaner textures in the forground(360 version)?

Actually the textures for the 360 version seem a bit larger/cleaner in the foreground, compared to the PS3's.

Seeing both these engine's in action, reminds me of the Voodoo days when it was a choice between 4XFSAA with blurry textures or 2XFSAA with more detailed textures, except in this case it seems like a matter of more AF'ing on the PS3's part.

Please add your thoughts on which version you prefer(graphics wise only) and why so?

Thanks for reading this,

Phil