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

Because hdd read speeds are slower than cpu/mem combo, raw (uncompressed) images would actually slow load times in contrast to any reasonable compression technique.

But HDDs on next gen are theorically twice faster than on old gen. The bluray drives are even more faster and have gzip decompression hardware. And if they load all the available levels in memory when you first start the game during the splash screen, I think it's easily feasible.

And they could still use lossless compression textures instead of compressed textures:

The power behind the Xbox One and the PS4 allows us to display the game’s textures without any compression on these consoles

It's obvious they are talking about lossless compression on next gen so the textures on old gen were compressed. I guess we'll see on the Digital Foundry article the difference.
 
I think we all agree about everything, except some of you guys are content with article's totally stupid use of words "compression" and "uncompressed". That's why I said "misleading".
 
I thought I could share with you those Infamous (PS4) pics I took (a while ago) from a Youtube 1080p video.

Remember that it's from a YouTube compressed stream and that I specifically chose these enclosed (apparently uninteresting) locations not really to show off the whole big city around the character.

The second picture is in fact the most interesting.

Infamous_1080p_1.png

Infamous_1080p_3.png
 
Looks horizontally upscaled from one pic. I count 13 pixels for 10 steps, or 980 wide, so I guess 960x720 in real terms if someone performed a more accurate investigation. Zero AA.
 
Looks horizontally upscaled from one pic. I count 13 pixels for 10 steps, or 980 wide, so I guess 960x720 in real terms if someone performed a more accurate investigation. Zero AA.

Thanks a lot.

One guy at gaf said that the game might use a dynamic framebuffer but I don't see any changes in IQ in those shots in terms of clarity. I have played the 360 demo and while it was soft as well it didn't look that bad plus the jaggies were not that apparent from what I remember so I thought that the softness was due to a possible post-process AA solution.

I'll download the PS3 demo to see how it looks on my set.
 
One guy at gaf said that the game might use a dynamic framebuffer but I don't see any changes in IQ in those shots in terms of clarity.

The game does use lower res transparencies, and there can be a lot of that in the opening room, so it could just be that.
e.g. http://i.imgur.com/pw69w25.jpg everything "behind" the smoke is unaffected while what's-his-face, who is in front of the smoke has derped edges.

http://i.imgur.com/sIPhtPi.jpg left wall is unaffected (behind the smoke) while everything from the middle to the right (in front of or intersecting the smoke) is derped.


----
http://i.imgur.com/T5jxqn6.jpg -> 1024 (16/20)
http://i.imgur.com/NpBRe6t.jpg -> 1024 (32/40)
http://i.imgur.com/Dp1XeFO.jpg -> 1024 (16/20)
http://i.imgur.com/PAGNqCn.jpg -> 1024 (16/20)
 
Still, that's like going from a max quality jpg to a png. Strikes me as very pointless. "Now our game takes up 4 times as much space on your tiny-ass HDD for no discernibly visible reason. Enjoy!"
Rayman looks pin sharp on every platfrom already. (the 3ds version of the first game was awful, though)

Should you be more worried about the amount space uncompressed textures take up in the video ram?

Can a gpu sample directly from a png compressed texture or does it have to be decompressed? If not, doesn't that mean that the textures take up twice the space in memory than it does on disk?

Seems pretty backwards to me that the space requirement of your textures grows as the amount of memory available shrinks the closer you get to the gpu/cpu.
 
They seem to have picked an odd resolution for it to run at: 1408x792

Based on some extended edge-counts from the game's stark tutorial section, our best guess right now is that 1408x792 is pretty close to Respawn's chosen rendering resolution. The overall effect is pretty similar to 720p overall though, and the implementation of an overly sharp filter across the entire image suggests that the Xbox One hardware scaler is used to blow up the image to 1080p. We're not exactly impressed by that and, we suspect, neither was DICE - hence the move on Battlefield 4 from Microsoft's scaler to a bespoke software solution between the preview and final code we played.

Update 13/2/14 10:06 GMT: Respawn has confirmed our '792p' pixel count, and has indicated that a resolution increase is on the cards for the final game. Meanwhile, The Verge reports that Respawn is working with Microsoft to improve the Xbox One's hardware scaler.

Source

*Seems like they are hoping to hit 900p without cutting too much out*
 
They seem to have picked an odd resolution for it to run at: 1408x792



Source

*Seems like they are hoping to hit 900p without cutting too much out*

What I find interesting is Eurogamer said last year after analyzing the (E3?) video, is they were running 1600x900 and there were lots of frame drops into the 40s.

I think the resolution bump may be marketing speak since there are frame drops and tearing in the game currently ~month from launch.
 
What I find interesting is Eurogamer said last year after analyzing the (E3?) video, is they were running 1600x900 and there were lots of frame drops into the 40s.

hm... this is the quote I found from their article:

The jaggies are a curious point in particular; even judged by the high quality feed we have availability direct from the Microsoft E3 conference there's more sub-pixel shimmering and rough edges than any other game on show, which suggests this may not be a full-blown 1080p title in its current state.
Might have been folks assuming 900p as a (then) likely explanation.
The perfect 60fps in the analysis implies to me it was likely PC footage at the time though. *shrug* [strike]Maybe it was just no-AA for the E3 feed.[/strike]

The later trailer in August was 1080p 4xAA, IIRC.

edit:

The gamersyde E3 footage looks 1080p. Maybe the shimmering was just broken AA. *shrug*
 
hm... this is the quote I found from their article:

Might have been folks assuming 900p as a (then) likely explanation.
The perfect 60fps in the analysis implies to me it was likely PC footage at the time though. *shrug* [strike]Maybe it was just no-AA for the E3 feed.[/strike]

The later trailer in August was 1080p 4xAA, IIRC.

edit:

The gamersyde E3 footage looks 1080p. Maybe the shimmering was just broken AA. *shrug*

You missed this part of their write up (my bold):

While 60fps is inevitably a slam-dunk for the Source engine if the active hardware here is indeed a PC, it's curious to see Respawn's shorter two-minute trailer, which precedes the gameplay demo, struggling to reach 50fps in spots and showing clearly the devastating effects of double-buffer v-sync when frame rendering runs over the ultra-tight 16.67ms budget. This trailer cuts together a montage of clips to show off more intensive, cinematic scenes alongside gameplay from new levels, where those featuring heavy effects work and broader draw distances visibly suffer from stutter. With one close-up of a Titan descending to just 40fps, it's plausible that some of these scenes are compiled using a much earlier, unoptimised build.

Also, the article says the game was 7v7 at the time. So they made a cut there, too.
 
Ah... interesting. The gamersyde feed of the trailer is also 1080p (with healthy amounts of AA). Wonder if DF only had a lower quality feed at the time. I still don't think we've seen any XO footage until the press event last week.

But anyways...
 
No PS4 vs XB1 comparison either. Hmmm

It says there at the bottom:
PS3/360: 720 (scaled internally) 30fps
XBOne: 720 60fps
PS4: 1080 60fps

'Scaled internally' is code for sub-720 rendering.
PS3 #1
pic_night_ps_ps3.jpg

#2
pic_daytime_ps_ps3.jpg

PS4 #1
pic_night_ps_ps4.jpg

#2
pic_daytime_ps_ps4.jpg

360 #1
pic_night_xbox_360.jpg

XBOne #1
pic_night_xbox_one.jpg


I can only link 6 pics, but you can see the last 2 pics of the 360/XBOne comparison in the original link.
 
Back
Top