Digital Foundry Article Technical Discussion Archive [2015]

Status
Not open for further replies.
IRC - there is no AA on 750 Ti and framerate is much worse. I don't know about other settings

The consoles use AA so the 750Ti will be using it too. It would be a ridiculous oversight and completely invalidate the article to miss such an obvious setting when deliberately aiming for visual parity.

As far as performance goes, the 750Ti is arguably a little faster in the video but both systems seem to trade blows.
 
DF are hiding some facts, if you omit important, objective facts you are misleading others.

Their wording is as much as harmful to the X1 as it can be. They say about the Xbox One version that they don't favour it because the framerate fluctuates between 20 and 40 fps. They do Know that's not true. During gameplay, which is what actually matters, the game Never goes below 30 fps.

Lies, lies and more lies.

Cinematics might be an entirely different thing, but they run fine for the most part. Plus it'd be nice if they talked about how texture streaming works in the game.

OK @Cyan. Look at that video at around 2:50 (they are comparing XB1 with 750ti cards), a couple of things:

- The XB1 game regularly drops at 29 or 28 on the fps meter in this short city scene.
- Even when the fps meter stays >= 30, there are constant frame time fluctuations above 33ms meaning the game if capped would each time drops under 30fps in a very similar way of the PS4 game.


But don't worry, even DF completely miss the point and believe a fps counter above or equals to 30fps automatically means a potentially judder-free 30fps capped game, they imply that the capped XB1 game would run much better than the PS4 when it's in fact totally false. Here they are harmful to the PS4 not the XB1 game.

Both are v-synced, but the issue for the latter case is that its frame-rate goes between 20-40fps, meaning there's no consistency to when your screen refreshes with a new frame.In theory, this issue is fixable on Xbox One with a patch, as theoretically it rarely dips below the 30fps mark

Now in the light of those new frame-time concepts, watch again the PS4 vs XB1 performance video by comparing only the frame-time values above 33ms, ignoring the rest, you'll see that in fact both games, if theoretically both capped, would run very similarly during basically the whole video. Obviously ignore also the cutscenes moments where the PS4 is capped at 20fps (in order to effectively hide judder, and it works believe me, a cutscene capped at 20fps is better than a ~25fps fluctuating one).

 
Last edited:
The consoles use AA so the 750Ti will be using it too. It would be a ridiculous oversight and completely invalidate the article to miss such an obvious setting when deliberately aiming for visual parity.

As far as performance goes, the 750Ti is arguably a little faster in the video but both systems seem to trade blows.

They gave exact settings in youtube video description and there's nothing about the AA. It could be in the "postprocess effect" category, but IMO it should be listed sparately. Also DF mentioned that foliage and shadows rendering visibility is superb on PC ultra settings, much better than on consoles. Point is that I haven't seen any pop-up on PS4 neither- visibility is great.
 
I just noticed that the Anisotropic Filtering is in fact better on the PS4 version.

O2wb.png


Xbox One gives the worst results here, and in Novigrad City's main plaza, PS4 hits a mid-point in filtering quality between the other two platforms....PC runs at 16x AF here, while consoles run at a lower quality setting, with Xbox One trailing behind PS4 in clarity.

Judging by the fact that PC is at a locked 16xAF, that in this screen PS4 AF is just a notch under PC 16x, and that PS4 AF level is between PC and XB1 in many areas, we can assume that PS4 often uses 8xAF vs 4xAF for XB1.

Which goes well with my actual observations. Indeed I several times thought that I was looking at 8xAF textures (when I erroneously thought I dunno why that the consoles version had 4xAF).

EDIT: Plz can a moderator move this post on the DF thread? Sorry about this.
 
mmm. If you are right @AlNets I have never seen such discrepancy in textures clarity just because of a 900p / 1080p difference which really should look like this example, where the difference is more subtle:

h3wb.png


Also I have just noticed a sharpen filter in effect in the PC pics that could again improve perceived sharpness of the PC version of textures and particularly vegetation.


But let's forget all of this guys because I am going to talk about a much more interesting aspect of the game that no one talked about before: the motion blur used on PS4. :smile2:

There are some interesting dynamic motion blur settings on PS4 and I am generally pretty happy about CD project decisions:

- In 99% of my playtime in the game, there is no motion blur applied (at all) in the near vicinity of the character. The motion blur is applied only at a certain distance (like a ~100m from the character).
- Until now the 2 places where I found a the usual unbearably strong motion blur even close to the camera is at the beginning in your house and a bit later in a castle when they change your clothes and you have to meet with a very important person. I suppose all future 'resting' moments in castles meeting with people will have motion blur applied too. But that's Ok cause we don't have to stay long in those places.

But apparently the lack of motion blur during 99% of the time close to the character doesn't bother a lot of people....I mean, did people even notice it? :rolleyes:

They could even deactivate the motion blur applied on the background and no one would see any difference...Ah yes, I think I have read some people complaining about the motion blur used in the console version, well, retrospectively I think they complained when there was one applied close to the character in the tutorial or in the castle after, because that one is some heavy stuff indeed... :mrgreen:
 
Last edited:
I just noticed that the Anisotropic Filtering is in fact better on the PS4 version.

O2wb.png

I don't believe this has nothing to do with AF being lower on XB1... I think this has more to do with LOD. Or a more aggressive form of it. Given that the XB1 has a lower resolution, it just exacerbates the LOD issue even more. Pretty much it's like looking at my PC edition of GTA V when compared to my PS4 edition... things at a distance just seems less vivid or detailed (clarity) on the PS4.
 
I don't believe this has nothing to do with AF being lower on XB1... I think this has more to do with LOD. Or a more aggressive form of it. Given that the XB1 has a lower resolution, it just exacerbates the LOD issue even more. Pretty much it's like looking at my PC edition of GTA V when compared to my PS4 edition... things at a distance just seems less vivid or detailed (clarity) on the PS4.
I think its just the lower image resolution blurring the details out
 
I think its just the lower image resolution blurring the details out
Still, I find the blurring on the XB1 game abnormal in some areas compared to others 900p / 1080p games.

Could the chromatic aberration applied on 900p on XB1 and at 1080p on PS4 explain this?
 
Still, I find the blurring on the XB1 game abnormal in some areas compared to others 900p / 1080p games.

Could the chromatic aberration applied on 900p on XB1 and at 1080p on PS4 explain this?
I believe it is explained here
Paired with CD Project's custom post-process anti-aliasing mode, edges are at least treated thoroughly. In effect, this gets us a blended half-tone on each stair-step that mimics a 2x MSAA pass, while a temporal component addresses any shimmer on trees in motion. But based on direct image comparisons, it seems Xbox One struggles to reconcile its upscaled resolution with this AA method, and as a result there's a big tail-off in clarity the further into the distance you look. Curiously, PC offers only a single toggle to enable this mode with no alternative, and Nvidia's own control panel settings for MSAA are incompatible with its engine.
 
On the Bone, most +33 ms frame times are follow or preceded (or both) by a 16 ms frame. This is a pacing issue and not evidence that - if capped at 30 hz - the Bone version would necessarily see the same types of frame rate drops as the PS4.

It seems pretty clear that XB1 game is running better the majority of the time. Except it's not capped, so it's actually running worse most of the time.
 
@Nesh So in fact what I took for 2x hardware AA was in fact a post AA mimicking a 2xMSAA... OK, my bad. But like on PS4, the vegetation on PC does not seem to be post '2xAA' antialiased at all. With or without the post AA, edges on trees and grass are identically aliased.

http://international.download.nvidi...wild-hunt-anti-aliasing-001-off-1920x1080.png

http://international.download.nvidi...-wild-hunt-anti-aliasing-001-on-1920x1080.png
The AA method seems to produce ugly results when its combined with the upscaled image of the XB1. Other than that it doesnt seem to do much of difference, probably due to the higher density of the pixels at 1080p. I suspect it tries to blur out larger sections in 900p because pixels are larger and the algorithm treats them as aliasing.
Just out of curiosity, can you capture the following?
one image at 1080p with AA
one image at 1080p without AA
one image at 900p with AA
one image at 900p without AA

Assuming of course that the PC is using the same method. 900p with AA might probably look blurrier than the 900p image without AA
 
@Nesh So in fact what I took for 2x hardware AA was in fact a post AA mimicking a 2xMSAA... OK, my bad.
It seems odd that a strictly post-process pass would turn 1x stairsteps into 2x stairsteps. This goes back to the temporal AA issue; if it's using TAA, the reprojection is probably what provides the apparent 2x sampling. Like, if the TAA isn't providing 2x sampling, what is it doing?
 
What is odd about temporal method to provide 2xSSAA?
Nothing. That's totally normal, and I intended for my post to imply as much. My confusion is with DF's wording which seems to imply that a spatial* post-process method is producing a 2x appearance, with "a temporal component" as a separate entity.

*I probably should have said spatial, in my earlier post I was using "post process" to strictly refer to such methods, i.e. FXAA and SMAA.
 
They gave exact settings in youtube video description and there's nothing about the AA. It could be in the "postprocess effect" category, but IMO it should be listed sparately.

Yep AA is in the postprocessing section of the graphics menu.

"All post effects on except vignetting."

Also DF mentioned that foliage and shadows rendering visibility is superb on PC ultra settings, much better than on consoles. Point is that I haven't seen any pop-up on PS4 neither- visibility is great.

Take a look at the first comparison tool picture on the left. The difference is obvious.
 
Status
Not open for further replies.
Back
Top