Digital Foundry Article Technical Discussion Archive [2015]

[London Geezer]

They're the same because when the TV matches the game output, full and limited look the same! It's only when the two don't match that things get ugly. Like we discussed a gazillion times on here.

 

[orangpelupa]

but its a direct-feed screenshot from xbox. not TV that get photographed
EDIT
just checked on PS4 screenshto and they do looks the same although different on TV

 

[Shifty Geezer]

but its a direct-feed screenshot from xbox. not TV that get photographed

The histograms are the same. Neither image is limited RGB which would bottom out at 16 and top out at 240.

 

[London Geezer]

but its a direct-feed screenshot from xbox. not TV that get photographed
EDIT
just checked on PS4 screenshto and they do looks the same although different on TV

Even more reason they would look the same

 

[Globalisateur]

PNGs on XB1 are probably grabbed directly from the framebuffer and so bypass the Xbox hdmi output hardware that allegedly modify ('popify' ) the normal gamma curve meant to be displayed.

 

[AlNom]

PNGs on XB1 are probably grabbed directly from the framebuffer and so bypass the Xbox hdmi output hardware that allegedly modify ('popify' ) the normal gamma curve meant to be displayed.

yeah,that was fixed relatively early in the preview program. Originally it was grabbing the output dictated by the system HDMI range setting.

It would appear the resource impact is either minimal or already consumed with the AF effect off, as adding it back on isn't causing any issues for the games. Although, that said, perhaps these game can patch it in because they have spare bandwidth, and other games will notice more of an impact leading to avoiding the patch?

Well, we are still not seeing 16xAF even on a last gen title. With engines switching to more complex materials, the number of textures associated with a surface ought to increase the associated cost of extra texture fetches/filtering or simply require more developer effort to tune which textures per material property per surface get the AF. ( albedo, specular, normal, roughness, etc)

tl;dr it depends on the game.


----

Side-note: It is also kind of irritating to have better texture filtering than the sun cascade shadowmaps.

 

[Arwin]

Doesn't Resident Evil: Revelations 2 have 16x AF on both consoles?

 

[chris1515]

yeah,that was fixed relatively early in the preview program. Originally it was grabbing the output dictated by the system HDMI range setting.


Well, we are still not seeing 16xAF even on a last gen title. With engines switching to more complex materials, the number of textures associated with a surface ought to increase the associated cost of extra texture fetches/filtering or simply require more developer effort to tune which textures per material property per surface get the AF. ( albedo, specular, normal, roughness, etc)

tl;dr it depends on the game.


----

Side-note: It is also kind of irritating to have better texture filtering than the sun cascade shadowmaps.

TLOUR and Resident evil Revelation 2 have 16 xAF

 

[AlNom]

Doesn't Resident Evil: Revelations 2 have 16x AF on both consoles?

TLOUR and Resident evil Revelation 2 have 16 xAF

I'm talking about DmC.... >_>

Anyways, we don't have a point of comparison for performance impact of AF for TLOU/RER2 since they shipped with it already on, and they'd be limiting the output to 60Hz.

All we know is that the performance was acceptable for shipping with it on. e.g. if it were off, GPU-time could have been even less although it'd be pointless since the target is 16.67ms (and then there's the CPU-bottleneck anyway), but you can't say there's no impact. Or rather, there may be other bottlenecks hiding it in those two cases.

 

[Arwin]

Sorry, misunderstood ... By the way, I wouldn't mind a small training on how to determine how many AF from a screenshot?

 

[orangpelupa]

yeah im curious too. i just know that more AF more clear but how to determind how much is it?

 

[iroboto]

Sorry, misunderstood ... By the way, I wouldn't mind a small training on how to determine how many AF from a screenshot?

Only way I can think of is: You need to recreate the shot and resolution and angle on PC and manually adjust the AF slider until you have the same image. Then you'll know the AF.

 

[Cyan]

I dont see much difference between the two

yeah whats the difference?

PNGs on XB1 are probably grabbed directly from the framebuffer and so bypass the Xbox hdmi output hardware that allegedly modify ('popify' ) the normal gamma curve meant to be displayed.

It seems to be fixed now if the issue ever existed.

Yesterday I played BF4 at Full RGB on my HDTV -which I also calibrated for Full RGB (computer mode)- to test crushed blacks and the game looked great, same with every single game out there. Not a single issue in any game. On my GF's TV there are crushed blacks everywhere if I set the console to Full RGB but it's caused by the TV.

The histograms are the same. Neither image is limited RGB which would bottom out at 16 and top out at 240.

Are you sure about that? I say this 'cos I remember the exact steps.. if anything I was confused as to whether the first screengrab I took was Full RGB or Standard RGB but they shouldn't be the same.

The process was (console set at Full RGB, defaulted by me)...-->FC4 character standing still-->take screenshot at Full RGB (double click Home button +X-->Paused the game-->Go to Settings-->Change display settings to Standard RGB (Colour Depth is set at native 36 bits of my TV)-->go to the game again-->take new screenshot.

The calibration settings are quite different in Standard RGB and Full RGB but there in no crushed blacks

 

[Shifty Geezer]

Are you sure about that?

Yep.

The process was (console set at Full RGB, defaulted by me)...-->FC4 character standing still-->take screenshot at Full RGB (double click Home button +X-->Paused the game-->Go to Settings-->Change display settings to Standard RGB (Colour Depth is set at native 36 bits of my TV)-->go to the game again-->take new screenshot.

The screenshot capture is capturing the frontbuffer before it is being sent to the TV, in pure full-range RGB. When using limited mode, it appears the full-range game output is converted to limited just before leaving the HDMI.

 

[Shortbread]

Digital Foundry: Hands-On with Project Cars

Note:

UPDATE: It's come to our attention that the build of Project Cars used as the basis for this article was not intended by the developer Slightly Mad Studios for technical analysis. We weren't aware of this, which was the unfortunate result of some miscommunication on our part with the game's publisher. It was an honest mistake and it is not our intention to misrepresent the game, so we've unpublished the videos that form the basis of the article.

The game's frame-rate is also a sticking point. Project Cars targets an ambitious 60fps on each platform, but the sheer breadth of options gives players the power to determine whether it hits this mark, or drops closer to 30fps. For example, our first race is on the Dubai Autodrome International circuit, a manic 35-car race with light clouds overhead, camera set to interior cockpit view and no damage physics enabled. Even with this number of AI racers, the game sticks to a 60fps line throughout, and only drops for one stretch on the circuit (to 50fps on Xbox One, and 55fps on PS4).

For moments where the frame-rate goes under, tearing is an issue - especially on sharp turns with lots of camera motion. This is a bigger deal once we push the boat out a bit more; our next race featuring a whopping 44 cars on the Circuit Des 24 Heures du Mans track, with a more taxing chase mode camera and heavy rain in play. The impact is immediate, and PS4 tears constantly with drops to 35fps at the busy starting grid, while Xbox One breaches the high 20s. This improves over the course of the race as cars splinter off into packs, though we never catch a glimpse of the 60fps line.

Finally, we push the Madness Engine one step further by engaging damage physics on each platform, selecting thunderstorms for weather, and this time allowing multiple car types on the Imola circuit. The hit on PS4 and Xbox One performance is huge; Microsoft's console puts in a result that lingers around the 30fps line, with its lowest read-out at 24fps. Meanwhile, PS4 has a rocky time with an output closer to 40fps (though it still wavers up and down), and a minimum rate of 30fps. We have a clear victor during these stress-tests, and it's fair to say that the PS4 edition does feel smoother in practice, though it's still far from ideal.

I'll stick with Drive Club, thank you very much.

Edit:

SMS dev: "It's an old build. I'm disappointed (and that's putting it very mildly) that this has happened, not so much with Eurogamer, but more that they were allowed to do this "analysis" at this stage with an old build."

Hopefully performance will improve...

 

[DrJay24]

" We have a clear victor during these stress-tests, and it's fair to say that the PS4 edition does feel smoother in practice, though it's still far from ideal."

LOL at the wording. Well at least they didn't mention the robustness of XBL.

 

[AlNom]

Hopefully performance will improve...

Maybe they ought to cut back on the transparencies for XO. They're not volumetric, are they? (not including the corpse rays)

 

[Shortbread]

Maybe they ought to cut back on the transparencies for XO. They're not volumetric, are they? (not including the corpse rays)

If possible, maybe an option or two within the game settings (users settings), on scaling back or disabling certain heavy effects/particles that are ROP heavy on the XB1.

 

[AlNom]

Have to also wonder about car model LOD. There were some funky drops with long stretches of road and only cars in the distance.

 

[Shortbread]

Have to also wonder about car model LOD. There were some funky drops with long stretches of road and only cars in the distance.

Wouldn't it have been better to scale back the 40'ish some cars during competition (maybe 25-30), on both consoles? Especially trying to hit 60fps...