Digital Foundry Article Technical Discussion Archive [2015]

[Deleted member 11852]

Funny, considering how lousy the PS2 documentation was fabled to be and how devs were left guessing for their launch titles.

I had access to a vast quantity of PS2 hardware documentation prior to launch and can categorically refute that Sony didn't have it available. It was extensive, detailed and well cross-referenced, however we were dealing with Sony's engineering guys so this information probably wasn't what was made available to devs. We also used to deal with Sony on their professional broadcasting equipment and, again, their low level hardware documentation was very good.

I learned more about MPEG encoding from Sony's detailed documentation on their custom encoders than I did from any other source and I lament not having access to their stuff any more.

 

[Phil]

Might you be referring to PDF documents? I'm pretty sure I've seen those documents too. In fact, I think I even attempt to print those hundreds of pages at some point (don't ask me why!). Very detailed stuff on EE and GS. Not sure now if found those documents through B3d or if they were delivered with the Linux kit that I bought for the PS2. I might still have them somewhere on some old backup harddrive from back then.....................

 

[iroboto]

a) yup (1920*540*2)
b) pretty sure it's offline only & only certain missions

edit:

I guess what might be interesting is how much further they could push the resolution on PS4.

i.e. Xbox goes from 720p to 1080p -> 2.25x pixels
PS4 goes from 900p to 1080p -> 1.44x pixels

Extra vertical res would be beneficial.

Just occurred to me that Xbox One was running 1080p@30fps (nearly locked) with a FB3 engine and PBR. For FrostByte engine, they did move the bar of performance on Xbox. Last PBR game(s) at 900p were NFS and DA:I.
I'm not sure if there is some trickery with split screen that overall reduces the load, and I thought that traditionally it's harder to do split screen (I think it is, could be wrong) then, that gives us a clue on what is bottlenecking XBO with FB3 games.

I'm not sure if's compute, as you get double the frame time to work with, but Xbox would have be sitting closer to 27 CUs to get 2.25 the resolution @1080p, and 12 CUs doubled would be at most 24.

 

[Deleted member 86764]

I think you're forgetting about framerate, iroboto? 1080@30 only takes marginally more rendering time than 720@60.

 

[AlNom]

Just occurred to me that Xbox One was running 1080p@30fps (nearly locked) with a FB3 engine and PBR. For FrostByte engine, they did move the bar of performance on Xbox. Last PBR game(s) at 900p were NFS and DA:I.
I'm not sure if there is some trickery with split screen that overall reduces the load, and I thought that traditionally it's harder to do split screen (I think it is, could be wrong) then, that gives us a clue on what is bottlenecking XBO with FB3 games.

There's certainly a heavier CPU/geometry load as the players can be anywhere e.g. visibility, geometry setup, physics/animation, asset streaming calculations (what to keep buffered or not between 2+ players). I'm not sure if each split-screen can be assigned separate threads. I guess you can see the trend in cut-backs when playing any split-screen game - reduced shadows (geometry/draw calls), less grass, generally lower LOD assets.

I'm not sure if's compute, as you get double the frame time to work with, but Xbox would have be sitting closer to 27 CUs to get 2.25 the resolution @1080p, and 12 CUs doubled would be at most 24.

They at least cut back on the post-process AA, so maybe they cut back just enough? I'm not too sure how the shading compares actually.

I think you're forgetting about framerate, iroboto? 1080@30 only takes marginally more rendering time than 720@60.

It was a weird way of comparing I think. He's going for "27" vs "24".

But yes, it'd be easier to just look at it as 2.25/2 = 1.125x pixel load

---------

So... I wonder:

1280*720 ~0.92MP
1920*540 ~1.04MP

Did they split the g-buffer between the 2 memory pools? Probably just coincidence.

edit9000: Like something insane

Render player 1 to esram, yadda yadda shade it, move it out to ddr3 (Player 1 CPU calcs in 16.7ms)
Render player 2 to esram yadda yadda shade it, move it out to ddr3 (Player 2 CPU calcs in 16.7ms)
Duct-tape
???
profit.

 

[iroboto]

I think you're forgetting about framerate, iroboto? 1080@30 only takes marginally more rendering time than 720@60.

There's certainly a heavier CPU/geometry load as the players can be anywhere e.g. visibility, geometry setup, physics/animation, asset streaming calculations (what to keep buffered or not between 2+ players). I'm not sure if each split-screen can be assigned separate threads. I guess you can see the trend in cut-backs when playing any split-screen game - reduced shadows (geometry/draw calls), less grass, generally lower LOD assets.


They at least cut back on the post-process AA, so maybe they cut back just enough? I'm not too sure how the shading compares actually.


It was a weird way of comparing I think. He's going for "27" vs "24".

But yes, it'd be easier to just look at it as 2.25/2 = 1.125x pixel load

---------

So... I wonder:

1280*720 ~1MP
1920*540 ~1MP

Did they split the g-buffer between the 2 memory pools? Probably just coincidence.

Yea there are likely significant cut backs in more than just frame rate to make it work.

It was suggested that because BF was shader heavy, thus, 720 was the most it could do. This could be still correct,because if you tone down enough effects, then eventually it should be able to run at 1080p.

In the video DF noted so far that they turned off/lowered AA to begin with. Nothing else was really mentioned. But that doesn't mean more couldn't have been reduced.
Leads to a honest question though, more effects in theory should have been removed for XBO than for PS4 given the performance specs showcased, but they mentioned they are basically identical. If PS4 had sufficient additional performance to give, AA should have been turned back on.

I think you're forgetting about framerate, iroboto? 1080@30 only takes marginally more rendering time than 720@60.

This is true. It has to render approximately 6912000 (12.5%) more pixels I guess at 1080@30 vs 720@60 if we are looking purely at pixel load.

My train of thought was all over the place, there's another related thread that we spawn off earlier in which we're trying to guess what's taxing both systems to not operate within their expected resolution intervals, in particular, 900 and 720. Since 720 > 1080 seems like a rather large boost, we can leverage some of that to rule out some possible bottlenecks the engine is running into.

The unfortunate scenario is bringing PS4 into this, when I should be looking at it in isolation but.
900@60 all features on 'high' becomes 1080@30 games features with no AA
720@60 all features on 'high' becomes 1080@30 games features with no AA

So I guess the question is ask is, what does AA impact the most?

 

[Deleted member 86764]

I don't think it's the AA causing the issues, I suspect a racer has lower render requirements compared to a large world 64/40 player shooter.

 

[Deleted member 11852]

Might you be referring to PDF documents? I'm pretty sure I've seen those documents too. In fact, I think I even attempt to print those hundreds of pages at some point (don't ask me why!). Very detailed stuff on EE and GS. ..

They mostly were PDFs and scream tics. It you have 3.8Gb then we have the same set.

 

[Deleted member 11852]

Might you be referring to PDF documents? I'm pretty sure I've seen those documents too. In fact, I think I even attempt to print those hundreds of pages at some point (don't ask me why!). Very detailed stuff on EE and GS. ..

They mostly were PDFs and CAD schematics. If you have 3.8Gb of them then we likely have the same set! I'm sure they had to liaise with a whole bunch of technical folks, especially within Government(s).

 

[iroboto]

I don't think it's the AA causing the issues, I suspect a racer has lower render requirements compared to a large world 64/40 player shooter.

the 2 player campaign is smaller in map size this is for sure. There are a variety of modes. But it's mainly 'horde mode' for SW:B. The reduced map size may have a lot to do with it. Though I'm unsure why that would be the case.

 

[AlNom]

the 2 player campaign is smaller in map size this is for sure. There are a variety of modes. But it's mainly 'horde mode' for SW:B. The reduced map size may have a lot to do with it. Though I'm unsure why that would be the case.

Memory?

 

[chris1515]

http://www.eurogamer.net/articles/digitalfoundry-2015-star-wars-battlefront-face-off

Face-Off Star Wars Battlefront

Edit autocomplete shit

 

[Clukos]

Decent AF, in a multiplayer game, on both consoles, at 60 fps. That is all.

 

[Shortbread]

Decent AF, in a multiplayer game, on both consoles, at 60 fps. That is all.

As game engines become more and more demanding/complex on PS4/XB1... will we see more and more 720p games pop up on PS4, or even worst, standard definition games on XB1? Or will we start to see more assets (textures, geometry, lighting, shadowing, etc...) scaled back, mostly matching PC's lower settings? Anyway it goes, these consoles IMHO, will show their age (budget hardware) quicker than previous generations.

 

[London Geezer]

Standard definition? On XB1?? I think that would be pushing it a little.

 

[Shortbread]

Standard definition? On XB1?? I think that would be pushing it a little.

That's why I mentioned the lower settings, for both - before that happens.

 

[Clukos]

Don't think we'll move below 900p/1080p for most releases, there's lots of things we haven't seen yet. Theoretical peak perf/tech for these systems should be during 2016/2017.

 

[iroboto]

As game engines become more and more demanding/complex on PS4/XB1... will we see more and more 720p games pop up on PS4, or even worst, standard definition games on XB1? Or will we start to see more assets (textures, geometry, lighting, shadowing, etc...) scaled back, mostly matching PC's lower settings? Anyway it goes, these consoles IMHO, will show their age (budget hardware) quicker than previous generations.

I don't want to say your comment is hyperbole, it's a possibility but I think you're forgetting about constraints. Fixed hardware yes, but also eventually fixed constraints on both fps and resolution. They both will have their limits.

Once those limits are hit, the engineers will have to find other ways of making it work. SW:BF I don't think is more demanding necessarily than other titles (perhaps at 60fps) but they are remarkably accurate with their values because of their source material which is why in people's heads that SW:BF looks so realistic, and other games may not compare as well.

This probably the difference between 3rd party PBR values and going in and taking your own readings.

I think if last generation was an indication of where things ended (much ahead of the beginning) the same should be true of this generation. Resolution shouldn't matter, SW:BF is indeed quite a ways more beautiful than most Xbox games.

 

[ultragpu]

I don't think devs would want to go so further so much that consoles would have to brutally strip down the res to sub 720p, you know for the sake of visual balance and perhaps a general acceptance from the gamers. I think there's a point when it's good enough is to be reached without crippling anything else and save the mega leap in visual to the next gen instead. After all, we're not developing for gtx 970 based machine this gen unless you wanna neglect the main market and suicide. So I'm saying as long as your game looks decent, fun with plenty of content and what not, keep it at 1080p or 900p for XBone, really no point in going for that extra level tessellation or shadow if 720p is the tradeoff, the pros just don't outweigh the cons.

 

[goonergaz]

Not forgetting they will get more from the hardware as they learn more 'tricks' and ways to optomise.