Digital Foundry Article Technical Discussion Archive [2016 - 2017]

Nice to see a great port for once.
And thanks for looking to do that test as well, could be interesting results.
Cheers

 

It seems alpha effects are responsible for the frame drops. The only time I see drops below 60 are with a full grid of AI producing spray/dust clouds. Notice how in that video, the framerate doesn't drop until everyone starts moving. Plus in that race you have the blue light/smoke markers along the edges.

 

Pro talk for the Pro thread.

 

Interesting...Battlefield 1 uses dynamic resolution on Xbox One...typically holds at 900p:

 

Quite an improvement for them, where typically XBO holds at 720p for FB games.

 

Pretty sure BF4, BF:Hardline, Battlefront were all fixed 720p on Xbox One correct?

Curious if PS4 is still using 900p fixed...if so it means DX12 has probably helped Xbox get close to PS4 performance....as I believe this is the first Frostbyte title with it implemented?

 

Battlefront was already a DX12 game... the famerate is also much worse than in Battlefront. Also, let's see how the dynamic resolution works in the multiplayer mode.

 

Didn't realize Battlefront was DX12 game...was it implemented at launch?

Framerate of BF1 is much worse than Battlefront? It seems pretty good in that Digital Foundry vid.

 

still waiting for PS4 results. But everything else is correct. Not sure when dx12 was implemented. Shouldn't benefit XBO unless they rewrote the whole pipeline to be dx12 based. Which I know they didn't since they have a dx11 variant still of BF1.

 

Battlefront used DX12 on XO only, on PC it is DX11.

 

Well, i think that we should wait for the DF analysis.

But so far :





Anyway, i'm suprised that the game doesn't run at 1080p on PS4 considering the resolution on XB1...

 

@mpg1
Not sure if you were around when we were looking into specific XBO customizations. But moving from 720p to 900p for Xbox is pretty massive, which meant that they needed to leverage a feature that DICE wasn't using before.

So it's entirely possible that DICE are using that feature now.
tldr; it could be the move to DX12 and leveraging ExecuteIndirect Feature combined with the customized XBO Command Processor.

link to that discussion here on B3D:
#1804

ExecuteIndirect as well as some other features on DX12 have been known to improve performance, EI seems to be the big one in particular (my understanding on certain code seeing upwards to 15-20% improvement IIRC, will need to dive back into the DX12 thread). We have no known understanding of the degree of improvement on performance the microcode on the XBO Command processor has. But DICE seems very enthusiastic about it with their presentation. (perhaps another 10-15% on top of that and you are getting close to meeting 900p.)

It's perhaps, the only _true_ secret sauce that XBO has over any other hardware out there that we know of today. I expect that feature to make it back into AMD graphics cards, and of course we should see it for Scorpio.

I'm not sure if it makes sense to make a hoopla about it, it's also the most hidden and guarded secret of xbox, and it took nearly forever to find out about it. Luckily DICE made a presentation about EI or we would never have known it existed in the first place. We knew they made customizations to the command processor, we just didn't know what for. We've gotten really used to just seeing a 30% performance advantage on PS4 which is almost always translated into resolution; I suspect the resolution for BF1 will be 1080p on PS4.

 

Yes my argument was that because they developed a dx12 version for PC it may have helped Xbox One version. It isn't just a matter if they used DX12 like with battlefront...it's whether the the developers have more experience using it/developed more techniques and fully implemented it.

 

yea its possible. Unfortunately a lot of our discussion is trying to determine where XBO APIS began and ended, that's often been a point of discussion. So we were never sure if XBO always had DX12 features, or some DX12 features. In any event, my full understanding is that they deprecated DX11.X Fast semantics on XBO SDK for DX12.X for Xbox One. They still have support for some form of general DX11.3X on XBO for broader audiences.

 

That being said it looks as though the PS4 has also received a bump:



Dynamic res between 900p and 95% of 1080p.


It's really interesting that they have seemingly cranked more out of these consoles. Obviously there was/is still room for optimization of Frostbyte 3 engine on these consoles.

 

If PS4 is capable of coming close to 1080p then I am really curious as to how PS4 Pro version will turn out. It's possible we will get 1800p+ checkerboard or some equivalent technique. I'm guessing this game will look REALLY nice on PS4 Pro with a 4K TV...sort of tempting since this is out in only a few weeks..

 

According to other analysis both versións are closer in terms of resolutión. When it comes to campaign the results are as follows:

Gamingbolt:
PS4: 1430 X 800 to 1800 X 1020
XBOX ONE: 1330 X 750 to 1800 X 1020

VGtech:
Ps4: 1422x795 and 1807x1016
XBOX ONE: 1320x744 to 1780x1000

Quite similar

 

Those numbers don't mean much without knowing the average resolution.

Also, this is only a campaign analysis. I believe others tested MP as well.
But the general takeaway is that ps4 seems to generally hold a higher resolution and is noticeably sharper. XB1 version seems to perform very slightly better though.

 

Gamingbolt... seriously ? This website always steals the data from DF, VGtech or NXGamer.

VGtech is reliable though. Maybe a more recent patch in the DF analysis ?

 

And this confirms that there is no magic about the improvement of the XB1 version... it's just basic optimization... Watch Dog 2 runs at 1080p on PS4, yet the PS4 doesn't have DX12 or any "special sauce".

But once again, we have a higher resolution in BF1 but worse performances compared to Battlefront... key word = compromises.