Digital Foundry Article Technical Discussion Archive [2016 - 2017]

Status
Not open for further replies.
It doesn't have to be though. Ori and the Blind Forest runs fantastically well and is gorgeous to boot on Unity. I have a theory that since Unity is generally (not always) used by smaller developers with smaller budgets that games based on Unity just don't get as much optimisation and polish as games based on say UE which is generally (not always) used by developers with a larger budget.

If you look at UE games by smaller developers with limited budget you'll also see quite inconsistent and sometimes extremely bad performance.
I think I agree with the point about Unity devs. I think they're more about getting it out there. UE is hardly fabulous though. At least it wasn't on PS3, even AAA titles. It just wasn't a well optimised engine on that hardware. I suppose if there's any platform it's weaker on now, it'll likely be XB1 as PS4 shares a lot with PC. The engine may be overlooking some PS4 optimisation options though via HSA etc.
 
I think I agree with the point about Unity devs. I think they're more about getting it out there. UE is hardly fabulous though. At least it wasn't on PS3, even AAA titles. It just wasn't a well optimised engine on that hardware. I suppose if there's any platform it's weaker on now, it'll likely be XB1 as PS4 shares a lot with PC. The engine may be overlooking some PS4 optimisation options though via HSA etc.

Agreed, but UE has a better reputation with regards to performance and features than Unity. Whether deserved or not.

Regards,
SB
 
DF is actually all over the map here. There older tests show no difference between PS4 and XBO internal drive speeds. And also, no major difference in loading times (in most cases) between an SSD and non SSD for XBO.

http://www.eurogamer.net/articles/digitalfoundry-2014-xbox-one-external-hard-drive-upgrade-guide

How could things have changed so massively?

Now their new article shows huge differences with no explanation.

This Fallout 4 HDD test seems to fall somewhere in the middle of all of the above http://www.eurogamer.net/articles/d...d-drives-boost-xbox-one-fallout-4-performance

Further, if you read comments, some will claim a 7200 RPM mechanical drive should be/is as fast/faster as a 5400 RPM Hybrid SSD (while much cheaper per GB). And DF's old 2014 article seems to bear this out. DF makes no mention of what speed their tested hybrid is (7200 or 5400) in their 2016 article (to be fair, Seagate seems to keep it hidden as well)

I'm confused, and it's annoying.

Also, as the Eurogamer comments mention, it would be really nice if somebody tested the load times with the XBO Elite console, which has a hybrid drive, somebody claimed it could be fastest because it uses internal Sata (3?)?

The trim issue in these comments being unclear also makes me wonder about purchasing an SSD or Hybrid for XBO use.
 
A hybrid drive shouldn't be concerned about TRIM, as the flash portion is managed by the drive itself as a cache, rather than listing itself as an SSD that an OS would send TRIM commands to.
That might be one thing that mitigates some of the uncertainty of SSDs for the consoles, although I am curious if there is any provision for disk defragmentation that a hybrid drive's platters could still benefit from.
 
It could also be that Microsoft and/or Sony have sourced different drives at different points in their life. That could lead to a wide variety of performance characteristics.

None of those articles states what condition the internal stock drive was at either. Was the drive full? Nearly full? Had it been in use prior to the review? Meaning was the game tested installed on the outer tracks or the inner tracks? That could have a significant impact on load times as well.

I'd also be curious to see whether Microsoft bothered to include (or will include) Trim on the XBO. Especially as it converges more and more with Windows. Did they foresee users adding external SSD drives? Or did they just leave it off?

Regards,
SB
 
Performance Analysis: Hitman
The final game runs better than the PS4 beta - and there are some unexpected surprises.

http://www.eurogamer.net/articles/digitalfoundry-2016-performance-analysis-hitman

On Xbox One, however, the boost in performance that we observed in Paris is more interesting. The busiest areas on the ship saw sustained drops into the low 20s, and we expected similar results in the more crowded Paris map. Yet the frame-rate manages to remain steady in the mid-30s instead. What exactly is responsible for the drops in the training mission and why are we seeing improved performance in a more complex map? It's not entirely clear, but we're reminded of Assassin's Creed Unity and the slight performance advantage evident on Xbox One in busy sequences.

If I were to hazard a guess, the ship and most outdoor areas are likely more GPU than CPU limited, hence the PS4 leads as expected. In the Paris area they mention far larger crowds of people (AI actors). It's quite likely then that it's more CPU than GPU limited in that case (more AI actors, more CPU required for AI and pathing). And the XBO lead is within the bounds of the CPU lead it has in hardware.

I'd expect anywhere with a large enough crowd of AI actors will potentially give the XBO that small 3-5 FPS advantage. But otherwise PS4 should lead in FPS everywhere else.

Regards,
SB
 
Is my PS4 dying or something?

The Division hovers around 20 something fps on indoor areas with only ONE enemy.

It randomly have 0 fps for a few seconds while traversing the world.

But DF says it relatively stable
 
Connection issues possibly?

hmm, from what i see, The Division prefer to rubberband and blink other network stuff rather than yourself and your world. (like... enemy suddenly in front of you, enemy moving back and forth. you yourself always able to animate just fine and shoot just fine despite the damage my come 5 seconds later)
 
http://www.eurogamer.net/articles/digitalfoundry-2016-performance-analysis-the-division

Resolution:
In terms of resolution, we're on familiar ground. PS4 features a solid native 1080p presentation throughout, while on Xbox One dynamic resolution scaling is in effect. Indoor areas and less demanding scenes see the game output natively in 1080p - just like PS4 - but as we move into outdoor areas and more detailed locations, we see the pixel density take a small hit, with 1792x1008 and 1728x972 framebuffers appearing. The knock-on effect is that street signs and fine details on distant objects appear a little softer on Microsoft's console - but it's a minor quibble, and most of the time the two versions look extremely close during gameplay. The difference between the two is more visibly felt when disabling the chromatic aberration filter, due to the increase in sharpness and clarity doing so provides.

PS4 Performance:
In terms of performance, there are no surprises, with both versions handing in similar levels of stability to what we saw in the beta. Aside from a rare dropped frame, PS4 delivers a stable experience throughout the game, with sequences set in detailed environments and shootouts remaining smooth and consistent at all times. The use of object and camera blur in combination with depth of field also helps, allowing the 30fps update to appear a little smoother that it actually is on both consoles. For example, 30Hz-related judder during slow camera pans is less noticeable than it was in Far Cry 4, and this gives the experience a pretty fluid and cinematic look.

XB1 Performance:
Moving over to Xbox One, and for the most part Microsoft's machine manages similar metrics to PS4. Combat scenes and traversing more open locations practically never impact on frame-rate. Occasionally, we see the appearance of a single dropped frame accompanied by a tear or two, but nothing that stands out when playing. Curiously though, there are times when the engine briefly struggles to keep up with the rendering load. Cut-scenes exhibit stutter from time to time, caused by drops in frame-rate and tearing, as do sequences where we move from some indoor areas to outside. We suspect that resolution may play a part in causing these hitches to occur. Xbox One holds to a native 1080p when indoors, and it's possible that the extra pixel workload is impacting performance in these scenes.
 
It doesn't have to be though. Ori and the Blind Forest runs fantastically well and is gorgeous to boot on Unity.

Ori looks to be a really simple 2D game. I think the GC based design of Unity (and their inability to update the Mono core) does not really fit any real time system. But I think GCs are crap anyway ("You can put the GC on a seperate core!" Of course that core has to compete for memory access with all the rest of the jobs you do.)
 
If I were to hazard a guess, the ship and most outdoor areas are likely more GPU than CPU limited, hence the PS4 leads as expected. In the Paris area they mention far larger crowds of people (AI actors). It's quite likely then that it's more CPU than GPU limited in that case (more AI actors, more CPU required for AI and pathing). And the XBO lead is within the bounds of the CPU lead it has in hardware.

I'd expect anywhere with a large enough crowd of AI actors will potentially give the XBO that small 3-5 FPS advantage. But otherwise PS4 should lead in FPS everywhere else.

Regards,
SB

I believe the CPU-lead is wishful thinking at best, which has been debunked every single time and to this day only manifested in 1-3 badly optimised Ubisoft marketing deal games, being patched out in the weeks after anyway. The reason for the difference in frames in this case is because PS4 renders more or better shadows(/AO?) in that scene:

t5LXIS7.png

^^^^ the pole things have shadows on PS4

BMr9DSj.png


^^^ here xbox shadows contain less samples and are more washed out or something.

There are many more differences but I don't really have time for this
 
It doesn't have to be though. Ori and the Blind Forest runs fantastically well and is gorgeous to boot on Unity. I have a theory that since Unity is generally (not always) used by smaller developers with smaller budgets that games based on Unity just don't get as much optimisation and polish as games based on say UE which is generally (not always) used by developers with a larger budget.
Well here's the thing, Ori is a 2D game and while it is fairly complex with it's metroid style levels, it is still akin to those Ubiart games in terms of tech. And yet it cannot run at perfect 60FPS as it still has drops.

Imo a developer would get more out of UE3 than they would out of any other engine right now, Unity is ofcourse easy to use and great for people who are not technically minded, but ironically it also requires a lot of experience and technical knowledge in order to find a workaround to its performance issues.
 
I believe the CPU-lead is wishful thinking at best, which has been debunked every single time and to this day only manifested in 1-3 badly optimised Ubisoft marketing deal games, being patched out in the weeks after anyway. The reason for the difference in frames in this case is because PS4 renders more or better shadows(/AO?) in that scene:

t5LXIS7.png

^^^^ the pole things have shadows on PS4

BMr9DSj.png


^^^ here xbox shadows contain less samples and are more washed out or something.

There are many more differences but I don't really have time for this
looks like gamma differences for me. also the angle is not 100% the same, so it might be the viewing-angle
PS4 image is general daker than the xbox's.
 
Status
Not open for further replies.
Back
Top