Digital Foundry Article Technical Discussion Archive [2015]

Status
Not open for further replies.
So, the grass causes slowdowns, but isn't foliage normally fill-rate limited, where PS4 has a huge advantage with twice as many ROPs? Unless the ROPs are bandwidth starved? If the flashlight is not casting shadows on foliage, what would be the obvious bottleneck for a non-shadow-casting light?

Edit: There is also fog in those scenes.
 
Last edited:
So, the grass causes slowdowns, but isn't foliage normally fill-rate limited, where PS4 has a huge advantage with twice as many ROPs? Unless the ROPs are bandwidth starved? If the flashlight is not casting shadows on foliage, what would be the obvious bottleneck for a non-shadow-casting light?
.

And GTAV showed PS4 fillrate handled a lot more foliage than Xbone so that's not the only factor here.
 
And GTAV showed PS4 fillrate handled a lot more foliage than Xbone so that's not the only factor here.

Well it's pretty much got to be one of two areas if we're looking at hardware...

CPU or something about ESRAM/bandwidth. And a third I suppose, main memory latency?

The X1 SDK seems to be improving leaps and bounds, so past games may not correlate to current experience. Then again, it'd be hard to think Capcom would be in a better position there than Rockstar's tremendous resources, even with a couple 5 months time advantage.
 
Well it's pretty much got to be one of two areas if we're looking at hardware...

CPU or something about ESRAM/bandwidth. And a third I suppose, main memory latency?

The X1 SDK seems to be improving leaps and bounds, so past games may not correlate to current experience. Then again, it'd be hard to think Capcom would be in a better position there than Rockstar's tremendous resources, even with a couple 5 months time advantage.

And you forget another factor software and maybe a GNMX game like The Crew for example. The game is using the old MT Framework. They use the PC version as the base version the port from PC to Xbox One is easy not so easy from PC to GNM PS4. They probably used GNMX and The Crew develop new code because some part of GNMX were they were not up to the performance they wanted... It was better than standard GNMX game but nowhere near GNM game performance...
 
After playing it for a couple of hours on a 970, I can say that the game is a turd port, no doubt about it. Micro stutter and random frame rate dip with this kind of graphic is unacceptable. Ps4 probably suffer from the same kind of "quality control" from capcom. The game had a last minute delay by a week close to release, who knows how it used to run on xbone and ms sent their eng to help optimized the game on their game.
 
Well I just beat RER2E1 on the Xone and it ran pretty smooth the entire time except in the corner of this one cabin towards the end of the game where the frame rate tanked for no reason. The game has some decent lighting in some moments but the textures vary from horrible to decent. The game was pretty fun though. I will most likely purchase the other episodes.
 
So, the grass causes slowdowns, but isn't foliage normally fill-rate limited, where PS4 has a huge advantage with twice as many ROPs? Unless the ROPs are bandwidth starved? If the flashlight is not casting shadows on foliage, what would be the obvious bottleneck for a non-shadow-casting light?

Edit: There is also fog in those scenes.
just maybe, it is just a memory-contention / cpu-thing. we know this is an old game ported to newer systems and the PS4 memory system can have a really big contention-problem if the memory access is not optimized. The PS3 had its own memory system für cpu and gpu. The xbox 360 had it's own small memory for the gpu which lowered the contention problem. Now the PS4 may suffer because it has only one pool of memory and the access-patterns may not be optimized for this. Also the CPU may something have to do with this. I don't think it has something to do with the 0,5 more CPU-Cores, more about the 9% more frequency. After all it is still "just" an old engine that has been ported so I don't think it is good at multithreading. I would guess that the game uses 3 CPU-cores max.
 
While DF seems to think this could be GPU related judging by their comments, shouldn't we be expecting some X1 multiplatforms to run a tad better due to CPU? X1 has a hardware advantage of 9% in CPU clock, and according to those leaked documents an extra .5-.8 of a core on top of that (presuming ps4 is still at 6 cores available). That's an ~18-23% advantage in theory (although I dont remember the timing when MS released the extra partial CPU core vs this games development time frame). That's half as much as the PS4's much ballyhooed 40% flops advantage.

Wait, am I missing something here? you're comparing apples to oranges aren't you? 40% GPU is more than double 20% CPU in raw numbers isn't it?
 
Rangers is saying that if a 40% advantage in one area is phenomenal, a 20% advantage in another area must, be at the least, significant. It can't be overlooked as a CPU advantage too small to impact the machines.
 
Wait, am I missing something here? you're comparing apples to oranges aren't you? 40% GPU is more than double 20% CPU in raw numbers isn't it?
you just can't compare those numbers. The GPU may have nothing to do because the CPU can not send enough to work on. So the GPU may have just nothing to do. <20% more CPU can be a really big difference, but ~40% more GPU can also be a big difference. Just depends on the scenario.
 
After playing it for a couple of hours on a 970, I can say that the game is a turd port, no doubt about it. Micro stutter and random frame rate dip with this kind of graphic is unacceptable. Ps4 probably suffer from the same kind of "quality control" from capcom. The game had a last minute delay by a week close to release, who knows how it used to run on xbone and ms sent their eng to help optimized the game on their game.

Are you getting framerate drops with a 970??
 
you just can't compare those numbers. The GPU may have nothing to do because the CPU can not send enough to work on. So the GPU may have just nothing to do. <20% more CPU can be a really big difference, but ~40% more GPU can also be a big difference. Just depends on the scenario.

well yeah - I get that...my point was apples to oranges anyway
 
It happens every time. People love to correlate hardware strengths/weaknesses with software performance on a 1:1 basis. Its poor post hoc logic.

Yes. And what about others recent games analysed by Digital Foundry? Those don't count? like Evolve which seems to be running better on PS4, wasn't this game supposed to be CPU heavy? Because that what people were saying when it was running better on XB1 during the beta! But no more of it when it now runs better on PS4 in the final version?

What we can say is that with the code to hand, the Xbox One version seems to hitch like this more often than its PS4 equivalent.

Also Battlefield Hardline, also a supposed CPU heavy game that also runs better on PS4:

demonstrates plenty of performance hitches and dips on both consoles, more noticeable on Xbox One

Both games even running at a higher res on PS4 being recent games on well known engines. So why no PS4 trumps XB1 in those cases? No PS4 CPU 20% more powerful than XB1 CPU here?

My point is when PS4 "trumps" XB1 in the majority of cases, even in CPU heavy games, nobody will ever assume anything about PS4 CPU but when XB1 games run better than PS4 games (in a handful of games only) then the Digital Foundry CPU narrative (that they meticulously crafted in previous articles and analysis) suddenly explain everything.
 
I guess what they're trying to say is that this crappy port is a demonstration that Xbox One is a more capable machine and that the CPU up clock is really showing its powers now.


Ya ok. Whatever.

I personally wouldn't even bother reading doing a technical analysis on something that's that far behind in terms of technical prowess. What's next? DF analysis on Pac man on both consoles?
 
... when XB1 games run better than PS4 games (in a handful of games only) then the Digital Foundry CPU narrative (that they meticulously crafted in previous articles and analysis) suddenly explain everything.

I don't really understand why this game could be CPU limited, but what else can it be?
 
GNMX. At least imo. If your company can't code in GNM the two platforms will be close together. With Dx11.3 coming you may actually start seeing X1 take a commanding lead especially in cpu limited scenarios with companies who don't have that low level coding talent on tap.
 
What's next? DF analysis on Pac man on both consoles?

What is wrong with analyzing and reporting on this apparent fubar? They would not have known otherwise. Folks would have been curious about a Resident Evil entry anyway.
 
Status
Not open for further replies.
Back
Top