Digital Foundry Article Technical Discussion [2021]

great work here Alex.

Good performance from PS5 here overall.

Looks like we have a clean up in aisle 2 however.
Let's try stay to technical folks.

 

yeah it performs as it should so around 5700xt (I reckon couse it would help to bring this cards in benchmarks but no rt support so probably thats why it missing) so nothing new here but there is some rt using here and performance not dying as in Watchdogs (but in watchdogs deffinitly more demanding rt)

 

The scene is very heavy in particles.

 

How one game performs doesn't necessarily tell the whole story of the overall system architecture. We're going to see game performance and IQ (especially from multiplatform third-party developers) on PS5/XBSX that can be reflective of various PC based hardware. One moment performance from XBSX/PS5 can look very similar to an RTX 2060, then the next moment very similar to 2070, or even possibly a 2080. The point being, depending on the game engine, type of game, optimizations and developer, nailing down PS5, XBSX, and XBSS true capabilities this early on in the generation -- especially when their SDK/GDK toolsets are quite new and always evolving -- can be quite challenging on establishing true performance capabilities.

So, there is no need for e-penis wagging and defending one's choice of gaming (i.e., console/PC warring nonsense).

 

Also with COVID having such a big impact on game development at the moment it's likely that many games have been behind on release schedules. Surly that's impacted time for optimizations on PC/XB/PS.

 

Great video as usual! Perhaps if you were to add a comment around how the performance is impacted in this scene on the PC GPU's by going from the higher alpha effect settings down to low, it might give some context around how much extra performance the PS5 is picking up.

 

Yes, that would make sense.

 

Thanks for the video @Dictator.

Can you estimate (or guesstimate) how much performance is being gained by the lower alpha effect resolution? The 2070S is 20% below the PS5, so would it lose closer to 5, 10, 15%?

Also, isn't the PS5 losing average framerate in that scene when compared to PC GPUs, considering the engine is most probably capped at a 60FPS? Even if in residual amounts, whenever the average goes above 50FPS there could be a bunch of frames where the PS5 could have rendered at less than 16.7ms but the cap wouldn't let it.

Perhaps people are ignoring it because they'd have a hard time quantifying it.

 

Actually I think I see my mistake here. Its not about the quality setting but rather the reduced resolution buffer. I'm guessing the PC uses full resolution regardless of quality setting thus making an estimate of the performance impact very difficult.

 

Great video! Kinda shocked by the posts here (and in the purged thread) -- @Dictator did a great job going over numerous settings that could have several MS of impact, and everybody is arguing that it's just a tiny change. Small visual changes can have a huge cost!

(Ps5 certainly still performs great, of course, it's a $500 machine)

 

I see RTX2070/S being used as a general comparison point, while it should be the RX5700XT i think. Anyway, nothing new to the table as said before, PS5 is performing as expected, 2070S as the closest match in CoD, and i think in general around there (it should, seeing the specs).
Best matching for DF going forward in the future should be 6700/XT+ (when its here, if ever).
Also, Turing is well, 2018.... A 3060Ti is outperforming PS5 (correct me if im wrong), so there he should start having 3060/TI when available.

Difficult enough before these factors already. But yes settings and resolution lowering do usually have large performance impacts yet with not that much return in visual fidelity.

 

after looking some benchmarks now I think ps5 in new call of duty perform above 5700xt probably but hard to compare different scene and settings

 

It should. Its currently the closest match from AMD though. A OC'ed 5700XT does align better though.

 

In one week we went from 1060 to 3060 equivalent. At the end of the year it'll reach 3090 power

 

Yeah based on this which is the only decent source I could find for benchmarks in this game (like the rest of the net apparently given the sites performance) it seems that the 2080Ti 36.3% faster than the 5700XT while Alex measures it at only 13.1% faster than the PS5 in his video. On the face of it that does suggest the PS5 is performing a bit better than the 5700XT. However when you consider the lower res Alpha effects which would reduce that PS5 advantage somewhat, as well as the vastly different scenes and settings (there is no RT being used in the 5700XT comparison for a start), it's very difficult to come arrive at any definitive conclusions from this. Nevertheless I would expect the PS5 to be performing a little better than the 5700XT.

 

I hope people keep this in mind if they are trying to use results from the Cold War PS5/PC tests to invalidate Gamers Nexus's video. He could've presented that information better, but I always took that video to be more a statement on what general level those specific games were utilizing the PS5 hardware, not a reflection of some upper limit of PS5's processing capabilities.

GN could've really mentioned something like this in his video and I don't think there'd of been anywhere near the level of pushback it got from some folks, because a lot of other people probably saw the mention and took it as a definitive statement of long-term performance ceiling...and he kind of left phrasing to imply as such which seemingly wasn't for the best.

 

The fact that the 6800XT is only marginal faster than PS5 (less than 30% faster) should tell you all you need to know about the effect of this higher resolution Alpha effect on performance.

 

A little hard to judge with a game that's technically crossgen, but I suspect that as more games on modern, pc focused engines (rather than updates to ps4 focused tech) come out this kind of small gap in xbox's favor will be the norm.