Digital Foundry Article Technical Discussion [2020]

Status
Not open for further replies.
He says the dynamic resolution drops lower on XSX.

xsx dynamic lowest measured resolution by nxgamer was 1440p on ps5 1620p

I think he said he noticed the XSX go that low once while the lowest he saw on the PS5 was 1620p. He didn't say whether that was in the exact same location but he did say he expects the PS5 does go down to 1440p too and potentially even lower for both consoles.

Sure we need to wait for more framerate tests for AC Valhalla. Based on what NX Gamer showed XSX seems to have an incredibly minor advantage in cutscenes (like 1% or so) and PS5 has a slight advantage in gameplay.

Yes but that's with the frame rate locked to 60fps. So you're only altering the average there with any drops below 60fps. If the frame rate were unlocked it might paint a much different picture.

I do enjoy and appreciate NXGamers video's but I'm sure we'll get a slightly more scientific analysis from Alex which should give us a better idea. I'm just really looking forward to seeing how it compared next to the PC as on the face of it it looks like the console versions - the PS5 in particular are punching well above their weight in this game (or rather the PC is punching below). This is the best comparison point I could find:

https://www.game-debate.com/news/29...c-performance-report-graphics-card-benchmarks

It shows an RTX 2080 averaging just 61.5fps at 1440p on the High preset. That's in the in game benchmark which is reasonably heavy vs gameplay but not unrepresentatively so and gameplay can get even worse performance in places. So unless the consoles are running at largely medium setting (which I doubt) then they seem to be performing very, very well here. For the record the 2080Ti seems to perform around 22% faster than the 2080 using this as a reference point which would put the 2080Ti at the same settings at around 75fps average.

Also from the game debate review we see the 2080 achieves 39.8fps average at 4K. So splitting the difference gives a rough 1800p performance of about 50fps on the 2080 or a little over 60fps on the 2080Ti - at High settings.

So it seems to me that either the consoles are running at largely medium settings (unlikely), they are running closer to 1440p in areas that more closely represent the PC benchmark and so far the console reviews have focused on less demanding areas for the pixel counts (maybe possible, but stretching) or the PS5 is performing as well as or better than a 2080Ti!

One final possibility may be that one or both consoles are using some form of reconstruction and the pixel counts so far aren't picking that up.
 
just watched the NX video, PS5 does not outperform XsX ? they're almost identical, even the load times o_O

From what I've seen of Valhalla on both nextgen consoles, performance seem incredible variable on both. Perhaps this is why we're seeing assessments with different conclusions.
 
VG Tech just made a video comparing Valhalla on both machines. The framerate is actually very similar between both versions. But still, the framerate seems to be a tiny bit better on PS5 judging by the stats.

impressive perf from ps5, does rx5700 (not xt) run this game on pc with similar performance ? ;) On techpowerup rx5700xt on avarage in 1440p 53 fps so rx5700 for sure even worse (but don't know what console settings are)
 
So what's the verdict so far? PS5 punching above its weight? Could it be the clock speeds that make these cross gen games run faster than we would expect? Jury is still out of course but what's the feeling here?
Did anyone expect much of a difference in launch titles when the consoles are so closely spec'd? I think we'll have to wait a bit before real technical comparisons can plot out a trend.
 
DF Article: https://www.eurogamer.net/articles/...dogs-legion-console-ray-tracing-vs-nvidia-rtx

Watch Dogs: Legion - how does console ray tracing compare to Nvidia RTX?
Cutbacks, compromises and optimisations - but they work.

Hardware accelerated ray tracing is now firmly established in triple-A titles on PC and with the transition to next generation Xbox and PlayStation hardware, consoles can now join the party. Watch Dogs: Legion is the first DXR-enabled title we've seen that also features ray tracing on consoles - which raises the question: to what extent can the new machines match up to existing PC hardware? What kind of compromises are required to bring RT to consoles - and what happens when we retrofit those cuts to the PC version? Let's just say that the answers are illuminating.

Hardware RT in Watch Dogs Legion is used for reflections, replacing the standard system based on a combination of screen-space reflections and cube maps. To make that clearer, cube maps are essentially non-dynamic 'probes', capturing environment detail and baking them into a texture wrapped around a cube - and typically thousands of them are generated in any given scene, drawn upon when needed by the game engine. Screen-space reflections capture what's on-screen, mapping that information into reflective surfaces like glass walls and puddles. This combination is often convincing enough but is rarely satisfying in reflecting moving objects (like people milling around the city) or in showing detail that isn't currently on-screen. Ray tracing is expensive but solves all of these issues - and Marvel's Spider-Man: Miles Morales is the best console rendition we've seen to date, up to a point.

...
 
So what's the verdict so far? PS5 punching above its weight? Could it be the clock speeds that make these cross gen games run faster than we would expect? Jury is still out of course but what's the feeling here?
There are some things about XSX that are not yet out there. I can’t comment as it’s not mine to say. But I will say if you want to know what’s happening that you need to look at more than just XSX vs PS5. In particular if you compare its performance vs other GPUs around it’s class from a front end perspective you’ll see a trend. Ie compare it’s performance against X1X and 6800 for instance. The FH4 DF video should be telling for instance.

I would say that XSX is having its ampere moment a touch here; perhaps some compromises were made to the front end to create more space for ALU while limiting silicon costs. And games currently designed heavily around a traditional rendering pipeline may / will likely not perform as well as its marketed number.

This wouldn’t be a problem I think necessarily except tying Series S which had little compute pretty much locked them into a more traditional
pipeline. a lot of re-optimization may be required for SeriesX beyond the norm.
 
Ray tracing analysis for WDL on Series X / S and PC


Superb video/article. Been waiting for this one with baited breath! I love the fact that @Dictator was able to match settings so exactly using the config files and I'm really surprised that the PC is using higher res textures than the next gen consoles. I don't think anyone saw that coming!

I must admit I'm pretty shocked at the performance level. With one quality analysis we've just gone from the XSX looking like it might be faster than an RTX3090 (given the 3090 falls below 30fps at max settings 4k) to the XSX being a little bit less performant than an RTX 2060S. Settings matter people!

I hope this isn't indicative of the RT performance we'll see in the RX6xxx series though.
 
There are some things about XSX that are not yet out there. I can’t comment as it’s not mine to say. But I will say if you want to know what’s happening that you need to look at more than just XSX vs PS5. In particular if you compare its performance vs other GPUs around it’s class from a front end perspective you’ll see a trend. Ie compare it’s performance against X1X and 6800 for instance. The FH4 DF video should be telling for instance.

I would say that XSX is having its ampere moment a touch here; perhaps some compromises were made to the front end to create more space for ALU while limiting silicon costs. And games currently designed heavily around a traditional rendering pipeline may / will likely not perform as well as its marketed number.

This wouldn’t be a problem I think necessarily except tying Series S which had little compute pretty much locked them into a more traditional
pipeline. a lot of re-optimization may be required for SeriesX beyond the norm.

Yes I speculated something similar on some thread or other recently. The XSX has only 81% the front end throughput of the PS5 and something like 66% of the RX 6800 (non XT) so that may well be holding it back in "last gen" games that are shader light.
 
So what's the verdict so far? PS5 punching above its weight? Could it be the clock speeds that make these cross gen games run faster than we would expect? Jury is still out of course but what's the feeling here?

My feeling is that people just looked at the paper specs and declared XBSX the faster console without proof and now everyone is in shock and awe when they see actual software running on both systems. Now you have the classic looking for scapegoats crap as to why that is when they believed the opposite for so long. "TFs" is the new "Bits" after all :rolleyes:
 
Status
Not open for further replies.
Back
Top