Digital Foundry Article Technical Discussion [2021]

Status
Not open for further replies.
DF Written Article @ https://www.eurogamer.net/articles/...-evil-village-demo-tested-on-all-playstations

Resident Evil Village demo tried and tested on PlayStation 5, PS4 and Pro
What's really going on with the '45fps' modes?

Last weekend, Capcom unleashed the first in a series of time-limited demos for Resident Evil Village. First up, we got to see the Village demo for PS4, Pro and PlayStation 5, with the promise of a Castle demo next week on the same systems, followed by a multi-platform release for both segments the week after. We took a look at the initial demo to see how Capcom plans to scale the game across the generations, what the ray tracing looks like, and how performance stacks up. Going into this, the developer told us to expect 45fps on certain consoles and game modes - confused messaging we can clear up having gone hands-on.

So, one demo, three systems and a total of five different play variations to get through - but that proved relatively easy. The truth is, there isn't much to the Village demo. There's a small exploration segment to begin with, but we quickly segue into what is effectively a series of cutscenes linked together with the most minimal of player interactions. The demo is effective in demonstrating the power and versatility of the RE Engine, it gives us strong indications of how beautiful the final game will be but as an actual sampler for the experience, it's minimal.

...
really love that super speed load time.
a return to 60fps + load times will likely be the characteristics that will define this generation of consoles (imo)
 
Last edited:
on the one hand its great on the other hope we will see also some graphics jump
yea I think that will just come in time - and I get everyone is impatient to see what true next gen games will look like. I mean I totally understand the feeling of wanting to see something new. I mean, that's why I bought a PS5 this generation ;) I wasn't going to miss out this generation as I did last.

But at the same time, we really only have so much time in a day, it's very refreshing to play games at higher frame rates and generally hit a button, and yawn, and start gaming.
Framerates and load times are definitely quality of life improvements. Whereas graphics are spectacles - something to be awed by, but once the awing is done -- QoL improvements are the staple to keep you going.
 
Last edited:
I am wondering with the adoption of new tools like ML lots of new assets and animations are "computer" generated. Textures, animation, deformations, text to speech and so on.
Few examples are Spiderman MM deformations, cyberpunk speach animation tech etc. This should greatly reduce game development time, as artists/programmers have more time to do other tasks. Why dont we see game development time getting rapidly reduced? Is it because tools are not mature enough? Is it adoption problem? Or maybe there are still not many tools and not widely adopted to engines like UE and so on.
 
Is it because tools are not mature enough? Is it adoption problem? Or maybe there are still not many tools and not widely adopted to engines like UE and so on.
this seems reasonable.
Things take a lot of time. There were a lot of features in DX12 that were exposed last generation that never took hold either and we are only starting to see them show up in engines now. It takes time, money, budget, talent, re-tooling etc etc. Generally, optimization usually occurs near the end of the title before release, so if they could get by without doing any of it, they'll try to get by. And if it's not enough, than their next release will have to be built with those features in mind from the beginning.

Just thinking back, all the titles released within 2021, were likely in development well before I think either PS5 or XSX initial alpha devkits were released. As most titles take at least 2 years in development now, likely 2.5-3 for the type of games that would leverage ML, and sliding to 3-4 with COVID.

I'm not expecting much ML in games (aside from those supported by 3rd party solutions) anytime soon.
 

RT mode is pretty good, mid-50s dips here and there aren't bad. The non-RT mode, with the locked 60fps is still the better option overall. While the game is essentially a cross-gen title, the graphics still look good to me. Plus, I'm a fan of the series, so it doesn't bother me so much.
 
RT mode is pretty good, mid-50s dips here and there aren't bad. The non-RT mode, with the locked 60fps is still the better option overall. While the game is essentially a cross-gen title, the graphics still look good to me. Plus, I'm a fan of the series, so it doesn't bother me so much.
Those wouldn't be a problem at all if PS5 had VRR (and if you owned a VRR TV).
 
I would consider playing this with RT on since performance seems stable enough, though admittedly, it does not make a dramatic difference. From the comparisons shown so far, we are looking at largely, luminance differences between RT and non-RT mode which is something we generally as humans don't pick up on as much as say blur, noise, sharpness, blockiness, shimmering, artifacts etc. The baked lighting does look very good as it is. Side by side comparisons will likely bias towards preference to RT no doubt. But some players will likely have a hard time seeing any sort of difference. Assuming the game sort of is 60 and some dips. RT method seems viable. Regular dips to 40 for other parts of the game, I would then go to non-RT mode.

I agree that some sort of DRS system would make RT mode more viable overall. But it is checkboard already, perhaps that introduces too many artifacts.
 
It's me or sometimes the PS4 Pro on performance mode looks better than PS5 with RT?
The texture in particular.
hmm.. I found the PS5 version to look superior in all scenarios. I will look again. Can you highlight something? Looks to me like higher texture filtering on PS5 and much higher resolution.
 
RT mode looks less constrasty, that can look worse for some.
yea, is this an issue with exporting HDR content back onto SDR? Monitor calibration? Or just people associating darkness with more detail (as a result of seeing more contrast)
 
I would consider playing this with RT on since performance seems stable enough, though admittedly, it does not make a dramatic difference. From the comparisons shown so far, we are looking at largely, luminance differences between RT and non-RT mode which is something we generally as humans don't pick up on as much as say blur, noise, sharpness, blockiness, shimmering, artifacts etc. The baked lighting does look very good as it is. Side by side comparisons will likely bias towards preference to RT no doubt. But some players will likely have a hard time seeing any sort of difference. Assuming the game sort of is 60 and some dips. RT method seems viable. Regular dips to 40 for other parts of the game, I would then go to non-RT mode.

I agree that some sort of DRS system would make RT mode more viable overall. But it is checkboard already, perhaps that introduces too many artifacts.
I think the RT will be more interesting in closed areas (like in the castle) notably with the much improved reflections. The old reflections looked really bad on RE2 and RE3.
 
hmm.. I found the PS5 version to look superior in all scenarios.
In many for sure, in particular in closed spaces, but from DF's video, for example at 4:15 the wood texture looks blurrier on the PS5 RT. Even the sad cheap 2d foliage looks worse.
Maybe just a problem with DF's editing?
 
In many for sure, in particular in closed spaces, but from DF's video, for example at 4:15 the wood texture looks blurrier on the PS5 RT. Even the sad cheap 2d foliage looks worse.
Has somebody checked the impact of RDO optimized compression in multiplatform games?
ETC compressors care only about perceptual image quality at the cost of high entropy, which makes those ETC compressed textures bad for further general lossless compression (just a bit better than compressing JPEG).
RDO optimizer works against this general ETC optimizer to reduce the entropy and increase the amount of compressable blocks.
At high λ, RDO sacrafices small details to make textures more compressable. There were a lot of talks on how PS5's games are smaller than Series X's ones, but for whatever reason, I've never seen texture quality comparisons anywhere.
 
In many for sure, in particular in closed spaces, but from DF's video, for example at 4:15 the wood texture looks blurrier on the PS5 RT. Even the sad cheap 2d foliage looks worse.
Maybe just a problem with DF's editing?
The sharpening filter could be stronger on Pro. They use a dynamic sharpening filter (Contrast Adaptive Sharpening) so results could be different depending of the whole image.

https://gpuopen.com/fidelityfx-cas/

Also the compression usually destroys the finer details, for instance because of higher resolution. I wouldn't worry about it.
 
Last edited:
Has somebody checked the impact of RDO optimized compression in multiplatform games?
ETC compressors care only about perceptual image quality at the cost of high entropy, which makes those ETC compressed textures bad for further general lossless compression (just a bit better than compressing JPEG).
RDO optimizer works against this general ETC optimizer to reduce the entropy and increase the amount of compressable blocks.
At high λ, RDO sacrafices small details to make textures more compressable. There were a lot of talks on how PS5's games are smaller than Series X's ones, but for whatever reason, I've never seen texture quality comparisons anywhere.
Well, the problem here is that we just don't know which texture is compressed with which compression algorithm. So there is not point to compare anything other than "that one looks better ... but we don't know why".
Also resolution differences, post-processing etc can also change how textures look. So it get's even harder to spot. Overall differences should be minimal.
 
DF Written Article @ https://www.eurogamer.net/articles/...-evil-village-demo-tested-on-all-playstations

Resident Evil Village demo tried and tested on PlayStation 5, PS4 and Pro
What's really going on with the '45fps' modes?

Last weekend, Capcom unleashed the first in a series of time-limited demos for Resident Evil Village. First up, we got to see the Village demo for PS4, Pro and PlayStation 5, with the promise of a Castle demo next week on the same systems, followed by a multi-platform release for both segments the week after. We took a look at the initial demo to see how Capcom plans to scale the game across the generations, what the ray tracing looks like, and how performance stacks up. Going into this, the developer told us to expect 45fps on certain consoles and game modes - confused messaging we can clear up having gone hands-on.

So, one demo, three systems and a total of five different play variations to get through - but that proved relatively easy. The truth is, there isn't much to the Village demo. There's a small exploration segment to begin with, but we quickly segue into what is effectively a series of cutscenes linked together with the most minimal of player interactions. The demo is effective in demonstrating the power and versatility of the RE Engine, it gives us strong indications of how beautiful the final game will be but as an actual sampler for the experience, it's minimal.

...

Here, I don't think I agree with this quote.

PlayStation 4 Pro? Obviously, there's no RT there, but there are modes for prioritising performance or quality. The latter mode is one to miss: Capcom suggests we'll get 30fps there, but what we're really getting is an unsatisfactory 30-40fps experience with 2160p rendering achieved by lower grade checkerboarding compared to PS5.

I'd imagine that the checkerboard rendering is the same on the PS4 Pro as it is on the PS5, however, when run at 30-40 FPS rather than 60 FPS, the temporal accumulation will be more prone to breaking down due to the greater time difference between temporal frames/data.

Basically, it could be an apple to apples comparison of why high FPS is needed for any temporal reconstruction (like checkerboard rendering) technique rather than evidence of "lower grade" checkerboarding.

I would postulate that if the title was capable of running at 120 Hz, that even the checkerboarding artifacts on vegetation would mostly disappear as there would be far more temporal data available in a short period of time for our brain to process.

Another way to think about it and something that I've said previously. For this generation any title using temporal reconstruction will likely look worse at 30 FPS than it would at 60 FPS even if the 30 FPS mode has higher quality assets.

Regards,
SB
 
Last edited:
Status
Not open for further replies.
Back
Top