Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

Pedantic: I do not see the words "authentic and fair", I see the words "accurate".

 

This 'dig' is specific to these side by side images, which are being presented as identical by posters here but very clearly not (take a look at them!)

 

really unnecessary negativity, vgtech is doing good job

 

This is projection, dude. I'll open up photoshop and make the image for you though if you don't wanna look yourself.

 

Ps5 in the green channel, xsx in the red channel. Significant difference in angle, seeing as how the points we're zooming in on are at the edges of the camera that's enough to really have no idea whether we'd see the same results in filtering, mipmapping, etc. Learn how renderers work, and stop nitpicking at completely unverifiable details to try to prove a weird theory about VRS.

 

Well, I'm not being passive aggressive, I'm aggressively pointing out how bad of a like for like this comparison is. That's what the original post you're calling out was about, and that's what this last post you're dismissing was about.

I'm sure VGtech does fine work in general, but for this particular comparison that's prompted like 20 posts it's really not that hard to line up the camera the same way twice.

Maybe VGtech never intended for these images to be used in the silly way posters have used them here, but if so, that's why you should provide analysis rather than just dumping numbers/images for fanboys to craft theories about. (Which I believe i've said before)

(Sorry for lumping you in with the other posters defending this comparison, you're right.)

 

Sorry, for replying aggressively, this may be mostly a misunderstanding. -- I was specifically replying to this post, where I lifted the language of 'the best' in response to these particular images being an identical comparison:

My criticism of VGtech here is limited to the accuracy of that one comparison and that I don't think dumping a dropbox of images like that is productive at all, but beyond that I have no opinion of their work.

 

The angle difference is not significant and it shouldn't be enough to cause such big difference of sharpness if that was only texture filtering based on angle. Am I really nitpicking? the difference of sharpness (which is important) can be seen in almost 1/4 of the image on the ground close to the character.

In the previous gen whole articles and threads (with probable causes, mostly unfounded) were made because of reduced AF seen in a few PS4 games. Now after a few months it's already the third game having possibly less AF on XSX and I should not have the right to discuss it with the probable causes, even unfounded?

 

Pedantic: Its not a game yet, its a demo.

 

I want to preference this by saying, I think all this is basically a wash. I have yet to see anything that makes me believe that there will be visual or performance difference that will be easily observed by most gamers.

According to Hargett, Sony GE will have a bigger impact than VRS, which may be true. But a lot of people are conflating two different features that serve different aspects of the rendering pipeline. VRS isn’t explicitly designed to mitigate a lack of culling performance.

It’s like saying more ALUs doesn’t hold a candle to more memory. That may be true But there is nothing from stopping a system from having more ALUs and more memory.

VRS is about shading less pixels while one of the aspects of Sony’s GE is about rendering less triangles. Culling on the XSX is handled by Mesh shaders if you forego the traditional geometry pipeline for certain geometry work. If Mesh shaders are just as capable as Sony primitive shaders what does it matter that culling performance has more impact than VRS?

here is more info on mesh shaders and culling.

http://meshshading.vzout.com/mesh_shading.pdf

Culling is one of the biggest strengths of mesh shaders. There are two high level types of culling you can do: per-triangle culling from within your mesh shaders or meshlet culling from within task shaders. While per- triangle culling was a already possible with the standard pipeline, meshlet culling is something new and very powerful. It will allow you to discard entire batches of vertices before even parsing them which is something the input assembler was incapable of doing efficiently.

 

I'm okay with less AF. Just stop calling it VRS until we can isolate it to be a VRS issue lol

 

As well as texture filtering and reconstruction, things like texture LOD bis or the use of a post process sharpening filter can have an impact on perceived texture quality.

Reconstruction can be affected by frame rate (potentially fewer candidate frames to reconstruct from), and a sharpening filter used more aggressively on a lower resolution image actually make texture details stand out more.

And these things can be additive: lod bias + filtering + reconstruction + sharpening all affecting the output. And some of this you'd probably tune on a per platform basis, and that might be done by different people.

 

OK. I found this comparison on another forum, here with exactly the same angle. Could be AF. it's the same problem as on Watchdogs Legion.

http://www.framecompare.com/image-compare/screenshotcomparison/FEEM1NNU

Seems to be only visible on the ground textures, the rest (the characters) looks identical so I think we can rule out image reconstruction difference.

 

I think AF issues tend to be more noticeable based on how they sampled it with respect to the the polygons they bound it to. If you have a lot of small individual textures making up an area, it's harder to notice. But I think once you have a texture cover a big surface, the AF quality will start to matter more. I noticed that with some games AF quality had minor impact to image quality. And some other games, it's like being with and without glasses.

 

Of course he would say that. But the fact is: On the original Twitter he mentioned a GE, not culling, and what was beeing discussed and who has a revamped GE?