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

Status
Not open for further replies.
if devs design a SW VRS that suits their need better, can they still run it with HW VRS to save some ressources ?

Moving Gears to Tier 2 Variable Rate Shading | DirectX Developer Blog (microsoft.com)

While we were able to implement VRS for all the passes that gave us the biggest bang for the buck, it was not plumbed into the entire engine due to time constraints. A deeper integration would allow VRS to provide even larger GPU savings.

“Software-Based Variable Rate Shading in Call of Duty” presented at SIGGRAPH 2020 (http://advances.realtimerendering.com/s2020/index.htm) has some interesting thoughts on this topic as well. They present a method leveraging how console hardware handles MSAA to emulate VRS on platforms without hardware VRS support and extra flexibility such as smaller tile sizes. In addition, they present an optimized way to apply VRS to compute shaders that uses ExecuteDispatchIndirect to ensure only waves with actual work are dispatched in contrast to our brute force method. However, Software-Based VRS also has some trade-offs including implementation complexity and the overhead of a de-blocking pass. One possibility is to use a hybrid of both techniques, switching between VRS techniques based on the characteristics of the rendering pass.
 
Why are people trying to say the tech is not without its downsides and is a perfect solution? VRS is not DLSS, it's not even a reconstruction tech. It's a clear trade-off system.

VRS, dlss, dynamic resolution, taa upsampling, and checkerboard are all clear tradeoffs. They all have incredibly noticeable artifacts if you care to freeze frame or zoom in 4x. If everybody has 200tf gpus we wouldn't use any of them, but since we don't, the techniques should be judged based on how good they look and how much performance they save.

VRS is blurrier than something like checkerboard, but it has no ghosting, no noisy breakup of samples on area lights/raytracing, and no other temporal artifacts. Also, in theory, vrs can be only used for parts of the screen that are disproportionately expensive (such as a part with many lights in a forward renderer, or certain vfx where overdraw is an issue, etc), or in areas where the loss of resolution is completely invisible (in large flat colored areas, in source pixels behind blurs, etc.) Vs a reconstruction tech where you're stuck downscaling the entire screen even if only a tiny portion is pushing you over your ms budget.

But it's relatively coarse and you're stuck with big resolution drops in the spots you use it for.

Everyone here understands the tradeoffs of these techniques except you and snc, and you guys have been trying to make up cases where it looks like a bad technique (by only comparing in like-for-like resolution shots for doom eternal, when the reality is xbox's base resolution is much higher, by making up phantom vrs to explain low res textures in games like outriders, etc.) for months
 
Yeah? Tell me about the differences discussed in the video. I'm all ears.
PS5 and Series X have higher settings than the other versions. Better textures, public, lighting, grass, anisotropic filtering ... however, all these characteristics will be analyzed in greater depth in its final version. Series S uses some Nextgen settings, but nevertheless lacks others (the absence of an audience, for example).
 
Some pretty funky resolution numbers knocking about this video. I wonder if the Xbox One X and PS4 Pro are using some upscaling technology


In order:

Xbox One X = 3008x1692
Xbox Series X = 2880x1620
PS5 = 2880x1620
PS4 Pro = 2560x1440
Xbox Series S = 1920x1080
PS4 = 1600x900
Xbox One S = 1280x720

Last gen machines pumping out higher resolutions, presumably at lower settings.
impeccable timing for release considering where England is right now
 
Re-watched due to the conversations being had.
Conclusion is relevant but an interesting watch regardless to complement the blog post in video form.

Can't find the other one I was thinking of.

Hadn't watched this one, thanks. The part about setting 2x2 shading rate for a pass and measuring perf gains is really good advice, if you don't see much of a perf difference by setting the entire frame to be 2x2 then it means the bottleneck for that pass is somewhere else and probably not worth looking into VRS. That way you can easily see where VRS is useful in your frame, the changes required are fairly trivial to get some initial performance numbers. Although, RDNA2 doesn't support 2x4, 4x2 and 4x4 shading rates, these would probably be useful in cases like the DOF scene in that video.

Edit: Not to mention AMD provide drop in open source code that you can easily integrate into your engine https://gpuopen.com/fidelityfx-variable-shading/
 
Last edited:
10-15% more framerate with VRS while PS5 looks overall >15% higher resolution in both 60fps modes. Easily noticeable without zoom using 2 identical scenes (source VGTech).

7iG3VHk.gif

I know it's not a discussion most want to hear about. They want to think VRS is a miraculous tech that is going to dramatically improve all parts of a game without downsides. But VRS is a sharp trade-off, one many wouldn't want in their game, it's not a silver bullet.
 
  • Like
Reactions: snc
10-15% more framerate with VRS while PS5 looks overall >15% higher resolution in both 60fps modes. Easily noticeable without zoom using 2 identical scenes (source VGTech).

7iG3VHk.gif

I know it's not a discussion most want to hear about. They want to think VRS is a miraculous tech that is going to dramatically improve all parts of a game without downsides. But VRS is a sharp trade-off, one many wouldn't want in their game, it's not a silver bullet.

Again vrs doesn’t smear the entire screen. It’s dynamic and only portions of screen where the algorithm thinks shading isn’t important, like the one you post of a small portion of the entire screen at 400% zoom. No matter what as noted by every analysis XSX is sharper looking in every mode. Vrs allows the XSX to get up to 29% better res with the help of the stronger gpu/bandwidth. Every dev will use it better over time as they get accustomed to it. TC did wonders with it in Gears
 
10-15% more framerate with VRS while PS5 looks overall >15% higher resolution in both 60fps modes. Easily noticeable without zoom using 2 identical scenes (source VGTech).

7iG3VHk.gif

I know it's not a discussion most want to hear about. They want to think VRS is a miraculous tech that is going to dramatically improve all parts of a game without downsides. But VRS is a sharp trade-off, one many wouldn't want in their game, it's not a silver bullet.

If you only notice it with screenshot comparisons then it's not really that bad.
 
10-15% more framerate with VRS while PS5 looks overall >15% higher resolution in both 60fps modes. Easily noticeable without zoom using 2 identical scenes (source VGTech).

7iG3VHk.gif

I know it's not a discussion most want to hear about. They want to think VRS is a miraculous tech that is going to dramatically improve all parts of a game without downsides. But VRS is a sharp trade-off, one many wouldn't want in their game, it's not a silver bullet.
quick ask; which ones are these on VG Tech? I can't figure out where you clipped from.

Thank you in advance!

Also if possible could you pixel count both? I don't need to see the proof, just provide the numbers.

Thanks.
 
Last edited:
I know it's not a discussion most want to hear about. They want to think VRS is a miraculous tech that is going to dramatically improve all parts of a game without downsides. But VRS is a sharp trade-off, one many wouldn't want in their game, it's not a silver bullet.

Stop making and presenting false statements. You're the only person ever talking about VRS as a silver bullet or miraculous tech. Everyone except for you has a reasonable view of it.

Consider this is your last warning to stop having discussions in bad faith.
 
7iG3VHk.gif

Are we sure that's even VRS? I'm absolutely not sure on this, but it looks like this is low res in texture space, not in screen space -- like from a streaming issue. There's a polygonal shape on the left side thats identical on ps5 and xsx. Wouldn't you expect either the texture to sharpen right around that seam, or the 2x2 vrs downscaling to also apply to the edge?

Edit: Maybe its just tricking my eyes, and because both images are blurry the polygonal seam looks the same to me even though its lower res. Would need to see this with like, textures turned off.
 
Last edited:
Please stop making definitive conclusions about nextgen hardware and nextgen technologies based on games made for ps4 and xbox one.

Xsx and xss were designed for completely different development methods using sfs, ms, vrs, rt and ml in the early stages of development.
Not for hastily ported games from the previous generation.
 
Status
Not open for further replies.
Back
Top