Next gen lighting technologies - voxelised, traced, and everything else *spawn*

RT shadows are not moving:

(time in the link)


Many of those artifacts could be programming bugs, or it could be that RT RT can not do better. I am not trying to determine if RT RT can not do it better, or the devs can not do it better(inside the deadline they were given).
 
They move just fine in the game, and even in all of the other scenes in the video, what you are referring to is a bug.

Here not moving either:
upload_2019-3-21_15-18-46.png

Often the updates of the shadows are too slow, it looks like not moving, but it moves a little bit:
upload_2019-3-21_15-20-46.png

How is this not an improvement?

Artistically speaking:
We are not able to use a ruler to prove shadows are correct. For me it looks bad with RTX On. I am an artist for long years. I draw since 5yo. For me it looks bad with RTX On.

Technologically speaking:
What vulnerability of shadow mapping could possibly cause to miss such a large area of shadows?! The shadow receiver and the caster are far from each other enough to avoid peter panning, and the rocks pop enough as to not miss the shadow because of the 2D floating precision problem of a shadow map depth buffer. It is more probably the bug from the other comparison i added as a screenshot.
 
Technologically speaking:
What vulnerability of shadow mapping could possibly cause to miss such a large area of shadows?! The shadow receiver and the caster are far from each other enough to avoid peter panning, and the rocks pop enough as to not miss the shadow because of the 2D floating precision problem of a shadow map depth buffer. It is more probably the bug from the other comparison i added as a screenshot.

One explanation is the devs loaded the ground as not self shadowing. And RT eats all the triangles in a bruteforced way so it detects self shadowing.

My bad here.
 
Here not moving either:
You can't pick the movements of the bottom shadows near the water because they are extremely soft, and because of youtube compression, focus on the hard shadows in the upper part of the image and see them moving.

One explanation is the devs loaded the ground as not self shadowing. And RT eats all the triangles in a bruteforced way so it detects self shadowing.
My bad here.
Yeah, the developer made so many objects in the game that way, even a flash light focused directly on them fails to cast shadows without RT.
 
You can't pick the movements of the bottom shadows near the water because they are extremely soft, and because of youtube compression, focus on the hard shadows in the upper part of the image and see them moving.


Yeah, the developer made so many objects in the game that way, even a flash light focused directly on them fails to cast shadows without RT.

I personally would disable RTX shadows in Tomb Raider. I would use RTX reflections in Horizon Zero Dawn, because the water surfaces in this game were pretty bad. But, only for the water surfaces. Not asking the modelers to add mirrors everywhere and make it all reflecting. I think if Horizon Zero Down had the grass popping of the Red Dead Redemption 2 and RT reflection for water surfaces, it would become a best game ever.

There is a saying "If a machine works, don't touch it, don't even look at the machine. When you start improving the old good working technology is where you break it all". RT should be introduced in a game only in those places where old school just can not do it.

If we take a person form the street who never painted, who never programmed and never played games, would he be able to perceive a difference? And the worst - ask him for 500$ more for one of the images over the other.

I am a developer and i can see it, but i develop for regular people. Most of times old school game techniques are ok enough.

I wish too, to drop it all into some buffer and make the hardware process resolutionX*resolutionY*((numbOfRaysCasted*numbOfAllTrianglesInTheScene)^numbOfBounces) 60 times per second.... but we have to wait 50 years more for such hardware.

(DXR recomputes the BVH for each frame of an animation, which is costly, so maybe some pieces of foliage is never updated inside the DXR structures, but updated for the rasterizer. This could explain why some shadows are still while the foliage moves)
 
Even in the non playable demos of Metro there are whole objects without shadows. Staying next to other objects who cast shadows.
Yes, but Metro uses shadow maps for shadows, Metro doesn't use RT for shadows at all.

I personally would disable RTX shadows in Tomb Raider
I personally wouldn't, if not for the soft shadows effect, then for the increase in the number and density of shadows, and the increase in the number of lights casting shadows.

HFTS and PCSS+ were used before to obtain the contact hardening effect of the shadows, but they came with a massive performance hit and limited visual improvement. In that regard it's my opinion that RT shadows are significantly better than these effects.
 
Can we assume with RTX there are no realtime shadow maps used at all?
In above video he showed a transition from RT to SM shadows in the distance. I guess those distant shadows are from a static shadow map cache?
If it is like that, we could compare RT vs. SM performance here.

I realize the caching advantage is too large to hope we could get rid of SMs anytime soon.
Again the most promising idea would be object space shading because stochastic texel updates only make sense with RT and it could become a win even performance wise.
(This sucks... i already decided against object space personally, but now it becomes interesting again.)

Visually i did not expect RT to be a big win here, and i still think so.
We would need a game with large area lights to show the advantage.
 
Yes, but Metro uses shadow maps for shadows, Metro doesn't use RT for shadows at all.

Yes. The GI of Metro skips whole objects too.


I personally wouldn't, if not for the soft shadows effect, then for the increase in the number and density of shadows, and the increase in the number of lights casting shadows.

Adding an extra light to RT is seriously affecting performance too. I am not able to tell if cube shadow mapping is more taxing than RT point lights. It should depend from the implementation. We are far from being able to add a light for each of the candles in the scenes in Tomb Raider. Or for the many flames in the photos you provide. I don't think our RT RT advances can handle in real time so many light sources.

From my experience, a light that does not cast a shadow makes it all look fake. It is visually horrible to have a light source and no shadows. Better to delete such toxic light sources, if the design of the scenery allows it.


HFTS and PCSS+ were used before to obtain the contact hardening effect of the shadows, but they came with a massive performance hit and limited visual improvement. In that regard it's my opinion that RT shadows are significantly better than these effects.

Not in Tomb Raider in my opinion. As in the last YT video, the guy noticed that the old school shadowing in TR, already takes in account the distance from the caster. Notice, RT casting one single ray will not produce soft shadows. RT has to cast multiple rays in order to obtain soft shadows. This is expensive. I am sure there is some blurring trick in TR to make shadows softer softer.

Soft shadows in WebGL(this is a very basic api, has no compute, no geometry or tessellation shaders)
(the last playable demo near the bottom of the page)

http://codeflow.org/entries/2013/feb/15/soft-shadow-mapping/

(It has artifacts too)
 
Can we assume with RTX there are no realtime shadow maps used at all?
RTX Medium in Tomb Raider still uses shadow maps for general shadows but uses raytracing for point lights. Only High and Ultra use RT for all shadows. Excluding LOD distance shadows I guess, since those seem to be shadow maps still for all settings with High having more aggressive LOD.
 
Actually, the space between the leaves produce a pinhole effect:

16831042551_152a35aba2_o.jpg


Way more complicated task than just casting shadows. Maybe after 100 years and for PC only.....
 
In real life, in a sunny day, shadows are pretty sharp:

upload_2019-3-21_17-38-10.png

upload_2019-3-21_17-38-41.png

upload_2019-3-21_17-40-20.png

upload_2019-3-21_17-43-47.png

Some times, the RT shadows of TR are just way too blurred.
upload_2019-3-21_17-48-16.png

Very unrealistic.
 
The softness of a shadow is dependent on the distance between the caster and the receiver. The larger the distance the softer the shadow. Your examples fail to account for this.

Those are real life photos, how the real life fails to cast shadows?!?!

To compute the softening of the shadows, it is used the distance between the receiving point and casting point, and the distance to the light source.
If the distance to the sun is so big, so it is ignored, then this should make you realize how small the softening of shadows is in a sunny day without clouds.

Softening will happen mostly due clouds.
 
Back
Top