Raytracing vs Rasterizer (next gen)

I have not seen any pure RayTraced games.

 

This makes no sense. Please define what you mean by 'RT based'. A game using RTRT hardware for global illumination is still using RTRT. So far, all games are hybrids to some degree except Minecraft, so it's impossible to compare an 'RT based' game to a 'non-RT based' games because such a thing doesn't exist. They all use ray tracing as a technique from AI to lighting to reflections (even if screen-space).

If your argument is consoles shouldn't have bothered including RTRT hardware as rasterization is more effective, you'll have to wait for several years to see how devs actually use that hardware and if they tell us its advantageous or not.

 

I have to nitpick that rasterization and RT should not be compared or related 'against' each other. This makes little sense although it's hard to avoid such thoughts currently, i admit.

Personally i expected more form Sonys games reveal. I expected to be shocked, but it did not happen. I'm even a bit disappointed. (but also calmed down about personal reasons which is good )
I think some more time has to pass, and probably we see more constant progress during next gen than we had for this gen. That's something good.

Well, if this could happen at all, it would not be their fault.
Maybe the move to HW RT came one generation too early. Personally i still think so.
But we don't know much yet about AMDs performance or options, and after seeing only the very first games it's too early to judge.

 

Were you expecting anything in particular? For most of us, I think the correct word is "pleased": we weren't shocked but we were pleased with the beautiful graphics (and added possibilities) next-gen machines can produce, even if the specs are not extraordinarily high. As in this current generation, the big expectations and the shocked reactions are a thing of the past, honestly. Yes, we may play a game and feel amazed at the graphics, but I don't think this is the feeling that defines a generation anymore.

Maybe the day when consoles have amazingly powerful CPUs (maybe GPUs, depending on the tech) able to perform real life simulations, fully interactive/destructible levels, etc., we will feel that again.

Ok, that sounds... weird. Just saying.

 

How do you know if Horizon is not using some form of raytracing? Wait and see some technological explanation and a deep dive into the game.

 

Speak for yourself! First thing of interest to me is lighting and shadowing. We've had great geometry for years, but haven't had realistic, uniform, consistent, dynamic lighting ever. Lighting hacks are really disruptive in games. Screen space reflections that phase and warp, and AO that sticks dirty halos around objects, are very objectionable. Similarly, great rendering with lousy animation can be really disconcerting.

You're entitled to your preferences, but please don't try and turn them into objective analysis when its entirely subjective.

Have they mentioned anything at all about the tech? Is there reason to think GG would ignore the RTRT capabilties and leave that part of the silicon idle?

 

If you don’t mind me quoting you

What do you think the recent inclusion and push towards RT is supposed to solve?

RT is everything to do with light transport.

it’s is the hardest thing to accomplish. Lighting and shadow computations require the most computational power which is why we’ve never done it until now.

you are truly all over the place with your opinions to the point of contradictory.

 

This is exactly the point I was making in some previous thread. I think it was me moaning about the hair in Pragmata.
Also I wouldn't be to sure that Guerrilla isn't using some form of Ray tracing.

 

Yeah i expected wonders that make me stop working on GI and jumping out of the window. So it was more personal anxiety than concrete expections, tbh. Just to explain my POV.

I remember i was more impressed at PS4 launch. Killzone with PBR and all those new screenspace tricks was mind blowing.
But yeah, it's ok i think. Impressing people with better gfx becomes harder and harder, and what they've shown is definitively good stuff.
UE5 has stolen the show, that's also a big factor.

Personally i think most promising opportunity of progress is character simulation. That's interesting and has no unpractical power requirements.

Destruction is difficult, though. Currently it's mostly precomputed fragments, which is already demanding extra work for both content creation and tools development.
At runtime it causes spikes in physics simulation, so it makes sense to have a low quality simulator, which is contrary to our expectations of 'real life simulation'. Having more CPU cores would not change this - we would still try to saturate them, so no easy way to handle such extra runtime spikes.
And finally, if we had a game where you can destruct absolutely everything, what does this mean for level design?
So i think it will remain a cosmetical feature for most games, but we'll see.

 

I think you misunderstand Ultragpu in that he doesn't have anything against Ray tracing but that the hardware isn't up to it yet so it might not be worth it.

I had similar feelings but seeing some first year games with it seems like it's got promise this upcoming generation.

 

I think accurate lighting/shadows/reflections is one of those things that you don't know how much it makes a difference until it's there.

 

Lay off the hyperboles. You're not doing yourself or your viewpoint any favors by being so out of touch. They make it impossible to take anything you say seriously. It turns your entire posts into the ramblings of a mad nan.

 

Don't judge a game from a trailer, which is designed to show you the best shots, judge it from actual gameplay wandering around the environment, that's where the defects in static GI start to show, that's where the SSR reflections become painful and lacking and where the shadow maps reveal just how lacking they are at doing 100% shadow coverage.

 

I guess we can assume the first games when switching to a new generation use still the existing engines and workflows, so increasing resolution is the first way to utilize more power. The shown games did this, increased details, and they also have RT running beside that, which proofs RT works good enough.
After some time i expect resolutions to go down (using whatever tricks like CB or upscaling) and IQ to go up.
Some games also have RT effects at half resolution, so there already is compromise as you suggest.
Looking at the cost of SSGI effects like McFly SSRT the performance hit is huge there too, but without a chance to get correct and stable results. (Though i suspect Horizon uses SSGI maybe, remembering the crabs)

There surely is no reason to worry about things like 'reducing geometric detail just to support RT' to happen in general. If this would be necessary, the devs would still aim to find the best compromise pleasing the people to sell games.
Even if they go into extremes here, the results could be interesting, both about rendering and how people react.
I have my own doubts about HW RT, mostly about questions of LOD and 'can it replace damn shadowmaps'. But even if neither works, it's still fine for reflections up to at a limited distance for example. Everybody will come up with something useful - we have to deal with limitations all the time.

Actually ROPs are at least as questionable as RT cores are, if we take UE5 as example. There is the point when fixed function HW is no longer necessary, there is a point when new fixed function stuff makes sense. But it's hard to tell when.

 

First off, I wouldn't use any launch title as a measure of what's possible much later in the generation.
You wrote in the title of this tread, raytracing vs rasterizer; but it's like you fail to realize that both are completely different rendering methods that could not be further apart. One is fast 2D drawing of a 3d approximation of what things should like, the other mimics physical light transport.

You are failing to recognize the apples and oranges comparison here; airplanes and rockets both fly. But only rockets will take you to the moon.

That means everything you learned about airplanes don't apply to rockets. To get RT to work, new techniques and methods and processes, workflows and tools need to be developed. Engines may need entire rebuilds, or whole new engines may need to be developed.
new hardware will likely be developed.

TF isn't what is needed to solve RT. They tried that, it doesn't work. We need RT specific solutions to solve RT.
If all you got i to brute force this problem using TF, you'll need a whole lot more than 20.
https://www.vice.com/en_us/article/ne9bbm/star-wars-nvidia-dxr-demo-gdc

you whine about RT performance without recognizing that the current GPUs are 20 some odd years of pure rasterization innovation after innovation.

Just because something is labelled as a GPU; doesn’t mean it’s great at solving all graphics problems.

 

Truthfully ray tracing and rasterizing can be roughly seen as two loops in reversed order:
for each objet in the scene
for each pixel it touches
do stuff

for each pixel on screen
for each object in the scene
do stuff

Path tracing, bidirectional path tracing, photon mapping and the likes are about light transport.