• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

State of RT today. Is it usuable?

RTX is basically getting early access to the future. Unbelievable results! Just wish it wouldn't take so long for it to pick up steam, but by next year it's gonna be a done deal. Come on 2020!

 
RTX is basically getting early access to the future. Unbelievable results! Just wish it wouldn't take so long for it to pick up steam, but by next year it's gonna be a done deal. Come on 2020!


It's going to take a long while before many games have RT that looks similar to what is in the official Path traced minecraft RT, and quake 2 RT.
Those two games can only have RT of that level due to how simple the geometry and materials are, the hardware requirements for RT go up exponentially with geometry.
Which is why most games implementing it have very limited and low quality ray tracing effect in part of the graphics pipeline.
 
It's going to take a long while before many games have RT that looks similar to what is in the official Path traced minecraft RT, and quake 2 RT.
Those two games can only have RT of that level due to how simple the geometry and materials are, the hardware requirements for RT go up exponentially with geometry.
Which is why most games implementing it have very limited and low quality ray tracing effect in part of the graphics pipeline.

Agreed, though personally all I care about is using it for GI. Otherwise, I'm still a lot more eager for other developments than RT. In particular I can't consider games with very aggressive lod distances 4K ready. That's the biggest one for me, lod & streaming.
 
It's going to take a long while before many games have RT that looks similar to what is in the official Path traced minecraft RT, and quake 2 RT.
Those two games can only have RT of that level due to how simple the geometry and materials are, the hardware requirements for RT go up exponentially with geometry.
Which is why most games implementing it have very limited and low quality ray tracing effect in part of the graphics pipeline.



This is false, ray tracing performance scales with the log of the number of triangles, it is linear with the number of light sources. There is no scaling factor that is exponential.

In fact the opposite tends to be true and ray tracing scales far better than most of the existing hack and cheats to emulate effects. There is an initial high cost to get any useful ray tracing output, which Turing can currently just do. But increasing the world geometry, textures or material complexity has very little effect. The hardware is also easily scalable without bottlenecks because by definition ray tracing is extremely parallel and every ray is independent. This is not the case with the hacked effects currently employed that all depend on various interactions with other frame data or even previous frames. Hence why SLI and XFX have more or less topped working.
 
This is false, ray tracing performance scales with the log of the number of triangles, it is linear with the number of light sources. There is no scaling factor that is exponential.

In fact the opposite tends to be true and ray tracing scales far better than most of the existing hack and cheats to emulate effects. There is an initial high cost to get any useful ray tracing output, which Turing can currently just do. But increasing the world geometry, textures or material complexity has very little effect. The hardware is also easily scalable without bottlenecks because by definition ray tracing is extremely parallel and every ray is independent. This is not the case with the hacked effects currently employed that all depend on various interactions with other frame data or even previous frames. Hence why SLI and XFX have more or less topped working.

Path tracing is even less troubled by geometry complexity - if you have anything like playable performance in the first place that is :s in my tests in Quake 2 RTX the performance difference from basically an empty box through to the max detail (LOL) the engine can do is basically within 1-2 FPS or 2%.
 
Path tracing is even less troubled by geometry complexity - if you have anything like playable performance in the first place that is :s in my tests in Quake 2 RTX the performance difference from basically an empty box through to the max detail (LOL) the engine can do is basically within 1-2 FPS or 2%.


It mostly scales with resolution, geometry has basically no impact in the end. Nvidia's RTX implement uses a BVH data structure accelerated through the CUDA cores, potentially with some additional hardware functions, so geometry look up is logarithmic. Increaisng triangle counts by a factor of 10 will barely make a difference. In fact, additional geometric complexity could increase performance if the geometry block different light sources then less rays need to be cast.,


The computational cost to do Ray tracing in Quake2/minecreft or in the latest games is basically the same. The difference is a modern game already struggles at 60FPS at high resolution so the added processing time of RTX is a struggle for current hardware. The hardware is highly scalable and the next iteration should see double to triple the performance and become much more viable.
 
It mostly scales with resolution, geometry has basically no impact in the end. Nvidia's RTX implement uses a BVH data structure accelerated through the CUDA cores, potentially with some additional hardware functions, so geometry look up is logarithmic. Increaisng triangle counts by a factor of 10 will barely make a difference. In fact, additional geometric complexity could increase performance if the geometry block different light sources then less rays need to be cast.,


The computational cost to do Ray tracing in Quake2/minecreft or in the latest games is basically the same. The difference is a modern game already struggles at 60FPS at high resolution so the added processing time of RTX is a struggle for current hardware. The hardware is highly scalable and the next iteration should see double to triple the performance and become much more viable.

Yeah on my 1070 the same scene at 800x600 = 30-40FPS, 1280x720 = 17-20FPS, 2560x1440 = 2-7 FPS by far the biggest hit is resolution.
 
A bit of both.

A lot of people seem to have become overly reliant on extremely high frame rates, so consider dropping to under 80fps for the feature as poor. Those of us who game at 30 or 60fps are more likely to find the current RT acceptable.

With that said I do consider it a rushed release. Nvidia chose a pricing strategy, and then had to find a way to make consumers justify it, and RT is what they came to.

Also as proven, RT is not exclusive to RTX cards, it is just that they can accelerate the effects via specialised hardware.
 
So yeah lots of rendering issues due to 25 year old engine, lack of proper materials for some types and looks of surfaces so I've had to do nasty hacks, etc. but if RTX can make a 25 year old engine look like this:

sG8n9LY.png

JDbhBss.jpg

What is the excuse for these developers with modern engines?

(Just imagine it is a nice medium gloss dark wood floor because stock Quake 2 has nothing close to that in its library).
 
Looks good apart from the floor Rroff :)

I'm not saying it is good as Quake 2 wasn't designed to do anything remotely like this - but if it is achievable even in a poor representation on a card that doesn't even have RTX hardware I want to see games pushing the boundaries of this, on GPUs that can do it at playable framerates, sooner rather than later.
 
Albeit a lot more impressive than your average PhysX demo and some things are done nicer than PhysX but if people like AMD had got behind hardware physics we could have had this kind of stuff properly implemented in games by now:

https://www.youtube.com/watch?v=n3RHoTaOZcY

People were doing that kind of stuff albeit with more limitations over 7 years ago with PyhsX

https://www.youtube.com/watch?v=8ztQX0C4wTE
Yeah shame. Had Nvidia not gone down the proprietary route AMD may have got behind it.
 
Yeah shame. Had Nvidia not gone down the proprietary route AMD may have got behind it.

A lot of the reason nVidia went down that route because no one else wanted to get behind it - AMD dabbled with and abandoned a number of their own proprietary implementations after nVidia went full lockdown on PhysX. With ray tracing they've integrated with DXR because MS has at least made some effort albeit not much.
 
Back
Top Bottom