on Optix6.0 my 2080ti is 4X faster than titanV

Just download and tried on my project. Now the 2080ti is 4 times faster than my titanV.
Compare with previous version, titanV is much faster than 2080ti.(40% around)
nice work! thanks for delivery on the first day of the new year!

That’s impressive!

What kind of project did you ran? Was it related to path tracing. Would love to see improvements levels for it.

thanks! Yes. path tracing.
Simple upgrade to 6.0 give you at least 3X performance boost on RTX cards. Optimize with GeometryTriangles from old Geometry give you another around 2X boost.
Best of all, 6.0 fixed all issue we have before. (mentioned in my previous post)
Applaud for the optix team!

Thanks for the info.

Lovely! Can’t wait to check how it scales.

zhangy, is there any chance you can provide more detailed info? I’ve been battling people very sceptical about RTX performance (something I find quite confusing as we’re just talking about accelerated BVH intersections). I’ve tried to tell them this is an order of magnitude improvement but they will not listen!

What kind of workload are you submitting? Is it static or dynamic? I’ve only seen this technology myself in the Microsoft DXR samples, which are extremely basic, and in 3D Mark, which I bought and played with the other day. In the latter case it’s a dynamic, real-time (there is a time constraint) scene. I’d love to know how this performs in other rendering scenarios, i.e. the somewhat more relaxed time constraints of a modelling viewport or a high quality render.

That is only half of the story. The hardware RT cores accelerate both BVH traversal and the actual triangle intersection.

My case is a static scene. Here is some statistics, but by no means for accurate measurement.
Scene info: Objects:2400, textures:300, materials:6800, triangles:320,000,000,
res:1920x1080, lights:1 bounce:8 total samples:100
on Optix6: 2080ti is 20 sec, titanV 85 sec
on Optix5.1: 2080ti is 210 sec