This is a pretty simple scene that shouldnt take this long IMO Attached are some screenshots and the render output from CPU vs RTX. For some reason my rendering times get slower with CUDA and barely faster with RTX. Another difference when comparing Vray CPU vs GPU is the amount of memory used. The RT Cores increase the ray-tracing calculations and are capable of typically speeding up the rendering by 40%. Chaos Group's V-Ray 5 Benchmark is a free standalone application used to quickly test rendering speeds on any combination of CPUs and GPUs.The Vray GPU rendering engine uses the RT Cores of NVIDIA RTX graphics cards (Turing architecture). For V-Ray GPU they are always enabled without an option to disable them.Vray cpu vs rtx ١٦/١١/٢٠٢٠. They are also enabled by default for new scenes. The consistent elements are automatically enabled when the scene contains an adaptive dome light so they don't have artifacts.
There's an option to enable or disable the new behavior in the Global Switches rollout under the V-Ray tab in the Render Setup window. VRayRawGlobalIllumination = VRayGlobalIllumination / DiffuseFilter). This is because the raw elements have to be derived internally from the corresponding normal elements in order to work with the consistent elements (e.g. The raw elements are affected only when the corresponding normal and filter elements are available, otherwise they're rendered as before. This change makes the elements more consistent but it's also needed for preventing artifacts in these elements with the adaptive dome light (and possibly in the future with other adaptive lights). In both cases they compose back to Beauty correctly but the different types of contributions are now split between the elements more consistently.
Some of the direct contributions that should be in the Lighting and Specular elements were written to the GI and Reflection elements instead.
Previously this behavior depended on the sampling of the lights and not just on the type of the contribution.