Render time using RT cores

[Fernando_Blanco]

HI,

We already know that a single rtx2800ti does the Vray gpu benchmark in arround 52 seconds.
And that the NVlink works with 2gtx cards using Vray.

Can we have an estimate of how much faster the Gpu bench could be once the RT cores are used?

Assuming the same gpu test could be done, how low can we expectet it to be?

And how far are we from beta of a Vray version using the RT cores?

Bet regards,
Fernando Blanco

[savage309]

There is no public NVIDIA SDK that supports RT Cores with CUDA yet. We expect such in 2019.

The RT Core support with the current V-Ray GPU algorithm will give speed up, but changing a bit the algorithms that we use has the potential to speed up even more. We have some proof-of-concept builds around and we can not say when they will reach a production ready quality, as it does not depend only on us but the driver/SDK as well.

Best,
Blago.

[oglu]

any news here.?

[vlado]

We do have some experimental builds working internally, but unfortunately we are still stuck waiting for NVidia to fix stuff.

Best regards,
Vlado

[oglu]

Does the new Optix version help?