2
$\begingroup$

This RTX family of Video Cards includes different generations of GDDR ram and amounts and multiple price points as well as so many clock speeds and OC options.

Rather than just throw money at a high-end card, which of these attributes best contributes to the Cycles render for the smartest purchase?

$\endgroup$
  • 1
    $\begingroup$ Too much possibility to the question. It is not that simple to render a scene by only some aspect in a GPU. Thing depends on how you use it, how you render it, what is the context and other hardware performance, even which OS platform you use can make a difference. Also, the ray tracing process may being include in future of Cycles rendering pipeline, that should be important as well. $\endgroup$ – HikariTW Nov 22 '19 at 14:03
  • $\begingroup$ General speaking, just go to see a performance website for rendering performance. You should find enough information there. $\endgroup$ – HikariTW Nov 22 '19 at 14:06
  • $\begingroup$ Possibilities is all the question is about - I'm trying to parse this blogs.nvidia.com/blog/2019/11/21/creative-apps-rtx vague claim by Nvidia and Blender's architect, Lommel (who might have been quoted out of context). Knowing where that performance website IS would answer the question to my satisfaction (or someone's experiments). Blender's RTX article in July 2019 linked to by Nvidia isn't enough info. $\endgroup$ – HelloHiHola Nov 22 '19 at 16:05
  • $\begingroup$ code.blender.org/2019/07/accelerating-cycles-using-nvidia-rtx $\endgroup$ – brockmann Nov 22 '19 at 16:20
  • $\begingroup$ You could roughly multiply GPU cores and boost-clock for performance BTW. Only same core generation work. $\endgroup$ – HikariTW Nov 22 '19 at 16:44
1
$\begingroup$

“With NVIDIA RTX, core ray-tracing operations are now hardware accelerated by the GPU, making this the fastest version of Cycles yet.” - Brecht Van Lommel said.

Cycles core computing method is ray trace. And if this thing is supported by GPU hardware directly, then it will be fast to compute ray trace part. So, RT core number will benefit Cycles most in general.


But the whole Blender Cycles use OSL/GLSL for rendering. Not every thing in those complex shader node bottle-neck is to compute ray trace result. See this chart in official Blender Optix support post:

enter image description here

You can see that not every scene got the same boost from using Optix. Since not every thing is boost by new architecture.

Also, there are some special unit to do other calculation in RTX family as well(texture, Render Output):

enter image description here

Those unit (including RT unit) performance are almost directly proportional to GPU Cores in RTX family. So I think don't bother on what core are best for Cycles. Cycles benefit all performance from them.

|improve this answer|||||
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.