nice to see some move by amd i guess ?
about the nvidia controversy , All the competition are unfortunately years behind.
nice to see some move by amd i guess ?
about the nvidia controversy , All the competition are unfortunately years behind.
And the fact that Blender is wildly used as a benchmarking tool.
HIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
That’s really cool. I just hope Nvidia RTX is also included in Vulkan and DX12 standards as a ray-tracing backend if it’s not yet. OptiX is still NVidia proprietary stuff.
That is really important. Where could I find information about it?
I hope OptiX is opensource else it goes against Blender’s license. Though Cuda is also proprietary… licensing is weird.
Still unless AMD releases something, there wont’ be any reason for implementing Vulcan or DX12 support. Look how long it took OpenCL vs CUDA…
Only rtx card right now, hopefully they add Pascal as soon as possible
But anyway ,another great news for the blender
Both the OptiX raytracer and CUDA are Nvidia drivers. The GPL is fine with drivers.
Build with RTX support from LazyDodo on Graphicall:
https://blender.community/c/graphicall/Cfbbbc/
Cheers, mib
So I assume this will only be officially added in 2.81?
It won’t be in 2.80 tomorrow if that’s what you mean. It also requires Nvidia 435.x drivers, which aren’t out yet.
Cool news. I hope this will be a part of the daily Master builds soon.
Guys, I want videos!
Someone who has a setup that can show the new potential of cycles make a video please!
visually there should be no difference, it just be faster… the code.blender.org post has some benchmarks graphs though.
just … I want to see this … ^ ____________ ^
Tried it with RTX 2070 and Fishy Cat with 128x128 tiles and 1000 samples:
did it also take you a lot of time loading the kernels ?
I just use whatever time the Blender measures in the upper left corner. Trying to render the one frame of Spring in test files causes OptiX out of memory error, while CUDA still manages to render it.
Clicked both multiple times just to make sure there’s no time differences because of caching.
That should always happen only once for the very first time you render. Then you should be all good pretty much up until you reinstall Blender.