Blender 2.90 Released With Intel Embree Usage, Broader NVIDIA OptiX Support

[phoronix]

Phoronix: Blender 2.90 Released With Intel Embree Usage, Broader NVIDIA OptiX Support

Blender 2.90 is out as a huge feature update to this widely-used, cross-platform and open-source 3D modeling solution...

http://www.phoronix.com/scan.php?page=news_item&px=Blender-2.90-Released

[Danny3]

Nothing for AMD, CPU or GPU ?

[GreenByte]

Nothing for AMD, CPU or GPU ?

Afaik in 2,90 Optix denoiser can be used on any GPU and CPU.

[andrei_me]

Nothing for AMD, CPU or GPU ?

Specifically like the CUDA stuff, not that I remember, but embree should run smoothly on ryzen CPUs, several years ago AMD had split the uber OpenCL kernel from Cycles in several pieces, which improved it's perf on AMD GPU.

Maybe with the release of CDNA, AMD will invest moreon Blender, but there's a limit on what they can do simultaneously with the amount of devs they have

[stargeizer]

Remember all the Optix and Cuda base denoiser was contributed by Nvidia, so it works only on Nvidia cards. What is new, is support for all cards from maxwell to current gen, since before, only RTX cards were supported. For AMD just use the Open Image Denoiser. It works on all kind of CPU's. It's not a bias towards nvidia, it just nvidia contributed themselves with the code.

For AMD Cycles, unfortunately OpenCL limitations are the showstopper. Devs are doing their best, but there's hope the next gen of GPU from AMD can allow feature parity between Nvidia and AMD.

[zxy_thf]

Remember all the Optix and Cuda base denoiser was contributed by Nvidia, so it works only on Nvidia cards. What is new, is support for all cards from maxwell to current gen, since before, only RTX cards were supported. For AMD just use the Open Image Denoiser. It works on all kind of CPU's. It's not a bias towards nvidia, it just nvidia contributed themselves with the code.

For AMD Cycles, unfortunately OpenCL limitations are the showstopper. Devs are doing their best, but there's hope the next gen of GPU from AMD can allow feature parity between Nvidia and AMD.

Theoretically HIP can be used to substitute CUDA, but first we need to have ROCm on RDNA.
Meanwhile I don't think OSS developers will purchase any CDNA card, unless there are <500$ options

[stargeizer]

Usually, Blender Foundation get GPU's directly from AMD and Nvidia, and then they send them to the main developers who required them. Since they are working on the Amsterdam headquarters, they have access to hardware. (Even Apple had sent to BF development machines and granted dev accounts, until Apple revoked and asked their machines back, some time ago).

Of course there's the problem of platform support (Linux, windows), but AMD has hired OSS developers working on it, so i think is just a matter of time proper support is added, at least the required support to allow GPU accelerated software to run on AMD cards on Linux, that is.

[Azpegath]

Michael, I think you missed that it also has initial Wayland support! I had to go through their entire news post to find it, but it is there, hidden in the depths!

https://wiki.blender.org/wiki/Reference/Release_Notes/2.90/More_Features

Linux

• Linux now has initial Wayland support, when building the WITH_GHOST_WAYLAND option. Some features are still missing in this implementation and X11 remains used by default for now.