OptiX is coming in Blender Cycles.
I have HD7870 2GB.
This is a request to make drivers which will support OptiX for Blender Cycles because after Blender 2.76, Blender developers dropped support for 7870 due to driver incompatibilities.
I second this. It'd be a pity if RPR would stay behind Cycles progress.
There is dnoise add-on based on optix for Blender which is better than integrated one for cycles. Could someone make this add-on works for AMD cards as well?
I saw that with OptiX and CUDA/RTX, loading the kernal in Cycles is faster than with AMD.
OptiX now supports viewport denoise in Cycles.
Is it possible for AMD driver to work with OptiX so that AMD card could use OptiX for rendering in Cycles?
Optix and Cuda are closed source products of NVidia, we can not support them unless they make them work with open standards such as OpenCL.
Is it possible to extend OpenCl like OptiX API ? To improve Cycles kernel load and rendering.
not really, gpu rendering needs to compile a kernel no mater the API you are using, weather you are compiling it once or every time you het render thats a different task, however Cuda and OptiX rendering dont require compiling because the render kernel is precompiled within the driver its self.
Can it be done in OpenCl so that kernel is precompiled within the driver ?
I`m not a developer but as far as i know open CL is a software technology that being open source can be used on any compute hardware, meaning Intel, Nvidia and AMD can all use the tech and any of their hardware CPU and GPU cand compute OpenCL calculations, meaning that its hard to make a precompiled kernel that contains every type of architecture and hardware feature. CUDA an OptiX is created just for Nvidia hardware and its easyer to create drivers for nvidia hardware since they allready have the info on what they use.
Since CUDA and OptiX are closed source software (because Nvidia said that some of the tech is licenced from others and they havent created everything) no universal driver or source code will be provided to the linux community making it hard to port it to other devices or hardware.
AMD are just cheap GPU's as they are. The dev's are too lazy to work to implement a better Blender Cycles support and too busy to work on Prorender that in reality only mac users need for the moment. They aren't even able to deliver a correct version the Adrenaline 2020 software, it's not able to display framerate with the overlay interface. Once our projects done will migrate all our PC to Nvidia. Enough with those cheap GPUs and those lazy devs that should be fired. Nvdia is not perfect but at least they work hard for the users.
It's a bit hard to reply seriously to a comment calling for us to be fired. Please keep civil here. This is your only warning.
However, that said, AMD does fund cycles development as well as working on ProRender (unfortunately my group does not work on the driver). Our hope is to improve both and provide a good experience with ProRender. Is there something specific with ProRender you find lacking? We actually focus quite a bit if not more on windows with RPR.
No no no... Please focus more on the mac side : ) Because RPR works better & faster on mac than on windows. Also, some more goodness on the road with 10.15.5 (metal api & RX drivers updates).
Btw, why RPR not to detach to separate program while transfer meshes & objects data as an addon bridge from 3d programs to RPR application? Just discourse.
Your gpu can handle raw calculations faster than NVidia, everybody knows that. The problem for me as a 3D artist for example is that most of my pipeline software is way faster with CUDA/OptiX than with OpenCL.
Blender normal map baking, Blender viewport denoising, Substance Painter maps baking are just all much faster.
The question is : What can you do for me so I will choose your RDNA2 card instead of the Nvidia Ampere cards?
Cause I'm sure it's all about the software at this point. AND I do really want to choose AMD (already having a ryzen 9 5950x, upgraded recently from my trusty and immensely helpful ryzen 9 3900x)
Hi, CUDA/OptiX used hardware solutions from Nvidia, that's not a software solution. So, for support CUDA/OptiX needs to add Cuda cores to AMD GPU. if it will be on the software side - it will be possible. Now for Ray Tracing boost AMD used Compute Units and NVidia - Cuda/Tensor.
For professional work with most 3d pipe lines and program solutions Nvidia still better then AMD because its support both tech OpenCL and CUDA\OptiX and AMD not, so maybe in the future this situation will be change, for now radeons just good gaming solutions and that all
For those people not in the mix, AMD dropped OpenCL support with RDNA2 in favor of Vulkan and D3D12.
I've actually been thinking about this for over a year now... so why doesn't AMD cooperate with Autodesk (especially 3dsMax) and Blender developers to create a renderer that's based on say D3D12 and/or Vulkan?
Both are open source, easily provided (much better use case compared to OpenCL), and works across all hw.