Blender Discussions

[cekuhnen]

I am comparing: NVIDIA GTX 1070Ti vs AMD RX 5700 XT AMD ProRender

And oddly the GTX 1070 Ti Windows openCL runs better than the AMD RX 5700 XT Metal macOS.

Any idea why? I would have assumed that that AMD card would perform much better specifically also with Metal!

[caretaker]

Have you tried bootcamping in windows and seeing how the amd card performs under openCL?

Because you are comparing apples to pineapples, they seem similar but are way different at their core. Radeon pro renderer is developed for OpenCL and Vulkan mainly, Metal suport is just a gift that the radeon team has granted.

[cekuhnen]

I know this is apples vs oranges to a certain degree.

I have bootcamp too and will try it there.

I personally was hoping AMD ProRender could become a good alternative to CUDA based Cycles in Blender.

[irz]

Hi, please,

try this tests:

1. RyzenGraphic_27 file for test with 1070Ti for Windows. Try to render with Cycles, but OpenCL also welcome.

2. RyzenGraphic_27_prorender file for Radeon Pro Render. For 2.2.1 and above. Try with Ti, RX, Windows, Mac...

Please, post your time render. Do not touch any settings.

For example:

RyzenGraphic_27_prorender with RX5700 XT, blender 2.82 and RPR 2.2.12:

1. Mac Os render time 00:06:60

2. Windows 10 render time: 00:07:40

[cekuhnen]

I am curious how you got this down to 7:40?

ProRender

Windows10 SSD 

First         25:28 sec 1 x RX 5700 XT

Second   13:40 sec 1 x RX 5700 XT

macOS Catalina MacPro 2010

First        18:23 1 x RX 5700 XT

Second   9:831 x RX 5700 XT

Windows10 SSD NVIDA 

First         43:17 sec

Second   14:49 sec 1 x GTX 1070Ti

Second   15:20 sec 2 x GTX 1070Ti

CUDA

7:27 sec  1 x GTX 1070Ti

5:02 sec 2 x GTX 1070Ti

So oddly with two GPUs the time is a tick slower?

openCL so bad for NVIDA dual GPUS?