[Success] AMD RX6000 Series working in macOS

dclive · Sep 5, 2023

Sorry to ask the blindly obvious to Davinci Pro pro, but ... if you jump on a Mac and run those same workstreams, you _do_ get 100% CPU and 100% GPU?

Edit: Newbie on dvp; just reading some forums earlier today to see what's accellerated.

ori69 · Sep 5, 2023

dclive said:
Sorry to ask the blindly obvious to Davinci Pro pro, but ... if you jump on a Mac and run those same workstreams, you _do_ get 100% CPU and 100% GPU?

Edit: Newbie on dvp; just reading some forums earlier today to see what's accellerated.

RX6600 vs M1

Edhawk · Sep 5, 2023

If you are going to compare systems at least try comparing two systems using the same GPU, i.e. a MacPro7,1 and your hack, both using an AMD Navi graphics card.

Don’t compare your hack with an Apple M1, or any other non-Intel Mac, as Apple Silicon (AS) systems use completely different processors and graphics. Plus Apple will have optimised their new AS systems to work with Davinci Pro.

luchina · Sep 6, 2023

I tested it on two Intel-based macOS, 2 Hackintosh with Intel 12th and RX 6900 XT, and a real Macbook Pro, and on the macbook the GPU usage is 100%.

Which leads me to understand that everyone is on the same version of macOS and Davinci Resolve, but only the Macbook actually uses the GPU to render...

oli.mathieu · Sep 6, 2023

ori69 said:
RX6600 vs M1

View attachment 571125

View attachment 571126

If your screen capture show the same Davinci project, It only means for me that RX6600 is more powerful than M1 GPU.
Those graph are relative!!

oli.mathieu · Sep 6, 2023

luchina said:
I tested it on two Intel-based macOS, 2 Hackintosh with Intel 12th and RX 6900 XT, and a real Macbook Pro, and on the macbook the GPU usage is 100%.

Which leads me to understand that everyone is on the same version of macOS and Davinci Resolve, but only the Macbook actually uses the GPU to render...

What is your Macbook's GPU?
meanwhile ...... RX6900XT is a beast
So 10% of RX6900 XT could easily be 100% of laptop GPU

But I agree if RX6900 XT is only 10% usage during rendering, it mean the bottleneck is somewhere else: Writing speed Drive may be ? Playback Speed drive may be ?
What are the models of those drives ?
What is you rendering speed (fps) ?

ori69 · Sep 6, 2023

luchina said:
I tested it on two Intel-based macOS, 2 Hackintosh with Intel 12th and RX 6900 XT, and a real Macbook Pro, and on the macbook the GPU usage is 100%.

Which leads me to understand that everyone is on the same version of macOS and Davinci Resolve, but only the Macbook actually uses the GPU to render...

I did tests of the same project with different graphics cards. I noticed that the weaker the graphics card, the higher the usage in Final Cut Pro.

And finally I fired up on an i7 8700 with a UHD630.

luchina · Sep 6, 2023

Whooops, I tested the same project here on Windows 11 with Davinci Resolve and the same thing, high CPU usage and very low GPU usage, that is, the same scenario I have on macOS... so you can even disregard that it's something on macOS/Hackintosh.

I don't have Final Cut to test, but I did a search and many people complain about AMD GPU with Davinci Resolve....

Test hardware: 12900K, Z690 Aorus Elite, 32Gb DDR4 4133Mhz and a NVME SN770 2Tb Gen4.

[edit]

I tested it with Final Cut Pro X, and the same thing... it uses 100% CPU and 10% maximum GPU.

This afternoon I tested adjustments with unfairgva (all values) and none of them had any effect... when rendering, either with davinci or with fcxp, the use is intense of CPU and little of GPU.

Middleman · Sep 6, 2023

@ori69,

Interesting results from those tests! Have you tried them all also with and without RestrictEvents and/or CPUFriend?

zirkaiva · Sep 17, 2023

luchina said:
I tested it on two Intel-based macOS, 2 Hackintosh with Intel 12th and RX 6900 XT, and a real Macbook Pro, and on the macbook the GPU usage is 100%.

Which leads me to understand that everyone is on the same version of macOS and Davinci Resolve, but only the Macbook actually uses the GPU to render...

Did you read this?