Testing our GPU in OSX/Linux/Windows in pro app (Blackmagic DaVinci Resolve Studio /AE/PR Pro,PS)

[fabiosun]

In this thread is possible to post  benchmarks with owned GPU ('s)

Candle benchmark for DaVinci Resolve Studio

 

 

Thread is open also for unsupported GPU in OSX like 6900 xt, 6800 Xt, 6800 (we hope in a near support for them) but also it is possible to post results for Nvidia GPU like RTX series that will never be supported by OSX (Sob 🙂)

I will start with mine GPU, you can see in this video

On YouTube a simple video showing initial Davinci resolve setup (OpenGL / Metal or Cuda if it is available)

If you like you can post your results using a screenshot for frame rate in different environment you use

It is welcome any new GPU benchmark (AMD rx 6900 xt and similar 🙂 🙂 )

 

You must play inside Color page and take note only there of your maximal FPS

 

 

Attached useful file for this benchmark

Candle benchmark.zip

[fabiosun]

Gigabyte Radeon Vega 64 Gaming OC 8 GB

 

OSX 11.1 (20C69)

Spoiler

 

Metal:

66 noise nodes

14/15/16 fps

 

6TRN nodes

19/20/21

 

OpenCL:

66 noise nodes

7/8/9 fps

 

6TRN nodes

6/7

 

 

Windows 10 Pro 64 Bit

Spoiler

 

OpenCL:

66 noise nodes

12/13 fps

 

6TRN nodes

17/18/19

 

 

[Arrakis]

 

Gigabyte Radeon Pro WX 9100 16 GB

 

OSX 11.1 (20C69)

Spoiler

Metal:

66 noise nodes

13 /14 fps

 

6TRN nodes

19 / 20 fps

 

OpenCL:

66 noise nodes

7 / 8 fps

 

6TRN nodes

9 / 10 fps 

 

Windows 10 Pro 64 Bit

Spoiler

OpenCL:

66 noise nodes

10 / 11 fps

 

6TRN nodes

17/18 fps

 

 

[tomnic]

if I remember well... 😉 Sigh

 

XFX Radeon VII 16 Gb

 

macOS 11.0.1

Spoiler

Metal

66 blur nodes

20 fps

 

6 TRN nodes

24 fps

 

 

[tomnic]

The same for my actual temporary 1080ti in macOS High Sierra, cough cough 😛 

 

MSI GeForce GTX 1080ti Aero 11Gb

 

macOS 10.13.6

Spoiler

 

CUDA

66 blur nodes

15 fps

 

6 TRN nodes

21/22 fps

 

 

 

Windows 10 Pro 64 bit

Spoiler

 

CUDA

66 blur nodes

14 fps

 

6 TRN nodes

19 fps

 

 

[Ploddles]

Davinci Resolve 16.2.8 Free Edition

 

Gigabyte RX 6900 XT

 

Windows 64 Pro (OpenCL)

 

66 Blur Nodes

20-22 fps

 

6 TRN Nodes

30-31 fps

 

Big Sur - won't run atm, no drivers (yet?)

[fabiosun]

@Ploddlesmaybe non studio version does not apply correctly filter used in this test

not sure

but I have a doubt

[fabiosun]

however, if the @Ploddlesresults are confirmed, making the due proportions of how much the AMD graphics card degrades with OpenCL and how much it performs well in metal .... If apple produces a driver ... there will be fun 🙂

[fabiosun]

 AMD Radeon Rx 6900 XT 16 GB

 

OSX 11.1 (20C69)

Spoiler

 

Metal:

66 noise nodes

SOB

 

6TRN nodes

SOB

 

OpenCL:

66 noise nodes

SOB

 

6TRN nodes

Sigh :)

 

 

Windows 10 Pro 64 Bit

Spoiler

 

OpenCL:

66 noise nodes

23/24 fps

 

6TRN nodes

33/34

17_Beta 6

Software is unrensponsive..not production ready with new GPU in my case

 

 

 

[jsq2627]

Sapphire RX 5700 XT 8G

Davinci Resolve 17.1 Beta 7 (Free Edition)

 

macOS 11.1

Spoiler

 

Metal

66 blur nodes

15 fps

 

6 TRN nodes

20 fps

 

OpenCL

66 blur nodes

8 fps

 

6 TRN nodes

10 fps

 

 

Edited January 28, 2021 by jsq2627
post davinci version

[fabiosun]

@6900 or 6800xt users😂

could try candle davinci benchmark with resize bar on and off and see if benchmark results are better or lower?

I am trying with a rtx 3090 in Windows and without resize bar on they are pretty similar of mine and @Ploddlespreviously posted results

thanks

[fabiosun]

Asus OC strix rtx 3090 (thanks @thenightflyer)

 

Windows 10 Pro 64 bit

Spoiler

OpenCL: (resize bar Off)

66 noise nodes

27 fps

 

6TRN nodes

34

 

Davinci 17.1

Software is working as it should

 

On 3/8/2021 at 4:04 AM, meina222 said:

@meina222can't see this image. 

[Arrakis]

😀 Just for information, since version 17.1 of Resolve Studio, there is no longer a need for the patch (Matlabs). In addition, the "libiomp5.dylib" file is no longer present in the application library and therefore inoperative.

[fabiosun]

Thank you for confirming this @Arrakis

in all previous 17.x betas were the same

[meina222]

Finally got a chance to retest the 3090 FE (single GPU as I don't have studio but used free version 17.1).

 

66 Blur nodes - 27-28fps 

6 TRN nodes - 31-32 fps.

[fabiosun]

[meina222]

@fabiosun, didn't know how to include image in IM. Lmk, and I can benchmark this puppy.

 

[fabiosun]

beatiful 🙂

 

you should try your two inserted yet beasts 🙂

 

[fabiosun]

 

Check GPUz memory temperature....😅

[wongdi]

davinci resolve studio 17.1.1

 

big sur 11.2.3

 

amd radeon pro wx8200

 

Metal: 13~14 fps

OpenCL: 8~9 fps

 

 

Metal

 

 

OpenCL

[fabiosun]

to complete without error this production Davinci Resolve Benchmark i have had to put a fan pumping fresh air to 3090 FE backplate

due huge GPU memory temperatures during this test..

results are great

[fabiosun]

[fabiosun]

Come on guys! help me repent and get back with AMD🙂

 

[fabiosun]

Thanks @valmeida 's great patience, we were able to do some tests via remote access to his machine

the good news is that some refresh issues in windows have also been fixed

the bad news is that for the moment the metal driver is not as performing as it should

At the moment the results are about

21/24 frames with 6 nodes

17 with 66 nodes

I also tried working with 8k red videos and the processing runs smoothly  in windows as in OSX

In windows we have results compatible with the other tests made in the past by me , @meina222and @Ploddles

 

For now, in OSX the Candle Benchmark results are comparable with a vega 64 previously owned by me

It is very weird because others aspect in editing and coloring are identical in windows

 

to give more datas and infos about my simple testing:

 

I have used three footage in 8k resolution on a multi layer timeline (RED camera footage)

Playback was set @ 25 frame/s

system is playing  @ right and constant speed i have set in project settings

we hope in future driver enhancement in OSX

 

 

[fabiosun]

again thanks @valmeida

we tested also in beta 2

No significative improvement in Candle Benchmark for now