Radeon RX 4XX/5XX standalone system, AMDRadeonX4250.kext (GVA support H264) does not support HEVC HW

[shuhung]

Testing the Radeon RX 580 4GB hardware encoding/decoding in Mac OS 10.14.5 beta 2.
The tests were done in FCPX version 10.4.6, using the 4k video file TCL Supercars 4k Demo.mp4 HEVC size 1.33 GB, also used VideoProc, HWMonitorSMC2 and Intel Power Gadget. I renamed my DGPU as Radeon Pro 580.
The picture 1 shows which GPU is used and loading.
The pictures 2 & 3 shows the time need it for exporting.
Best exporting time is market in red.


Test 1: DGPU RX 580 enabled, IGPU disabled, SMBIOS iMacPro 1,1.
HEVC -> H264 : 4:01 pictures 1&2
H264 -> HEVC : 3:09 picture 3

View attachment 398838View attachment 398837
View attachment 398836


Test 2 : IGPU Intel UHD 630 enabled, DGPU disabled (unplugged), SMBIOS Mac mini 8,1.
HEVC -> H264 : 3:53 pictures 1&2
H264 -> HEVC : 7:04 picture 3

View attachment 398839View attachment 398840View attachment 398841


Test 3 : DGPU RX 580 enabled, IGPU Intel UHD 630 enabled, DP cable in DGPU, SMBIOS iMac 19,2.
HEVC -> H264 : 2:28 pictures 1&2
H264 -> HEVC : 6:07 picture 3

View attachment 398842View attachment 398843View attachment 398844

If want to verify export H264 performance on FCPX, use video clips more than 60 minutes for H264 export test. result will be differnce, AMD VCE should faster than IGPU + DGPU.

 

[CMMChris]

It also makes no sense to compare on different SMBIOSes. Run all these tests on each of the SMBIOSes used.
Also, renaming your card to Radeon Pro 580 doesn't do anything.

 

[justinhe]

It also makes no sense to compare on different SMBIOSes. Run all these tests on each of the SMBIOSes used.
Also, renaming your card to Radeon Pro 580 doesn't do anything.

It makes sense. Because iMac Pro 1.1 only has dGPU, macmini 8,1 only has iGPU, iMac19,2 has both. His test is very accurate.

 

[Psychilles]

It makes sense. Because iMac Pro 1.1 only has dGPU, macmini 8,1 only has iGPU, iMac19,2 has both. His test is very accurate.

The SMBIOS is the system definition right? Or am I missing something? And does that really improve performance?

 

[justinhe]

The SMBIOS is the system definition right? Or am I missing something? And does that really improve performance?

SMBIOS does not improve performance, but you have to use correct SMBIOS based on your hardware settings.

 

[mattystonnie]

Yes, its the system definition, it makes a difference for power management depending of your pc configuration, ex: cpu, motherboard,etc. For my build B360,i5 8400,Intel UHD 630, RX580.I’m using SMBios iMac 19,2.

 

[mattystonnie]

If want to verify export H264 performance on FCPX, use video clips more than 60 minutes for H264 export test. result will be differnce, AMD VCE should faster than IGPU + DGPU.

My second job is to film and edit videos, for me DGPU +IGPU for H264 was always faster compared with DGPU alone.
This is my personal choice and is what worked best,everyone is free to test and use whatever they want.


I renamed my RX 580 and Injected to belisted in PCI device list and in apps, like a real mac, doesn’t make a difference in performance.

 

[shuhung]

My second job is to film and edit videos, for me DGPU +IGPU for H264 was always faster compared with DGPU alone.
This is my personal choice and is what worked best,everyone is free to test and use whatever they want.


I renamed my RX 580 and Injected to belisted in PCI device list and in apps, like a real mac, doesn’t make a difference in performance.

Difference may be affected by CPU performance, my result DGPU is slightly faster on export H264, both FCPX and DaVinCi same result.

I mainly use DaVinci Resolve to measure H264/HEVC transcoding performance, easier for verify H264/HEVC transcoding fps and total processing time, better metal /openCL support than FCPX.

IGPU standalone, round 110 fps 1920x808 H264 transcoding speed.
IGPU + DGPU, round 163 fps 1920x808 H264 transcoding speed, extra 50 fps provided by AMD DGPU openCL/metal.
DGPU standalone, same or slightly faster than IGPU + DGPU, should higher than 165 fps.

 

[mattystonnie]

Difference may be affected by CPU performance, my result DGPU is slightly faster on export H264, both FCPX and DaVinCi same result.

I mainly use DaVinci Resolve to measure H264/HEVC transcoding performance, easier for verify H264/HEVC transcoding fps and total processing time, better metal /openCL support than FCPX.

IGPU standalone, round 110 fps 1920x808 H264 transcoding speed.
IGPU + DGPU, round 163 fps 1920x808 H264 transcoding speed, extra 50 fps provided by AMD DGPU openCL/metal.
DGPU standalone, same or slightly faster than IGPU + DGPU, should higher than 165 fps.

Maybe because UHD 630 has better suport in Mojave,Mac mini 8,1, iMac 19,2,iMac 19,2 has the UHD 630. For the test we need someone else with UHD 630.

I'll try DaVinci, Thank you.

 

[Psychilles]

Maybe because UHD 630 has better suport in Mojave,Mac mini 8,1, iMac 19,2,iMac 19,2 has the UHD 630. For the test we need someone else with UHD 630.

I'll try DaVinci, Thank you.

I have one. In the 8350K. Right now igpu is turned off. Is it as simple as turning it on in the bios?
What should I test?