Contribute
Register

Enabling igpu alongside RX580 for better Handbrake performance

Status
Not open for further replies.
@miliuco
Boring race anyway!
OK, I have just disabled igpu in bios and reset NVRAM and switched to the imacPro EFI you gave me. Machine has booted fine.

Running the exact same test as before, and this time the CPU threads are all below 25% and the GPU is showing utilisation, albeit quite low (average 13.9%).

The average FPS is slower than before at 37 and overall time predicted to around 10-15 mins more (bearing in mind this is a small - 4.4GB - source clip).

Not sure now whether to revert to iMac19,1 and re-enable my igpu as that did give the higher frame rate (not that Handbrake is my primary goal on this machine, just a side consideration).

Leaving aside the Handbrake issue, would you say there is any benefit or difference in running the machine as iMac19,1 compared to Imacpro1,1 in everyday useage? Thanks
 
Last edited:
@miliuco
... Not sure now whether to revert to iMac19,1 and re-enable my igpu as that did give the higher frame rate (not that Handbrake is my primary goal on this machine, just a side consideration).

Leaving aside the Handbrake issue, would you say there is any benefit or difference in running the machine as iMac19,1 compared to Imacpro1,1 in everyday useage? Thanks
At least EFI folders work fine on your PC, it was inexplicable that you couldn't use iMac19,1 because it is a PC very compatible with that SMBIOS.

To change from iMacPro to iMac you have to do the reverse. Being in macOS, you replace one EFI with another - restart - DEL key to enter BIOS - iGPU active - AMD continues as main card - Save and exit and stop in this dialog with the pointer over Yes - change monitor cable from iGPU to dGPU - click the mouse button to act on Yes - it restarts as iMac19,1.

Both SMBIOS are valid for your PC, it is assumed that iMac manages CPU power better (for being CPUs of the same generation) but it is the theory, in practice I have not noticed differences.
What I have seen is that on GeekBench, both Metal and OpenCL, the RX 580 achieves somewhat higher scores with iMac than with iMacPro.
 
@miliuco
Many thanks - I'll switch it back to imac19,1 EFI then.
An interesting experiment which has demonstrated (and reaffirmed to me as @qbe stated) that Windows based encoding using NVENC appears to be the fastest option.

Can't thank you enough for your assistance and time on this though. Much appreciated!
 
@miliuco
Boring race anyway!
OK, I have just disabled igpu in bios and reset NVRAM and switched to the imacPro EFI you gave me. Machine has booted fine.

Running the exact same test as before, and this time the CPU threads are all below 25% and the GPU is showing utilisation, albeit quite low (average 13.9%).

The average FPS is slower than before at 37 and overall time predicted to around 10-15 mins more (bearing in mind this is a small - 4.4GB - source clip)...
I have done 3 tests with 3 different SMBIOS:
  • iMac18,1 + iGPU as main card
  • iMac19,1 + iGPU headless mode + AMD as main card
  • iMacPro1,1 + iGPU disabled.

I only see the iGPU working converting video from Handbrake in iMac18.1 mode and never above 15%, with the CPU working at 80%.

On iMac19,1, only if Handbrake encodes with VideoToolBox I see the dGPU working at 12% (like you) with the CPU at 70% (H264) or 20% (H265). If Handbrake encodes without VideoToolBox, Radeon remains at 0%.

In iMacPro1,1, if Handbrake uses Videotoolbox I see the dGPU go up to 12%, the behavior is very similar to iMac19,1 with less FPS in iMacPro.
 
@miliuco
Aside from the CPU utilisation difference in your test, it appears that my results were not a weird anomaly then, which is a relief.
Thanks again.
 
@miliuco
Aside from the CPU utilisation difference in your test, it appears that my results were not a weird anomaly then, which is a relief.
Thanks again.
You are wellcome.
 
re tests

as x264/x265 encoder/decoder is only part of a chip i always asumed its never gonna tax your gpu fully. Filters and effects are different than en/decoders.

What is important is the time/fps.

cpu is gonna be used as gpus arent as flexible as cpus so cpu still has to do some (post)processing depending on your settings. thats why cpu quality is usually better, allows for much much more settings.
i work in video and my job involves handbrake only for preview use, i havent done the quality test thoroughly for videotoobox.
using fcpx>handbrake(cpu) workflow results in better overall quality at smaller sizes than just straight fcpx exports (using gpu). this can of course be different if you use handbrake gpu (with some cpu processing). especially on hack :)

just did a speedtest, with handbrake using same HD source and output settings:
x264 cpu - 113fps
x264 videotoolbox - 210fps
x265 cpu - 47fps
x265 videotoolbox - 200fps

with videotoolbox cpu is around 50%
this is using 5700xt
 
@qbe
Wow those numbers are awesome.
 
@qbe
Those numbers are with the PC in your signature? Awesome!!!
 
yes, but that all depends on handbrake settings ofc
but interesting in results is that going x265 with gpu would be probably 3-4 times faster than cpu, which is a lot of time saved
 
Status
Not open for further replies.
Back
Top