MP 7,1 Final Cut Pro with multi 3rd party GPU utilization

[awkward_eagle]

Just installed 2x Radeon Pro WX 9100s today after removing the 580X. Opened a 4K FCP project with source files being 4.6k ProRes 4444 footage shot on an Ursa Mini Pro. Turned off background render and deleted generated renders for the sequence. Turned viewer to Better Quality.

GPU memory on both cards filled up pretty fast but the processors were barely being used at all. Handful of clips use Optical Flow retiming but still played without stuttering. This is without any effects applied. Excited to see how FilmConvert Nitrate and Neat Video work. It's amazing how well ProRes / FCP are optimized on this hardware with the new Metal rendering engine.

It's been a while since I've done any grading in Resolve but my memory is that when it goes to cache the timeline to ProRes is uses maxed out GPU compute. When rendering live corrections I think it just uses as much as it needs. Will jump in and start testing Resolve soon.

 

[Casey P]

Yeah, FCPX has really been optimized to not require heavy processing in normal use. I actually wish there were some options to allow continuous background rendering to use all of the extra power available.

 

[awkward_eagle]

Given how well this runs with 2x gpus, makes me think I don't need an Afterburner. That might change when doing grading or multicam.

 

[mtnbikfrut]

This thread has been insightful. I have a 580x in my 2019 MP. I have lots of ram, the 16core chip. I'm looking to speed up rendering and exporting. I can purchase the 5800X from apple business for $900. Is this the route I should go? Think renders will be about 2x faster with double the ram and double the teraflops ?

 

[h9826790]

This thread has been insightful. I have a 580x in my 2019 MP. I have lots of ram, the 16core chip. I'm looking to speed up rendering and exporting. I can purchase the 5800X from apple business for $900. Is this the route I should go? Think renders will be about 2x faster with double the ram and double the teraflops ?

For hardware encoding, the CPU performance and memory size doesn't really matter (unless it's way too low).

The bottleneck is on the GPU's video engine.

I am still using the 5,1. The Radeon VII's video engine can do 2x faster than the RX580.

Activate AMD hardware acceleration

Thanks, you just remind me that I can make a performance comparison of the RX580 and Radeon VII (of course, about the HWAccel performance, but not gaming, etc). Both done on my cMP, same CPU, same RAM, same SSD, just the GPU is upgrade, and Mojave minor update. Both tests convert the same Sony...

forums.macrumors.com

So, I expect if you move to the Vega / Navi, you should able to get better encoding performance as well (if you use hardware encoding).

 

[awkward_eagle]

This thread has been insightful. I have a 580x in my 2019 MP. I have lots of ram, the 16core chip. I'm looking to speed up rendering and exporting. I can purchase the 5800X from apple business for $900. Is this the route I should go? Think renders will be about 2x faster with double the ram and double the teraflops ?

Did you mean the W5700X? Word is the Navi drivers are unstable, causing panics and power management problems. The 16gb of vram will definitely be used by FCP. I’m sure it’ll get sorted out but it might be a while.

Drivers for the Vega cards are solid. I have a pair of WX9100s in one system and a pair of the Pro Vega II MPX in another. Both have been solid and haven’t shown any of the problems others have reported with the W5700X. WX9100s still a good option, but a little overpriced compared to the W5700X, but more stable.

Pro Vega II MPX still best overall option but price is high if that’s enough of a barrier.