Since few days I have two Vega 64, and have some problem with FCPX 10.4.8 on 10.15.2, but everythings works well on 10.14.6. The problem is in BruceX. Under 10.15.2, graphics cards crushes during rendering. Does anybody have similar situation?
Good morning Mr sasamedovarski
My actual FCPX Mac setup computer (iMac i9 2019 Catalina 10.15.2) has two GPU: on board Vega 48 and Vega 56 eGPU. I haven't yet see any problems in FCPX. Just opposite: works fast and great!!!
I love last FCPX feature of second GPU for renders and share
sasamedovarski wrote: It‘s not bad result, but I excpected about 5 sec on that configuration.
On my 10-core Vega 64 iMac Pro running 10.4.8 and Mojave 10.14.6, the 5-run average of BruceX is 4.9 sec per individual BruceX timeline range.
Note: BruceX should be exported to a ProRes 422 file, not an H.264 file. That could account for some differences.
Also BruceX is so fast on recent Macs starting with 10.4.7 that you cannot get an accurate figure unless you append the BruceX timeline to itself 5 or 10 times, run it and divide the execution time by 5 or 10. That is much better than doing a bunch of 5-10 sec runs and trying to average those by hand. There's too much variability that way.
Even considering that, it's only one test. It doesn't mean dual GPUs aren't being used for real world workloads. It might be more meaningful to do a Neat Video test with that configured to use only GPU processing. However that would require standardized Neat Video versions, test clip and standardized parameters.
The easiest way using Neat Video is using a test clip which everyone could download, position on frame 1, open the Neat Video interface and select Tools>Preferences>Performance>Optimize Performance. It will run a quick benchmark and list the achieved frame rate using all combinations of CPU, GPU, inc'l CPU-only and GPU-only.
My 4-core 2017 i7 iMac 27 with 32GB RAM and 8GB Radeon Pro 580 running 10.4.8 on Mojave 10.14.6 did BruceX in 9.0 sec, averaged over five runs. Method: append 5 BruceX timelines, export to ProRes 422, divide elapsed time by 5. That is roughly 1/2 the performance of my 10-core Vega 64 iMac Pro, which is roughly consistent with the difference in GPU performance.