Why FCPX is not faster with RX580 8gb

[carambo]

Hi, I have upgraded my MP with a new RX580 8gb Pulse, to replace my old 5870, and I am very surprised that actually I don't see any benefit in FCPX. I thought a 8gb RAM in the video card would allow to process more frames and speed up the editing process, particularly when adding effects, compare to the 1gb in the 5870, but no
I am on High Sierra with 32go RAM.
Anyone has explanations about this ? And any recommendations how to leverage this new card, like using another editor (Premiere,...) ? Txs

 

[Socket]

have you tried a BruceX Test?
you have both card installed at the same time? If yes, you must swap out the old one, because FCPX "work" with the lower one (don't know why)...if you have installed only the 580, then it works better

 

[carambo]

have you tried a BruceX Test?
you have both card installed at the same time? If yes, you must swap out the old one, because FCPX "work" with the lower one (don't know why)...if you have installed only the 580, then it works better

No I have only the RX580 installed, but FCPX is not better, or may be marginally, but I was expecting much more improvements

 

[namethisfile]

Hi, I have upgraded my MP with a new RX580 8gb Pulse, to replace my old 5870, and I am very surprised that actually I don't see any benefit in FCPX. I thought a 8gb RAM in the video card would allow to process more frames and speed up the editing process, particularly when adding effects, compare to the 1gb in the 5870, but no
I am on High Sierra with 32go RAM.
Anyone has explanations about this ? And any recommendations how to leverage this new card, like using another editor (Premiere,...) ? Txs

I have a similar sentiment. I recently acquired a Pulse RX 580, as well. I came from RX 460 and I didn't see my experience in FCPX snappier, faster or smoother between RX 460 and RX 580. Even, though, when running Geekbench 4 OpenCL and Unigine benchmarks shows the RX 580 trouncing the RX 460... under FCPX, there doesn't seem to be any observable difference.

This is when applying transitions, scrubbing on the timeline, etc. Also, export time doesn't improve, as well, using H264.

So, I feel like I wasted $240+tax+shipping on the RX580.... Is there a "Doh!" emoticon? If so, ***Doh-moticon***

 

[h9826790]

RX580 is mainly to improve rendering.

For time line, as long as using ProRes, that should be storage speed limiting. And if not using ProRes, and directly editing H264 / H265, then Apple doesn't let the RX580 do hardware decoding in FCPX yet. That's why it can't be smooth.

For export, encoding is purely CPU job, upgrade GPU won't help anything. Again, Apple not allow us to use the RX580's hardware encoding in macOS yet.

 

[carambo]

RX580 is mainly to improve rendering.

For time line, as long as using ProRes, that should be storage speed limiting. And if not using ProRes, and directly editing H264 / H265, then Apple doesn't let the RX580 do hardware decoding in FCPX yet. That's why it can't be smooth.

For export, encoding is purely CPU job, upgrade GPU won't help anything. Again, Apple not allow us to use the RX580's hardware encoding in macOS yet.

Txs for the info. To your knowledge, is it the same with Premiere or another editor, or is it related to the OS and Mac hardware, whatever is the software ?

 

[h9826790]

Txs for the info. To your knowledge, is it the same with Premiere or another editor, or is it related to the OS and Mac hardware, whatever is the software ?

OS limiting. The same card on cMP in Windows can do H264 and H265 hardware decoding / encoding flawlessly.

 

[slamjack]

Same here. In general, i edit jpeg or PNG image sequences in FCPX (cartoon animatics) — timeline is not smooth, it is laggy, dragging and dropping is laggy, although rendering codec of the project is 422 LT, and all the images are compressed and held on PCI-e SSD.

RX580 definitely has bottle neck, caused by OS now, and Mojave did not change anything.
What modern cards (nvidia, vega 56?) don't have these limitations?

 

[thomasthegps]

Apple doesn’t want an old mac to compete with their newer machines so they have purposefully disabled hevc encode/decode on the gpu. Fcx and other apps cant utilize the gpu features if they aren’t exposed by the os...

Don’t expect any other card to offer better performance because apple has made it difficult for nvidia to use the gpu decode api, and all other amd cards use the apple driver that purposefully disables gpu features... (This has been proven with the 10.14.1 update. The RX 580 had hevc decoding abilities in 10.14.0 but apple decided to disable it in 10.14.1.)

This is by definition programmed obsolescence.

 

[smirk]

This is by definition programmed obsolescence.

Programmed obsolescence would be if older computers blocked hardware decoding, while newer ones enabled it. You would naturally be prompted to buy a new machine to get the added performance.

 

[thomasthegps]

Programmed obsolescence would be if older computers blocked hardware decoding, while newer ones enabled it. You would naturally be prompted to buy a new machine to get the added performance.

Which is what apple is doing... the hevc support in 10.14.0 probably cam from the imac pro. As soon as apple realized that decode functionality worked for the rx 580 they disabled it...

 

[kohlson]

have you tried a BruceX Test?

In my experience BruceX is not a good indication of GPU-FCPX performance.

 

[h9826790]

In my experience BruceX is not a good indication of GPU-FCPX performance.

BruceX itself has nothing wrong. However, you can't use it as indicator for H264 export benchmark. The test was decided to measure rendering speed, and only export in ProRes.

If you want to measure rendering + export H264, you can always mod the process, so that FCPX will export the BruceX project in H264, and you can time it accordingly.

If you said BruceX is not accurate because it can't reflect the H264 export time.

It's like running an OpenCL test and complain it's not a good indicator for DirectX performance.

 

[ProTools808]

Apple doesn’t want an old mac to compete with their newer machines so they have purposefully disabled hevc encode/decode on the gpu. Fcx and other apps cant utilize the gpu features if they aren’t exposed by the os...

Don’t expect any other card to offer better performance because apple has made it difficult for nvidia to use the gpu decode api, and all other amd cards use the apple driver that purposefully disables gpu features... (This has been proven with the 10.14.1 update. The RX 580 had hevc decoding abilities in 10.14.0 but apple decided to disable it in 10.14.1.)

This is by definition programmed obsolescence.

Hi there,
Same situation here! May I ask you: If I get "Downgrade" the OS X...I mean if I get back to High Sierra (10.13.6) will be faster FCPX with my "old" nVidia GTX 970 4GB GPU+CUDA driver versus RX580?

 

[h9826790]

Hi there,
Shame situation here! May I ask you: If I get "Downgrade" the OS X...I mean if I get back to High Sierra (10.13.6) will be faster FCPX with my "old" nVidia GTX 970 4GB GPU+CUDA driver versus RX580?

Absolutely NOT.

Even my 7950 can kill my GTX 1080Ti in FCPX.

FCPX won't use CUDA at all. And not optimised for Nvidia GPU.

My above reply was in 2018. And in 2019, I found a way to activate the hardware encoding ability for cMP in macOS.

Activate AMD hardware acceleration

[For existing OpenCore users, you can always download my latest OpenCore package at post #1314 for update] Full HWAccel achieved now! Everything is covered in the OpenCore thread. The step by step tutorial to enable AMD hardware acceleration is now at here, but I strongly recommend you go...

forums.macrumors.com

And this will make your RX580 able to export H264 / HEVC in FCPX. That will be way faster than CPU encoding (even you have dual X5690).

 

[ProTools808]

Hey Mr

h9826790,​

Thanks for your answer, but unfortunately I'm totally not understand this:

Activate AMD hardware acceleration
[For existing OpenCore users, you can download the latest OpenCore 0.6.3 package at post #1314 for update] Full HWAccel achieved now! Everything are covered in the OpenCore threads. The step by step tutorial is at here, but I strongly recommend you go through the whole thread before you do...

forums.macrumors.com

...I'm absolutely newbie about this topic :S

​

 

[ProTools808]

Even my 7950 can kill my GTX 1080Ti in FCPX.

Do you mean, the 7950 is faster then GTX 1080Ti?

 

[ProTools808]

...and how about 7950 3GB or 7970 3GB vs RX580 8GB?

 

[h9826790]

Hey Mr

h9826790,​

Thanks for your answer, but unfortunately I'm totally not understand this:

Activate AMD hardware acceleration
[For existing OpenCore users, you can download the latest OpenCore 0.6.3 package at post #1314 for update] Full HWAccel achieved now! Everything are covered in the OpenCore threads. The step by step tutorial is at here, but I strongly recommend you go through the whole thread before you do...

forums.macrumors.com

...I'm absolutely newbie about this topic :S

​

What you don't understand?

"For existing OpenCore users...", since you are not existing user, you can ignore that sentence.

"Everything are covered in the OpenCore threads...", the OpenCore thread link is there, you can go to there an have a look. However 'I strongly recommend you go through the whole thread..." means that's not mandatory. If you don't want to read anything, you can skip that, but still can get HWAccel works.

"The step by step tutorial is at here...", you can click on that link, then follow the step by step guide at there to enable HWAccel.

 

[h9826790]

Do you mean, the 7950 is faster then GTX 1080Ti?

Yes, in FCPX

 

[h9826790]

...and how about 7950 3GB or 7970 3GB vs RX580 8GB?

7970 is a bit faster than 7950, RX580 is about 100% faster than 7950

 

[ProTools808]

What you don't understand?


If you don't want to read anything, you can skip that, but still can get HWAccel works.

"The step by step tutorial is at here...", you can click on that link, then follow the step by step guide at there to enable HWAccel.

Please don’t get me wrong, I’m all interested in what you describe, but unfortunately it’s too complicated for me.
...but thank you for your help

 

[h9826790]

Please don’t get me wrong, I’m all interested in what you describe, but unfortunately it’s too complicated for me.

That's why I created that "easy to install package".

You still have to disable SIP by yourself, learn how to mount EFI partition, etc. But I insist that's mandatory for any OpenCore to learn. Just in case, if anything goes wrong, the user still has the basic knowledge to perform some necessary recovery actions.

Anyway, if you want to play safe, after you download my OpenCore 0.6.3 package, you can follow the steps in this video to install OpenCore onto a USB thumb drive. So, if that doesn't work as expected, you can simply pull out the USB drive, then nothing is changed.

Of course, you don't need to install Big Sur (which the video just want to demo that my package also enable Big Sur support on the 5,1). But if you want to run the latest FCP, you have to go beyond Mojave (e.g. Big Sur).

 

[OkiRun]

Please don’t get me wrong, I’m all interested in what you describe, but unfortunately it’s too complicated for me.

It is time you bought new mac.

>++<

 

[ProTools808]

It is time you bought new mac.

>++<

x)) yep, but unfortunately I living in Hungary ... this machine is extremely expensive in Hungary! You have to imagine getting it in USA for $ 100,000 ...Like this !! It's nonsense!
The salary in Hungary around 500, 600 euro monthly. When you paid the bills and food maybe you can save 100-200 Euro in a month..
...so, if I want a new Mac Pro, I have to wait some years