Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MrRedfield

macrumors member
Original poster
Jul 8, 2019
30
4
Hello,

I currently have a EVGA Geforce GTX 680 GPU and final cut pro x seems to export very slow. 4 minutes for a 1 minute 4k clip. In Adobe premier pro, same clip takes 1.5 minutes. I believe I have the same settings in both programs. (Screenshots below).

I also notice that Adobe uses more CPU then FCPX during export.

is FCPX more optimized for AMD cards? Would I noticed a big difference with a RX580 vs my current GTX680?

Thanks!

Screen-Shot-2019-07-19-at-12-13-15-AM-1.png
Screen-Shot-2019-07-19-at-12-11-54-AM.png
 

Ritsuka

Cancelled
Sep 3, 2006
1,464
968
Which Mac are you using?

Anyway, you are probably comparing the performance of the H.264 encoder. You should export to a less compressed format like ProRes to see the actual difference in the render performance.
 
  • Like
Reactions: keysofanxiety

MrRedfield

macrumors member
Original poster
Jul 8, 2019
30
4
Hi, Thanks for the quick reply. I was selecting H.264, I'll try prores next.

My machine is a 12 Core MacPro (Upgraded to X5690s) running at 3.46Ghz and 24GB of memory 1333Mhz.
 

MrRedfield

macrumors member
Original poster
Jul 8, 2019
30
4
Which Mac are you using?

Anyway, you are probably comparing the performance of the H.264 encoder. You should export to a less compressed format like ProRes to see the actual difference in the render performance.


You were right, Just ran the test again. This time prores. Adobe = 01:32 and FCPX 00:40

Anyway to speed of H.264?
Thanks!
 

Ritsuka

Cancelled
Sep 3, 2006
1,464
968
There was an hack to let FCPX use the hardware encoder in your rx 580. I think the post was here on macrumors but I can't remember where right now.
 

iluvmacs99

macrumors 6502a
Apr 9, 2019
920
671
Hello,

I currently have a EVGA Geforce GTX 680 GPU and final cut pro x seems to export very slow. 4 minutes for a 1 minute 4k clip. In Adobe premier pro, same clip takes 1.5 minutes. I believe I have the same settings in both programs. (Screenshots below).

I also notice that Adobe uses more CPU then FCPX during export.

is FCPX more optimized for AMD cards? Would I noticed a big difference with a RX580 vs my current GTX680?

Thanks!

Screen-Shot-2019-07-19-at-12-13-15-AM-1.png
Screen-Shot-2019-07-19-at-12-11-54-AM.png

H.264 is a very demanding interframe codec which requires the CPU to do most of the brute force encoding and decoding. Adobe Premiere is somewhat efficient as well as with Davinci Resolve which rely on most of the cores available plus some use of the GPU to encode h.264. Which is why you are seeing some speed improvements with Adobe compared to FCPX. It is not to say that you can not speed up FCPX for h.264 encoding; it's that it needs a software called Compressor to help encode to h.264 faster and FCPX and Compressor both work seamlessly as it supposed to be. With Adobe Premiere, people use Adobe Media Encoder to encode to h.264 faster.

The problem with the GPU is that, it is not meant to handle the variable GOP interframe footage of compressed h.264 efficiently as it does with a multi-core/multi-thread capable CPU (6 cores and up) and therefore your final output may suffer in quality. GPU encoded h.264 final product, while fast is usually inferior to CPU encoded h.264 footage. So you would prefer to have a blend of CPU and GPU. The GPU portion of the encoding is for the rendering of titles and the LUTs applied, so having a faster GPU can help the overall h.264 encoding experience if you have a heavy use of renders and LUTs in your movie.

With Core i series CPUs, they have this feature called Quicksync which helps encode h.264 material much faster than the CPU process without sacrificing a lot in quality. And that is actually my approach. I use my Mac Pro 5,1 with the RX580 GPU to edit my h.264 footage and then save it in Resolve archive format into my Mac Mini server. Then I just use my Macbook Air to load the resolve archive format and then render it to h.264 via Quicksync. Starting with the Haswell CPU and onwards to Coffee Lake (late Mini 2018), you can use Quicksync to encode h.264 material and preserve quality at the same time. There is very little difference between CPU encoded vs Quicksync, but I could get 1:1 encoding with my Macbook Air without a GPU, whereas with my Dual 8 core Xeons 1:1.7 -- meaning a 1 hr 4k footage to 1080p HD takes 51 mins to encode on the Air and 1 hr and 51 mins with my Mac Pro with dual 8 core Xeons. I do not use GPU encoding alone as the quality suffers somewhat in exchange for greater encoding speed and I don't save to an intermediate intraframe codec like Pro Res from h.264 to create proxies, because doing that would introduce a generational loss. So I would prefer less conversion to preserve quality.
 
Last edited:

MrRedfield

macrumors member
Original poster
Jul 8, 2019
30
4
H.264 is a very demanding interframe codec which requires the CPU to do most of the brute force encoding and decoding. Adobe Premiere is somewhat efficient as well as with Davinci Resolve which rely on most of the cores available plus some use of the GPU to encode h.264. Which is why you are seeing some speed improvements with Adobe compared to FCPX. It is not to say that you can not speed up FCPX for h.264 encoding; it's that it needs a software called Compressor to help encode to h.264 faster and FCPX and Compressor both work seamlessly as it supposed to be. With Adobe Premiere, people use Adobe Media Encoder to encode to h.264 faster.

The problem with the GPU is that, it is not meant to handle the variable GOP interframe footage of compressed h.264 efficiently as it does with a multi-core/multi-thread capable CPU (6 cores and up) and therefore your final output may suffer in quality. GPU encoded h.264 final product, while fast is usually inferior to CPU encoded h.264 footage. So you would prefer to have a blend of CPU and GPU. The GPU portion of the encoding is for the rendering of titles and the LUTs applied, so having a faster GPU can help the overall h.264 encoding experience if you have a heavy use of renders and LUTs in your movie.

With Core i series CPUs, they have this feature called Quicksync which helps encode h.264 material much faster than the CPU process without sacrificing a lot in quality. And that is actually my approach. I use my Mac Pro 5,1 with the RX580 GPU to edit my h.264 footage and then save it in Resolve archive format into my Mac Mini server. Then I just use my Macbook Air to load the resolve archive format and then render it to h.264 via Quicksync. Starting with the Haswell CPU and onwards to Coffee Lake (late Mini 2018), you can use Quicksync to encode h.264 material and preserve quality at the same time. There is very little difference between CPU encoded vs Quicksync, but I could get 1:1 encoding with my Macbook Air without a GPU, whereas with my Dual 8 core Xeons 1:1.7 -- meaning a 1 hr 4k footage to 1080p HD takes 51 mins to encode on the Air and 1 hr and 51 mins with my Mac Pro with dual 8 core Xeons. I do not use GPU encoding alone as the quality suffers somewhat in exchange for greater encoding speed and I don't save to an intermediate intraframe codec like Pro Res from h.264 to create proxies, because doing that would introduce a generational loss. So I would prefer less conversion to preserve quality.


This is some great info. Thank you so much, really helps.

So the MacbookPro that I traded for this MacPro is faster with H.264. Crazy, guess I have to get another MBP.

Also, I did the BruceX test with the GTX 680 and my old HD 5770 (on High Sierra). Here where my results.

HD 5770 = 1:14
GTX 680 (Apple Driver) = 2:34
GTX 680 (Web Driver) = 1:21

That's crazy to me, it does seem FCP likes the AMD/ATI cards better. Guess it's time for me to get a RX580.

Thanks again
 

kohlson

macrumors 68020
Apr 23, 2010
2,425
736
There was an hack to let FCPX use the hardware encoder in your rx 580.
This: https://forums.macrumors.com/threads/activate-amd-hardware-acceleration.2180095/

@h9826790 is still quite active on monitoring this. I would encourage you to read all the posts. There are 2 ways to do this hack, each with pros and cons. I've done it on my 6-core 4/5,1, and on a friend's 2012 12-core. We're both at 10.14.5. I have had one freeze, but otherwise no issues, and have great hopes for rock-solid reliability in 14.6.
 
  • Like
Reactions: h9826790

MrRedfield

macrumors member
Original poster
Jul 8, 2019
30
4
This: https://forums.macrumors.com/threads/activate-amd-hardware-acceleration.2180095/

@h9826790 is still quite active on monitoring this. I would encourage you to read all the posts. There are 2 ways to do this hack, each with pros and cons. I've done it on my 6-core 4/5,1, and on a friend's 2012 12-core. We're both at 10.14.5. I have had one freeze, but otherwise no issues, and have great hopes for rock-solid reliability in 14.6.


Just installed the RX580 last night and installed that hack. Worked great! Night and day really. Its now so fast. Even the BruceX test was less than 20 seconds. I couldn't be happier.

Guess I'll be putting that GTX in another system now.

Everyone thank you!
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.