Weird rendering/exporting issue

Two identical partitions (on the same SSD) Same Mojave version, (10.14.6) exactly same Fcp version (10.4.10) and BruceX is taking half the time on one on them !  Half 

(Monitor doesn't show any background process on the slow one. (ICloud, Drop Box, Time Machine etc disabled...)  

On this machine I have a RX 580 on a "old" cheesegrater MacPro "vitamined",  not very "official " but nevertheless It should be the same. 

The only "special" thing I did on the fast one was an installation of Avid Media Composer... shouldn't be related.

Is it possible to have a "bad" Metal access on some cases.

I updated the slow one to Catalina... the same.

Stu Wart wrote: ...Same Mojave version, (10.14.6) exactly same Fcp version (10.4.10) and BruceX is taking half the time on one on them !...

If the exact same procedure is not followed, sometimes BruceX can have widely varying run times. Examples:

(1) Background rendering should be disabled, otherwise export time may vary based on how long it auto-rendered. If testing two systems and only one has background rendering disabled, this could make a big difference.

(2) Export should be to ProRes 422 not H264 or any other codec. Otherwise it is testing encode performance (not render performance) which can vary greatly based on hardware.

(3) Time Machine and (ideally) Spotlight indexing should be disabled. Normally this doesn't make a huge difference but it can introduce variability. To disable TM, click on the icon in the menu bar, open TM preferences and clear the checkbox "Back Up Automatically". Remember to re-enable that when finished testing.

To disable Spotlight indexing, add each drive (including system drive) to the privacy tab of the Spotlight app in System Preferences. When finished testing remove those from the privacy tab. This will also rebuild all the indexes: support.apple.com/en-us/HT201716

(4) Re-boot MacOS before starting a test sequence, and wait a few min before doing the test. I realize you may have already done this due to separate partitions.

(5) On newer machines where BruceX has a total run time of less than about 20 sec, it is better to append 5 BruceX timelines, do the timed test, then divide the result by 5. Even with all of the above procedures, there can be some run-to-run variation of a short test. Doing many individual short runs and taking the average is probably less accurate than doing one longer run of a 5x timeline. After appending the timeline 5x, do a SHIFT+Z to visually verify it is actually 5x not anything else.

Thank you but

1- "Gollner rules" are respected , (no BG render , etc)

2- I am exporting 1920x1080 ProRes 422 for both

3- Everything in background  disabled in both case

4- Reboot of course


I remember the fast one went "OS 10.14.6 right away" (fresh install)  and the slow one first 10.14.4 and then update to 10.14.6 

Stu Wart wrote: ...I remember the fast one went "OS 10.14.6 right away" (fresh install)  and the slow one first 10.14.4 and then update to 10.14.6 

Verify they are both on the latest (and same) version of Apple Pro Video Formats. Mis-matched versions of PVF can make a big performance difference for some tasks.

Starting with Mojave that updated in System Preferences>Software Update. Before Mojave you had to download that from Apple.

To check the PVF version, do About This Mac>System Report>Software>Installations, then scroll down to Pro Video Formats and note the highest installed version number. 

Nope, not Video Pro Format related 

Run both GeekBench CPU *and* GPU test (Open CL and Metal versions) on each partition's MacOS and see if there's any difference. Maybe there is some difference at the system layer, below FCPX.

Geekbenches  are the same ("slow partition" with Fcp ) even performing better