Heavy CPU Usage

Hello,

Context:

Primary Millumin Machine:

-Mid 2015 Retina MBP

-16GB RAM

-Radeon R9 M370X GPU

-2.8 GHZ CPU

-1TB SSD

Backup Millumin Machine:

-Mid 2015 Retina MBP

-16GB RAM

-Radeon R9 M370X GPU

-2.5 GHZ CPU

-500GB SSD


OSX installed on the same day last year and tweaked the same on each machine.

Note the differences between the two machines are CPU speed and SSD storage. When the project is set to a single 1920x1080p60 output and playing a 1920x1080p30 video, the Primary machine's CPU is racing super high with VTDecoderXPCService between 130-260% and Millumin 3 at around 150% and FPS drops to around 30, whereas the Backup machine is running at a relaxed pace with FPS at around 60. I've reset the machines and this persists. Not much else is running on the machines. 

What could be happening? Why is the slightly heavier duty machine struggling?  

See photos. 

Comments

  • *Neither Chrome nor Safari are allowing me to post photos or attachments.

  • "Request failed with status code 403."

  • Hello @RoggieRog,

    Indeed, you machine should run easily such a project.

    One explanation would be that the first machine is getting hot, because of dust in fans for example. In this case, the GPU alters its performances to keep a temperature below 90°.

    A few questions :

    • When you hit the button "Optimize", what recommandations are displayed ?
    • Do you have the problem if no output is connected to your Mac ?
    • What is your codec ? ProRes 422 ?
    • Could you try to convert your files to HAP codec (use AVF Batch Converter if needed) ?
    • What is your version of Millumin ? 3.18.g ?
    • What is your version of macOS ? 10.15.7 ?
    • What is your graphic card ? Geforce 750M ?


    Thank you for your answers.

    Best. Philippe

    PS : the upload bug on the forum is fixed

  • Hi Phillippe,

    Here are screenshots answering most of your questions:

    Primary machine GSI-MBP-01:


    Backup machine GSI-MBP-02:


    -No issues without external display connected. Issue begins once display is on/connected.

    -ProRes 422. I'll convert to HAP and see if there's a difference. Again, the oddity is that the more powerful machine is the one that's taxed a LOT more.

    I just opened up and cleaned both machines with compressed air. The primary was surprisingly dustier than the backup. Nonetheless, the issue persists.

  • It seems as though my last post was deleted after an edit. Here it is again showing HAP encodes versus ProRes on the same machines. The HAP really brings down the heavy CPU usage compared. That said, I understand that HAP is decoded by the GPU whereas ProRes by the CPU, but again, why is my heavier duty machine struggling compared to the lighter duty machine?

    Primary machine GSI-MBP-01 with ProRes:


    Primary machine GSI-MBP-01 with HAP:


    Backup machine GSI-MBP-02 with ProRes:


    Backup machine GSI-MBP-02 with HAP:


  • Hello @RoggieRog,

    Not sure to understand : do you have a problem (of FPS) if you use HAP codec ?

    For info, this codec is mostly decoded by GPU (thus it's efficient) while ProRes-422 is decoded mainly by CPU (via VTDecoderXPCService process). Of course, Millumin is mainly using the GPU.

    As your 2 machines are very close (the CPU being not that different), the performances should be close as well. The difference of CPU usage for VTDecoderXPCService between the 2 machines is kind of puzzling. Sorry, I've no clue so far.

    A few more questions :

    • If no display is connected, how VTDecoderXPCService usage compares between the 2 machines ?
    • If you switch the displays between the 2 machines (using backup's monitor with main machine I mean), what are the results with Millumin ?
    • What is the usage of VTDecoderXPCService if you play a 4K ProRes-422 movie with Quicktime (without and with displays connected) ?
    • Check the System Preferences / Displays to spot any difference between your 2 machines : do you find some ?

    Best. Philippe

  • Hi Phillippe,

    -With no display connected, playback in Millumin has no heavy affect on CPU's. It only begins with external monitor connected.

    -Switching displays does not affect performance.

    -When playing back a UHD 29.97 ProRes file in QT with no monitors connected, the performance is nearly the same--the backup machine with the slightly slower CPU has a higher CPU percentage use. Although, when playing back with external monitors connected, the more powerful machine has over 600% CPU usage versus around 200% for the slower machine. This tells me that it's not Millumin, but something with the machine or the something with the OS.

    Whereas when playing back the same UHD 29.97 file in both ProRes and HAP in VLC, whether or not a monitor is connected, the CPU usage is nearly the same on both machines--the backup machine with the slightly slower CPU has a higher CPU percentage use.

    -Display prefs are identical. Also, automatic GFX switching is off on both machines.

    -GSI-MBP-01 primary machine:


    GSI-MBP-02 backup machine:


    Thank you very much for following up. I don't expect you to spend your time assisting me with this troubleshooting any further as it's not a Millumin issue.

  • Hello @RoggieRog,

    Indeed, I don't think the problem comes from Millumin.

    I would recommend to reinstall macOS from scratch (be sure to format your drive first). Also, you can have a try with macOS Catalina or Big Sur to see if it changes something.

    Lastly, if the issue persists, ask Apple to call you (see their website) since it may be a hardware issue.

    Let us know the results.

    Best. Philippe

Sign In or Register to comment.