Linux performance compared to macOS
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
Hello everyone
I am a 20+ year Apple user, new to Linux and trying to make a slow transition away from macOS to Linux.
At the moment I am dual booting my machine with macOS 11.1 and Manjaro 20.2.
I am already getting about 60% of my work done in Linux and so far I am very happy about that.
Something I noticed is that my u-he plugins perform worse on Linux compared with how they perform on macOS.
My system:
* Intel Xeon W-3235 12-Core CPU @ 3.30GHz
* 48GB 2666MHz
* 2TB SSD
* AMD Radeon Pro 580X
* RME Fireface UFX in CC Mode
** ALSA (jack2 is installed, but not configured)
* Manjaro GNOME 20.2
* 5.9.1-1-rt19-MANJARO kernel
* Bitwig Studio 3.3.1
** 1024 samples block size @ 44100 Hz
** Plug-in Hosting Mode: Individually
** Prefer VST 3 over VST 2: enabled
** Prefer 64 bit over 32 bit: enabled
* u-he plugins revision 10721
** multi-core setting enabled where possible
Test scenario:
I made two test projects (one with only Bitwig devices and one with only u-he devices) and ran them with the exact same settings in macOS and Manjaro. While the Bitwig only test project performed more or less identical on both operating systems, the u-he only project performed much worse on Linux.
The u-he project consists of 20 tracks with Satin, Presswerk, ACE, Bazille, Hive, Repro 5 and a return with Twangstrom and ColourCopy. On macOS there is about 50% CPU headroom left. On Linux I can only playback about 5 tracks before the DSP Performance Graph starts to touch the roof and to create crackles.
Do other users have comparable results to mine? Of course, it would be great if I can achieve comparable performance to macOS on Linux.
Here's a couple of screenshots I made:
* Bitwig only macOS: https://www.dropbox.com/s/05csr3z60dee4 ... s.png?dl=0
* Bitwig only Manjaro: https://www.dropbox.com/s/pt1drgubmzpyi ... o.png?dl=0
* u-he only macOS: https://www.dropbox.com/s/pmmlba6xv7037 ... s.png?dl=0
* u-he only Manjaro: https://www.dropbox.com/s/jmkwfpmodn7n1 ... o.png?dl=0
Many thanks for your help and the exchange in advance!
Amadeus
I am a 20+ year Apple user, new to Linux and trying to make a slow transition away from macOS to Linux.
At the moment I am dual booting my machine with macOS 11.1 and Manjaro 20.2.
I am already getting about 60% of my work done in Linux and so far I am very happy about that.
Something I noticed is that my u-he plugins perform worse on Linux compared with how they perform on macOS.
My system:
* Intel Xeon W-3235 12-Core CPU @ 3.30GHz
* 48GB 2666MHz
* 2TB SSD
* AMD Radeon Pro 580X
* RME Fireface UFX in CC Mode
** ALSA (jack2 is installed, but not configured)
* Manjaro GNOME 20.2
* 5.9.1-1-rt19-MANJARO kernel
* Bitwig Studio 3.3.1
** 1024 samples block size @ 44100 Hz
** Plug-in Hosting Mode: Individually
** Prefer VST 3 over VST 2: enabled
** Prefer 64 bit over 32 bit: enabled
* u-he plugins revision 10721
** multi-core setting enabled where possible
Test scenario:
I made two test projects (one with only Bitwig devices and one with only u-he devices) and ran them with the exact same settings in macOS and Manjaro. While the Bitwig only test project performed more or less identical on both operating systems, the u-he only project performed much worse on Linux.
The u-he project consists of 20 tracks with Satin, Presswerk, ACE, Bazille, Hive, Repro 5 and a return with Twangstrom and ColourCopy. On macOS there is about 50% CPU headroom left. On Linux I can only playback about 5 tracks before the DSP Performance Graph starts to touch the roof and to create crackles.
Do other users have comparable results to mine? Of course, it would be great if I can achieve comparable performance to macOS on Linux.
Here's a couple of screenshots I made:
* Bitwig only macOS: https://www.dropbox.com/s/05csr3z60dee4 ... s.png?dl=0
* Bitwig only Manjaro: https://www.dropbox.com/s/pt1drgubmzpyi ... o.png?dl=0
* u-he only macOS: https://www.dropbox.com/s/pmmlba6xv7037 ... s.png?dl=0
* u-he only Manjaro: https://www.dropbox.com/s/jmkwfpmodn7n1 ... o.png?dl=0
Many thanks for your help and the exchange in advance!
Amadeus
-
- KVRist
- 74 posts since 22 Apr, 2018
A few ideas for troubleshooting. I haven't directly compared Linux and Mac so I have no thoughts specific to that scenario. That being said I find things runs really well on my ancient Sandy bridge system with 8gb of ddr3 and reaper.
I'd configure Jack, it should help. Also, I've found the the multicore mode on repro worsens my performance. If you are using it try going without it and see if that improves anything. I'd try another daw to rule out any bitwig specific behaviour. I don't know about in bitwig but in Reaper unarming tracks that aren't being live tweaked significantly reduces the CPU usage. Good luck!
I'd configure Jack, it should help. Also, I've found the the multicore mode on repro worsens my performance. If you are using it try going without it and see if that improves anything. I'd try another daw to rule out any bitwig specific behaviour. I don't know about in bitwig but in Reaper unarming tracks that aren't being live tweaked significantly reduces the CPU usage. Good luck!
- Banned
- 2288 posts since 24 Mar, 2015 from Toronto, Canada
what apple product did you buy that is this config? is this a hackintosh?* Intel Xeon W-3235 12-Core CPU @ 3.30GHz
* 48GB 2666MHz
* 2TB SSD
* AMD Radeon Pro 580X
Spotify Soundcloud Soundclick
Gear & Setup: Windows 10, Dual Xeon, 32GB RAM, Cubase 10.5/9.5, NI Komplete Audio 6, NI Maschine, NI Jam, NI Kontakt
Gear & Setup: Windows 10, Dual Xeon, 32GB RAM, Cubase 10.5/9.5, NI Komplete Audio 6, NI Maschine, NI Jam, NI Kontakt
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
Thanks for your help!Robinrobo wrote: ↑Sun Dec 27, 2020 9:13 pm A few ideas for troubleshooting. I haven't directly compared Linux and Mac so I have no thoughts specific to that scenario. That being said I find things runs really well on my ancient Sandy bridge system with 8gb of ddr3 and reaper.
I'd configure Jack, it should help. Also, I've found the the multicore mode on repro worsens my performance. If you are using it try going without it and see if that improves anything. I'd try another daw to rule out any bitwig specific behaviour. I don't know about in bitwig but in Reaper unarming tracks that aren't being live tweaked significantly reduces the CPU usage. Good luck!
From what I understand ALSA sits underneath Jack and therefore can't perform worse than ALSA + Jack.
Anyways, at the moment I don't need the flexibility that Jack offers. But I have already uninstalled Jack and installed Jack 2 and will certainly dive into it very soon.
Last edited by amadeusp on Wed Dec 30, 2020 5:08 pm, edited 1 time in total.
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
Awesome, thanks!
Code: Select all
cpupower frequency
Code: Select all
powersave
Code: Select all
performance
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
- Banned
- 2288 posts since 24 Mar, 2015 from Toronto, Canada
NM... look like other poster fixed it for you.amadeusp wrote: ↑Wed Dec 30, 2020 5:15 pmIt's a Mac Pro:
https://browser.geekbench.com/v5/cpu/5611027
Spotify Soundcloud Soundclick
Gear & Setup: Windows 10, Dual Xeon, 32GB RAM, Cubase 10.5/9.5, NI Komplete Audio 6, NI Maschine, NI Jam, NI Kontakt
Gear & Setup: Windows 10, Dual Xeon, 32GB RAM, Cubase 10.5/9.5, NI Komplete Audio 6, NI Maschine, NI Jam, NI Kontakt
- KVRian
- 1051 posts since 6 Jun, 2016 from San Marcos, Texas
The Xeon stuff is interesting. I built a Xeon system in 2016 and it's the best desktop I've ever built for myself.
In a layman way of putting it, Xeon/server stuff isn't about pursuing the top-end. Rather, it's more about a guaranteed performance floor. If that makes sense.
In a layman way of putting it, Xeon/server stuff isn't about pursuing the top-end. Rather, it's more about a guaranteed performance floor. If that makes sense.
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
I was working with a Hackintosh for a couple of years and although it was an amazing time (the machine and the amount of stuff I learnt about computers in general really was absolutely great for me) I wanted to have a system that was more robust so that I would actually spend more time making music and less time learning and fiddling with Hackintosh stuff.lunardigs wrote: ↑Wed Dec 30, 2020 10:56 pm The Xeon stuff is interesting. I built a Xeon system in 2016 and it's the best desktop I've ever built for myself.
In a layman way of putting it, Xeon/server stuff isn't about pursuing the top-end. Rather, it's more about a guaranteed performance floor. If that makes sense.
The Xeon based Mac Pro is no let down in that regard. I sold all my Eurorack stuff to be able to afford it and I do not regret it. The machine is super reliable, pretty capable and extremely flexible due to the large mainboard with tons of room to expand and experiment. Of course, there's also downsides to it, the T2 chip for example. And yes, I could have built a threadripper for a fraction of the cost. But for my current situation with the ongoing transition to Linux and the possibility to have a fully working macOS at hand at any time is pretty much ideal for me.
-
- KVRist
- Topic Starter
- 37 posts since 13 Oct, 2011 from Basel
Yes, thanks. It is working perfectly now.telecode wrote: ↑Wed Dec 30, 2020 7:55 pmNM... look like other poster fixed it for you.amadeusp wrote: ↑Wed Dec 30, 2020 5:15 pmIt's a Mac Pro:
https://browser.geekbench.com/v5/cpu/5611027