Topic by mqbernardo

  • Avatar
  • mqbernardo
  • Member
  • Offline
  • Registered: 01-08-2021
  • Posts: 26

Topic: PTQ 7 on new Mac M1

Just got myself a new MBP M1 Pro 14" base model. Unfortunately my DAW (Ableton Live) is still in beta for native ARM and isn't the most reliable regarding CPU meter in my experience.


Using only the core drivers and playing a C-E-G-C chord starting from middle C, in the standalone version, I get from 7% (Grotrian) to 11% (Ant. Petrof) CPU hit. Steinways, Steingraeber and Bechstein are 10% and Bluthner about 8%, using the Prelude preset with a buffer of 256 samples, 44.1 KHz sample rate  (2.9 seconds latency).

using 128 samples at 44.1 puts the CPU hit from 5% (again Grotrian) to 7% (Petrofs, Steingraeber and Steinways).

going for 512 samples (11.6 seconds latency) again at 44.1 takes it from roughly 18% (Grotrian and Bluethner) to 25% (the others).

Going for 64 samples or higher than 512 didn't seem to change these limits. Using the plugin version inside of Ableton Live produced roughly the same values, using any of the different formats.

Pianoteq was the first plugin I installed and I am still going through the 1st stages of adaptation. Besides that I am not much into the technical side of computers, but coming from a 2015 MBP model I was expecting more of wow factor. maybe I´ve been hyped by the buzz in the webz.

Wonder if anyone else has new ARM Macs with PTQ and wants to share experiences/expectations  with it?

cheers,
Miguel.

Reply by joelw

  • Avatar
  • joelw
  • Member
  • Offline
  • Registered: 17-11-2020
  • Posts: 4

Re: PTQ 7 on new Mac M1

I have been using Pianoteq on M1 (and M1 Pro) for around a year, and I am quite happy with their performance. Pianoteq runs so well and I don't see any practical blocker while performing.

I have no very good idea about how the CPU hit was measured in Ableton Live; does it denote single-thread CPU usage (which seems likely), or multi-core CPU usage where 100% means all the 8-10 CPUs are maxed out? Assuming it is the former, the numbers you show here doesn't look wrong or surprising to me because you only played few notes. The CPU load is not necessarily linear to the number of notes you're playing because of some constant overhead.  Performance can be better tell when the polyphony and audio load is high.

What do you see in the "Perf" tab in Pianoteq's preference panel, which shows you the audio load and performance index? You should be able to see a noticeable difference as you change the buffer size. Also, why don't you run Pianoteq as standalone and measure the process CPU usage which sounds to me like a more "reliable" measure?

Also note that Pianoteq hasn't been optimized with Silicon chip, and there are some unresolved issues regarding the performance in the presence of "little core" (see https://forum.modartt.com/viewtopic.php?id=8125). I also feel Pianoteq's multi-core support is not fully optimized in a sense that it often produces crackles and audio load gets maxed out while many other cores are still idle (like only 200% CPU usage in total), which would suggest Pianoteq does not utilize all the available cores for rendering. M1's single-core performance is of course better than that of 2015 MBP Intel, but not dramatically; the maximum performance gain of M1 comes from utilizing multi-core extensively.

Last edited by joelw (25-11-2021 01:25)

Reply by Groove On

  • Avatar
  • Groove On
  • Member
  • Offline
  • Registered: 14-09-2017
  • Posts: 199

Re: PTQ 7 on new Mac M1

joelw wrote:

II also feel Pianoteq's multi-core support is not fully optimized in a sense that it often produces crackles and audio load gets maxed out while many other cores are still idle (like only 200% CPU usage in total), which would suggest Pianoteq does not utilize all the available cores for rendering. M1's single-core performance is of course better than that of 2015 MBP Intel, but not dramatically; the maximum performance gain of M1 comes from utilizing multi-core extensively.

This may have to do more with the MacOS CPU scheduler rather than Pianoteq. On my Odroid N2+ (ARM / 6 cores) - Pianoteq load balances very nicely across all 6 cores. I've tried some Linux distributions that have less optimal CPU schedulers and in those cases, weird load balancing occurs where Pianoteq threads keep getting assigned to the same 1-2 cores, causing performance problems.

FWIW - on my 2015 MacBook Air, turning off all notifications in MacOS resulted in a more stable and straighter performance index (blue line in the graph). The MacOS notification scheduler can steal a lot of CPU cycles, it shows up as a more jagged blue line in Pianoteq.

Last edited by Groove On (25-11-2021 06:27)

Reply by mqbernardo

  • Avatar
  • mqbernardo
  • Member
  • Offline
  • Registered: 01-08-2021
  • Posts: 26

Re: PTQ 7 on new Mac M1

joelw wrote:

I have been using Pianoteq on M1 (and M1 Pro) for around a year, and I am quite happy with their performance. Pianoteq runs so well and I don't see any practical blocker while performing.

I have no very good idea about how the CPU hit was measured in Ableton Live; does it denote single-thread CPU usage (which seems likely), or multi-core CPU usage where 100% means all the 8-10 CPUs are maxed out? Assuming it is the former, the numbers you show here doesn't look wrong or surprising to me because you only played few notes. The CPU load is not necessarily linear to the number of notes you're playing because of some constant overhead.  Performance can be better tell when the polyphony and audio load is high.

What do you see in the "Perf" tab in Pianoteq's preference panel, which shows you the audio load and performance index? You should be able to see a noticeable difference as you change the buffer size. Also, why don't you run Pianoteq as standalone and measure the process CPU usage which sounds to me like a more "reliable" measure?

Also note that Pianoteq hasn't been optimized with Silicon chip, and there are some unresolved issues regarding the performance in the presence of "little core" (see https://forum.modartt.com/viewtopic.php?id=8125). I also feel Pianoteq's multi-core support is not fully optimized in a sense that it often produces crackles and audio load gets maxed out while many other cores are still idle (like only 200% CPU usage in total), which would suggest Pianoteq does not utilize all the available cores for rendering. M1's single-core performance is of course better than that of 2015 MBP Intel, but not dramatically; the maximum performance gain of M1 comes from utilizing multi-core extensively.

Thanks for the reply. As I hinted and you probably have guessed, computers are mostly black boxes to me.
With my previous Mac, PTQ 7 performance was more than adequate for my beginner piano needs, but using it inside a mix with more instruments and effects could be a problem, depending on the number and CPU-intensiveness  of the processes going on. Not a problem per se, as I could freeze tracks or record before the applying heavy effects and all that but still a drag.

I did use both the standalone and the plugin version for this assessment  "test". I don't think Ableton Live is very good at the multi-core stuff, and I realise that M1 gain comes mostly from multi core but still I was expecting something  more dramatic. As for the performance tab I get (with 128 samples, playing around 6-12 polyphony) about 100-109 performance index. Previous Mac, about 130. The blue line, whatever that means) stays close to the top of the rectangular box.

cheers,
Miguel.