How to measure CPU-usage of plugins?

[Arno Peeters]

It can very well be that the answer to this is buried in this forum somewhere, but I have yet to find it.
Of course you can watch the CPU Usage when you remove or add any plugin and I'm well aware that (say) a denoiser uses more CPU than an EQ, but is there a reliable and easy way to find out how, say Clarity VX measures up against VoiceDenoise or Convology XT against AltiVerb? Has anyone ever tried recently?

Curious if I can spare me some time here.

[Darryl Ramm]

I doubt there is anything useful out there and anything would be next to useless for real users to interpret/take advantage of. I do expect plugin developers to be benchmarking their plugins. And the is DAWBench (or rather two versions of DAWBench DSP (DSP==regular audio) and VI (virtual instruments)) that runs some plugin workload but it's Reaper centric, and really used for benchmarking computers and interface latency.

And some of the plugins that are CPU pigs can vary depending on what you are doing/how they are configured so may have many things to keep track of.

What really matters is if something on your system used how you are going to use it triggers a CPU error. And that's different than just how much a meter moves that claims to be showing CPU usage. And we sure know the Pro Tools CPU meters are not always accurate about what is going on (I have no technical idea what they are measuring, somebody can explain that to me would be great, and I understand how sampling based CPU usage/profiling works ). It's not just me, plugin developers have warned about using Pro Tools CPU meters on here before.

And how you use it is critical. Is the plugin part of a long chain of plugins that tend to be stuck on a CPU core. Is it that core near exhausted? Or the plugins on that core smashing the cache and seeing degraded performance? Is that plugin or others got erratic issues directly triggering CPU errors, etc.

And even if you could qualify something. How would you compare that on different processor cores/cache sizes etc. Whats the HW Buffer size? Or is that plugin streaming data thought the large disk playback buffer? What sample rate? Again, lots of stuff to keep track of.

As you are already thinking I suspect the best thing you can do is for you to compare plugins in working sessions. Swapping them in and out where they would be used in production. Even then I'd not have a lot of confidence in the Pro Tools CPU meters, I'd want to look at it with macOS tooling as well. The other thing is can you just work around any CPU usage issues? I suspect many of us are able to push our systems way further than in the past thanks to using freeze and commit. I know I can, to the point that I just don't hit CPU usage problems (8 core Intel MBP) unless I have a very occasional totally buggy plugin somewhere and then it's a matter of throwing out that plugin. Yes I know that won't work for everybody.

[Eric Lambert]

FWIW I use the Mac's Activity Monitor (in the Utilities folder) since I've learned not to trust PT's CPU monitor.

[Arno Peeters]

CPU pigs

Haha, that's a new one for my vocabulary!

Yeah, I was afraid I was asking for this answer. I'll narrow it a bit further as to why I want to know. It is not so much that I get CPU overloads mid-session, (I do a mainly podcast and documentary) but when committing and bouncing say an hour show, it definitely has effect on how long that takes.
As an example I'll take WavesFactory's TrackSpacer: I use that to duck the music using speech. The first version of this plugin was very useful, but uses an enormous amount of CPU. The update they released was way less of a pig but not nearly as transparent. Yesterday I saw the release of VoxDucker, which I still have to test. If it works better using less CPU, all for it. But it would be nice not to have to repeat the process, just to see how much faster it is.

[sdelsolray]

Studio One Pro's "Performance Monitor" is a light year ahead of Pro Tool's Activity Monitor.

[Eric Lambert]

Originally Posted by sdelsolray

Studio One Pro's "Performance Monitor" is a light year ahead of Pro Tool's Activity Monitor.

Wow, it sure is!

[Stephen Bond]

Not sure what I am missing but just to be clear, latency is not the true indicator of CPU cycles used. Sure, it might be but is can also be totally bogus.

I mean, how hard on the CPU is a latency correction plugin that is reporting 10000 samples in the Pro Tools mixer??

https://www.eventideaudio.com/plug-i...on-time-align/

I put money on this plugin costs hardly any CPU cycles compared with a basic reverb.

Stephen

[smurfyou]

Originally Posted by sdelsolray

Studio One Pro's "Performance Monitor" is a light year ahead of Pro Tool's Activity Monitor.

Wow how useful.

The term "plugin nap" is funny. Wonder if it was a literal translation from German. Plugin schlafchen? Excuse my ignorance.

[Darryl Ramm]

Plug-in nap is the equivalent of Pro Tools dynamic processing. Which I and I suspect others have found to have problems at times… but generally gotten better over time. How they offer that up here as per-plugin raises the question of do some plugins have problems with nap and can a user effectively find them? Vs the Pro Tools approach of global dynamic processing on or off. I have no experience with StudioOne. Anybody used this fine grain nap control?

[sdelsolray]

Originally Posted by smurfyou

Wow how useful.

The term "plugin nap" is funny. Wonder if it was a literal translation from German. Plugin schlafchen? Excuse my ignorance.

Well, the Studio One software team is mostly in Germany.