CPU usage in Live 10

Discussion of music production, audio, equipment and any related topics, either with or without Ableton Live
Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Thu Mar 21, 2019 8:45 pm

pottering wrote:
Thu Mar 21, 2019 6:05 pm
"Anticipative processing" is just an extra audio buffer, that adds to latency like any buffer and doesn't fit the real-time model of Ableton Live.
...........
Rising Live's audio buffer to a huge size similar to the size of Reaper's anticipative processing's buffer size should get the performance closer.

Other causes of supposed higher CPU usage in Live are processes running that people don't realize are running real-time using CPU, like warped audio and muted plugins.

Point is, people keep talking about "inefficiency", and "good/bad programming", but they are even not comparing the same things in Live and Reaper.
Thing is none of that matters, not a bit of it, if I can do exactly the same thing in Reaper without the CPU hit.
All the various methods DAWs use to control CPU outside of super high over all latency involve raising the buffer on tracks that are not being used. Even Ableton at one point had a low latency setting for armed tracks. Apparently they gave up on it though.

I'm not trying to throw Ableton under the bus here, but it's obvious that certain DAWs are tentative at best code wise when it comes to setting up unarmed tracks for a slightly higher latency as a CPU saver, and Live is one. Again though, I haven't seen much of a performance cut in operations involving real time use in DAWs like Reaper. DP for most things also doesn't suffer from it. DP has an Effect Performance window that allows you to monitor this so you can watch a track switch over and notice if there's a latency, so far I haven't seen it.

I'm not going to bank on it, but I would bet at some point Live gets a rewrite, because using programs like Reaper it's obvious that the days where you need a 35% + overhead reduction to make real time performance seamless are over, or close to over.
In general though it's not a major concern of mine, but if I was in Live only, VEP would come in really handy for sure. Some Cinesamples Kontakt patches barely run on my computer.

pottering
Posts: 638
Joined: Sat Dec 06, 2014 4:41 am

Re: CPU usage in Live 10

Post by pottering » Thu Mar 21, 2019 10:20 pm

nm
Last edited by pottering on Sat Mar 23, 2019 2:19 am, edited 1 time in total.

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Fri Mar 22, 2019 2:49 am

pottering wrote:
Thu Mar 21, 2019 10:20 pm
It matters because it explains where the differences in performance between Live and Reaper come from, which is not "bad coding" from Ableton's part (to start with, CPU usage is not the only factor to consider when weighting "good" or "bad coding"), and also provides a suggestion for people that only want to use Live (rise audio buffer to a size similar to Reaper's "anticipative processing" and turn on "Reduced Latency When Monitoring" option).
Again, I can't see any performance oriented advantages that Live has over Reaper in terms of real world testing.
Reaper doesn't glitch when I add a plug in while the sequencer is playing, I get no noticeable latency when arming and playing a track etc. etc.
The reason initially was probably real time processing for sure, but I can't say with modern computers that I see any rational reason as to why Live is so much less CPU efficient.

All of that is true with Logic and DP here, they both stutter when adding plug ins sometimes, but Reaper points out clearly that it's possible to be CPU efficient and still have a strong playback engine. I'm not trying to throw Live under the bus here, but the excuse that it's a real time performance DAW runs short when there are DAWs like Reaper out there that are supremely CPU efficient and still don't break under pressure.

Nokatus
Posts: 1064
Joined: Fri Jul 01, 2005 7:06 am

Re: CPU usage in Live 10

Post by Nokatus » Fri Mar 22, 2019 11:44 pm

pottering wrote:
Thu Mar 21, 2019 10:20 pm
provides a suggestion for people that only want to use Live (rise audio buffer to a size similar to Reaper's "anticipative processing" and turn on "Reduced Latency When Monitoring" option).
I know how it can feel like it, but this isn't at all the same thing ;). The whole point is being able to work in realtime, contextually, within a large project/composition. In Live, "Reduced latency when monitoring" cannot override the actual main audio buffer individually on a track to track basis, that's a given. So if you rise the audio buffer "to a huge size", as you put it, that's the minimum latency you are getting with anything you arm and play, in any case, and then you can't compose by really performing the parts anymore.

With Reaper's anticipative processing you can leverage the CPU power for calculating in advance everything you aren't specifically controlling in realtime, independent of the audio buffer. I'm quite sure Live does something similar behind the scenes (why wouldn't it?), but stays on the safe side and within the restraints of prioritized live use and everything -- but that's where the a) quality of implementation and b) restrictions of said priorities come into play. So, in Reaper, for example if I'm routinely working with just a couple of ms realtime latency when composing, I can still have anticipative processing turned on the whole time, run an absolute ton of Kontakt instances and processing for different buses/sections etc etc... and yet, when I arm something to record a realtime passage on any given instrument on the fly, the timing is very tight, as it bypasses the anticipative processing for that instance/chain and thus has the straight responsive feel of that very short audio buffer. Boom. It really is magnificent :)

pottering
Posts: 638
Joined: Sat Dec 06, 2014 4:41 am

Re: CPU usage in Live 10

Post by pottering » Sat Mar 23, 2019 2:17 am

OK, sorry.

I won't bother you both anymore.

Zero interest in discussing Reaper, my mistake chiming in this thread.

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Sat Mar 23, 2019 4:06 am

I don't really get why anyone erases their posts when they've been quoted?
It's just awkward. :x

The point wasn't to chat about Reaper, DP, Logic etc. all do it as well. It's just to point out that Ableton still have a CPU heavy DAW on their hands at a time when other DAWs seem to have gotten around the issues that one has always attributed it's heavy CPU use to, i.e. it's real time performance oriented legacy.

Beyond that Live is lovely, I especially appreciate it on the laptop where I'm going to be working for the next couple weeks. :)

Nokatus
Posts: 1064
Joined: Fri Jul 01, 2005 7:06 am

Re: CPU usage in Live 10

Post by Nokatus » Sat Mar 23, 2019 8:47 am

Yep, I love Live also, and have used it for fifteen years :D ... And again, just now, the latest project I did was Live all the way. Now that I have kept those massive "traditional" composing projects elsewhere, I've been happier and feel like I'm using these tools according to their strengths, without feeling like I'm forcing it, in any environment I work in. Everyone's mileage may vary, of course.

The reason I like to mention Live's CPU use, at least on occasion, is how Live 10 is evidently more CPU heavy in use than a previous version, for no apparent reason -- that is, in exactly the same use case as before, Live 10 does tax the system more. Given that Live was already in the heftier end of the scale before this, I would be more comfortable seeing these current developments bringing that load down, not hiking it yet higher.

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Sat Mar 23, 2019 9:39 pm

Nokatus wrote:
Sat Mar 23, 2019 8:47 am
Yep, I love Live also, and have used it for fifteen years :D ... And again, just now, the latest project I did was Live all the way. Now that I have kept those massive "traditional" composing projects elsewhere, I've been happier and feel like I'm using these tools according to their strengths, without feeling like I'm forcing it, in any environment I work in. Everyone's mileage may vary, of course.

The reason I like to mention Live's CPU use, at least on occasion, is how Live 10 is evidently more CPU heavy in use than a previous version, for no apparent reason -- that is, in exactly the same use case as before, Live 10 does tax the system more. Given that Live was already in the heftier end of the scale before this, I would be more comfortable seeing these current developments bringing that load down, not hiking it yet higher.
I'm the third man out here, love Live, also love old school DAWs, use DP10 and Reaper all the time as well. :)

I really haven't noticed any higher CPU use in Live 10 than 9. I could test it easy enough, I have to prep a couple songs for travel etc. later so I can run some tests. I kind of doubt any truly noticeable increase in CPU in 10, mostly over the years a DAW might take another 5% to run, and everyone loses their mind. :)

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Sun Mar 24, 2019 10:31 pm

Nokatus wrote:
Sat Mar 23, 2019 8:47 am
The reason I like to mention Live's CPU use, at least on occasion, is how Live 10 is evidently more CPU heavy in use than a previous version, for no apparent reason -- that is, in exactly the same use case as before, Live 10 does tax the system more. Given that Live was already in the heftier end of the scale before this, I would be more comfortable seeing these current developments bringing that load down, not hiking it yet higher.
So, running a strait forward test of Live 9 VS 10. Using Collision with the Electric Guitar and a fast two note 16th beat one measure run in Live 9 Suite, on a Macbook Pro 2012, 2.7 four core using AKAI's MPC Live as an audio card. I can run 24 instances before audio starts crackling.

I get exactly the same number in Live 10 Suite.

The reason to use Collision over third party is simple. Ableton could beef up their VST or AU spec for modern standards, and that could subsequently cause CPU issues among other things with third party plug ins that wouldn't necessarily be Ableton's fault. This way it's all in house for the most part. My findings plainly point to Live 10 not being more of a CPU pig than 9, that doesn't mean that some third party plug in won't run worse in 10, but it's far more likely that a plug in that hasn't been updated in a long time is at fault there over Ableton making 10 run worse etc.

And as I've stated a gajillion times on this forum, the current Performance Test thread and the file etc. that people are using to test their set ups is not an accurate way to measure your computers performance, period. Only to failure tests actually let you know what a computer can handle, especially in this age of 4-12 core machines.

Also, kudos to Ableton, the built in plug ins seem much much more optimized than they used to be, 24 instances on a 7 year old laptop! 48 simultaneous notes. 8O (the note count would be higher with more notes per plug in and less instances of the plug in BTW)

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Sun Mar 24, 2019 10:44 pm

For fun, taking the Collision preset up to 11 notes at a 16th note run, 11 instances or 121 notes simultaneously. Could probably get to about 130 with 12 note polyphony. Recently had a debate about this, but more instances does not load balance if you're using the same plug in and sound. OS's and DAWs do not use the same RAM or whatever for the same plug in etc.

Nokatus
Posts: 1064
Joined: Fri Jul 01, 2005 7:06 am

Re: CPU usage in Live 10

Post by Nokatus » Mon Mar 25, 2019 12:58 am

Thanks for taking the time, it's reassuring to know that it's that much on par on some systems :). And yep, to failure is the only valid way to determine how the totality of any given system performs, that's very important, and agreed.

I recreated your test using 44.1 kHz sample rate and a 128 samples long audio buffer on a 2012 desktop system, in turn. A caveat: it didn't have Live 9 installed, so I went with the Collision in Live 8. Might not be comparable?

Fast two-notes-at-a-time pattern, 16th notes, in Live 8 small crackles at 86 instances. Loading the same project in Live 10, I need to go down to about 76-77 instances to be at the same level of slight crackles appearing. At 86 instances the project is still workable and there are no large dropouts, but it crackles markedly more, as in, enough to be disturbing. Either there's a difference between the algorithms of these Collision versions under the hood, or it's something else. There's also the possibility that the rest of the system might bottleneck differently in different setups and test cases, of course; something might start to give before Live does. This RME system generally doesn't bottleneck early, and stays clean all the way to effective 100% load. But that goes more into speculation territory anyway.

Good to know that it isn't that bad. In any case, many of those real-world projects that I had pushed right to the brink on an earlier Live version (and of course include 3rd party plugins) do need some set management in Live 10 in order not to exhibit crackles.

cmkadmin
Posts: 9
Joined: Fri Mar 15, 2019 10:55 pm

Re: CPU usage in Live 10

Post by cmkadmin » Mon Mar 25, 2019 1:00 am

I have a strange CPU usage issue as well. It appears in some of the Ableton packs, where one rack instrument in the same category will blow the CPU out of the water at a fairly low level of polyphony.

For example: Electric Keyboards -> Basic Tonewheel Organ, sits at around 5%, while the Organ Mustard Ballpark sits at around 30-40% with just normal playing. If I hold sustain down (yes, I realize you probably would never do that with an organ :D ), the Basic Tonewheel Organ never gets above 30% - and that's probably with all 61 keys sounding, but the Organ Mustard Ballpark blows up almost immediately - and in fact is not always playable by itself. I've seen this with a number of other sampled instruments as well (Mallets as an example)

I have no issues with Kontakt, other NI instruments or other VST's (Halion, Arturia) or in Cubase or FL Studio. Running 44Khz with Focusrite Scarlett 18i8.

If anyone wants to give it a try to reproduce the Organ, I figured I would throw that out there.... Little upset because I just bought Live 10 Suite.

Nokatus
Posts: 1064
Joined: Fri Jul 01, 2005 7:06 am

Re: CPU usage in Live 10

Post by Nokatus » Mon Mar 25, 2019 1:20 am

Continued a bit from above: as predicted, but still interestingly, the percentage-wise difference between versions becomes less when using a longer buffer. Bumping the buffer from 128 to 512 samples, Live 8 runs 143-144 instances of Collision in that same project before getting the crackles, and Live 10 about 136 instances. That's a pretty good percentage for such a margin. So... yeah.

Edit: thanks for chiming in, cmkadmin; in cases like that, I'd check what specific stuff resides in the rack :) and how it might effect the load. Also note the "test to failure [audio breaking up], not just numbers", heh. I can't check those rack presets myself, as I don't have those packs installed at all.

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Mon Mar 25, 2019 4:16 am

Nokatus wrote:
Mon Mar 25, 2019 12:58 am
Fast two-notes-at-a-time pattern, 16th notes, in Live 8 small crackles at 86 instances. Loading the same project in Live 10, I need to go down to about 76-77 instances to be at the same level of slight crackles appearing. At 86 instances the project is still workable and there are no large dropouts, but it crackles markedly more, as in, enough to be disturbing.
OK so please take this as healthy criticism, to failure is the coin of phrase, but what you want to do is take the CPU to failure and reduce plug ins until you get no crackles or other audio artifacts in your signal at all! The reason for this is obvious, this is the ability within a 5% range of your audio system to handle a project without it being compromised to the point of breaking down. I can get more or less the same level of crackles for a large amount of plug ins, then it abruptly stops the engine, but none of us work like that.

You might be surprised like I was, when Live 10 and 9 resulted in the same CPU level using the same plug ins. :)

Machinesworking
Posts: 11098
Joined: Wed Jun 23, 2004 9:30 pm
Location: Seattle

Re: CPU usage in Live 10

Post by Machinesworking » Mon Mar 25, 2019 4:45 am

cmkadmin wrote:
Mon Mar 25, 2019 1:00 am
I have a strange CPU usage issue as well. It appears in some of the Ableton packs, where one rack instrument in the same category will blow the CPU out of the water at a fairly low level of polyphony.

For example: Electric Keyboards -> Basic Tonewheel Organ, sits at around 5%, while the Organ Mustard Ballpark sits at around 30-40% with just normal playing. If I hold sustain down (yes, I realize you probably would never do that with an organ :D ), the Basic Tonewheel Organ never gets above 30% - and that's probably with all 61 keys sounding, but the Organ Mustard Ballpark blows up almost immediately - and in fact is not always playable by itself. I've seen this with a number of other sampled instruments as well (Mallets as an example)

I have no issues with Kontakt, other NI instruments or other VST's (Halion, Arturia) or in Cubase or FL Studio. Running 44Khz with Focusrite Scarlett 18i8.

If anyone wants to give it a try to reproduce the Organ, I figured I would throw that out there.... Little upset because I just bought Live 10 Suite.
There are a couple reasons for this, certain patches of the same plug in regardless of the DAW used, use different amounts of CPU.
Live's CPU meter is estimating a single core on your multi core cpu system, and it's level before failure. If you run any test of CPU on a computer you will see this behavior play out as some odd jumps in "CPU use". So one instance of a heavy CPU use plug in or patch will sit at 40% and two will only jump to 45%, but put four or more instances on a four core laptop for instance and you will see it jump to 85%. Up until 4 Live loads every new instance on a new cpu core, so the meter only goes up marginally because every core is only within 45% of failure, but a 4th or 5th plug in will double that on at least one CPU, so it reads that your at 85% etc. It can happen at 4 because Live itself also takes some CPU.

Basically, it's not a big deal if one plug in hit's 45% cpu on your system, but for that track you might want to use sends to add reverb etc. just to lighten the load. Warning though, it's very possible to overload one core and get crackles etc. by loading a lot of plug ins on one track. Most of us have probably run into this loading mastering plug ins on the master fader etc.

Oh and one last thing, a Rack is not just a sampler, there are more than 5 FX built into each of those organ racks, like a preset, how they're used can add more CPU for sure.

Post Reply