2020 DAW shootout

[machinesworking]

Makes sense, interested to hear those results too, thank you once again for taking the time to do this!

Mostly I'm stuck trying to figure out the direction I want to go in DAW wise, (interested in using MPE, my main two DAWs Live and DP don't do it natively), so it's good for me to know, and I would hope people would try it themselves and see what types of performance they can expect.

Bitwig remains an anomaly. Reaper is for the most part solid, but I really would like to use a DAW that gets articulation or Expression maps like Cubase and Logic eventually, and I get the impression it's always going to be half assed in Reaper. Logic is kinda disappointing CPU wise right now, which is sort of a bummer, since Alchemy, MPE and Articulation maps are all there. All this is subject to my older CPUs of course, if I end up getting a modern i7 laptop it could be different if the DAWs and plug ins code for them.

[.jon]

This is just dead wrong.
There are basically two main tiers with DAWs.
Live, Bitwig etc. are in their own category, that use vastly more CPU. This is probably due to the way they treat tracks and interaction with tracks. In Bitwig and Live pretty much all tracks are treated as active, there's little extra buffering of unarmed tracks. Everything is treated as 'live'.

No, it is how things work. You don't seem to grasp how the magic box with the apple logo actually makes the sound- it's not the host. A DAW uses ridiculously little CPU for it's tasks of routing audio and passing data to plugins. The DSP happens in the plugin. The plugins use the CPU. Their algorithms, the DAW has no effect on these algorithms whatsoever, none, zero. It's the exact same amount of processing power needed in every single host, because it's the plugin code doing all the work. Literally and precisely the exact same computation in every computer running the same plugin at the same sample rate.

The second category is the more traditional DAW approach, DP, Logic, Reaper, Cubase etc.
These DAWs almost to a one use some sort of buffer, or pre rendering of tracks that are not currently armed for recording. This significantly increases track counts, and for years Logic was the king of the hill. Reaper currently is, and DP has caught up. Previous to version 9 of DP it was around 80% of what Logic or Reaper could do. Same with Cubase, for years Cubase had significant CPU hits at any buffer size besides 1076 on OS X, making it difficult to work in, a couple years ago they improved it and track counts increased.

Very good, this part you seem to understand- as the DAW can't affect the plugin code in any way, the only way to make processing easier is to increase buffer size. You have a limited number of CPU available per each sample block, so the only way to achieve more is to increase the buffer. That doesn't mean "Bitwig and Live use more CPU", it literally means they don't add extra latency on top of your audio interface block size.

You're just really misinformed here, this is all well known, it's not any mystery. What is a mystery is how you're getting such terrible results with Reaper? I'm not really a Reaper user, mostly DP and Bitwig, and Live since 2004, but I'm not unaware of its CPU consumption.

It's well known that "Reaper low CPU" is nothing more than a misconception spread years ago by clueless people who got confused by Reaper's CPU meter defaulting to hide RT CPU, and who don't understand what makes the sound in their boxes.

No that's not it. These tests are the same on a single quad core i7 here, Live VS Reaper, Reaper is always almost twice as many plug ins VS Live. With Live mostly clocking in around 60-70% of other DAWs.

Anticipative FX Processing, Under Preferences, Audio, Buffering> The default setting is on, and 200ms. If this is off it explains your poor numbers with Reaper. The other issue might be armed for recording tracks, Reaper does badly with armed tracks as Anticipative FX Processing doesn't work on those.

I've had a Reaper license forever and know the piece of shit well enough to script and theme it, but thanks for the tips. The numbers are with anticipative on, without it Reaper gets nowhere near any properly coded DAW. It's just what it is.

[machinesworking]

the DAW has no effect on these algorithms whatsoever, none, zero. It's the exact same amount of processing power needed in every single host, because it's the plugin code doing all the work. Literally and precisely the exact same computation in every computer running the same plugin at the same sample rate.

Sure sounds great, then why is it possible to have all DAWs at the same buffer settings and get wildly different results?

Why do some plug ins spike the CPU in some DAWs but not others?

It's obvious that almost all DAWs use a second buffer for instrument tracks that are currently not playable. In any of these tests you can see this by playing notes into armed tracks and watching the CPU go up. It's not a bad thing, they aren't "hiding RT CPU". If record enabled tracks respond at the buffer setting you set in preferences then I don't care what algorithm the DAW is using to get better performance.

The results I'm getting are consistent, there's no mystery in them, and they're very different for each DAW, there's no way it's just buffer size that's the variable here.

I see a bunch of variables, chipset instructions and how well the plug in and DAW handle those (or even if at all in the case of AVX), variations on how the DAW handles unarmed tracks by adding buffering, spiking caused by either the plug in or the DAWs handling of the plug in.

If it's just buffer size then explain why Bitwig consistently handles Diva at such a better rate than other DAWs? It's supposed to be a low buffer cheating DAW right?
I think you're putting all your eggs into one basket, there are other considerations when it comes to CPU use, a plug in and a host.

I'm still confused as to how you rigged the test to get Reaper to behave badly? Arming the tracks for recording would do it. Live and Bitwig don't respond much differently with armed tracks but the other DAWs do.

[bk]

Seems like the actual midi file will make a difference as well.
If I play it legato and notes overlap more, and you play it staccato, we'll get different numbers, for sure.

[bk]

Why not post the actual midi file, for consistency, or specify a certain number of voices per instance, played at the same time?

[Passing Bye]

Logic 10.5 is out, do you think anything changed because of new live oriented workflow CPU vise, any possibility for testing it out?