Performance Problems live VSTi's at under 128 buffer vs Logic, fascinating discovery

Hi there…

Some performance tests with heavy halion sonic patches, all done at 44K. What I am doing is picking a patch where playing chord progressions uses an entire core for one instance. All tests done on the crystal well macbook pro in signature, turbo disabled, clock at steady 2.8ghz… This is to avoid high temps and throttling. I am using the inbuilt headphone out, using a cable plugged straight into an IK iloud.

On playback using Asio guard, Cubase, like Pro tools, is superb at spreading load of pre recorded midi or audio tracks in playback mode with heaps of native FX…

But Logic is killing all of them at low buffer to “monitor” external audio through the DAW with native fx, or playing VI’s in realtime.

I am using a very hungry halion sonic patch for this test called “chord builder 124”.

  1. In Pro tools 2018.3, I can play it at 128 buffer with three note chord progressions…At 64 buffer there are dropouts, at 32 buffer not even one key can be pressed.
    If i duplicate the track and record arm it as well, so all halion sonic tracks are being triggered “live” at the same time i press a key, i get some dropouts at 5 instances. So, 4 is playable… that’s the max when the instrument tracks are all “armed”. Don’t even think about 32 or even 64 buffer in this scenario.


2)…Cubase Pro

I can play EIGHT of them “live” at 128 buffer, yes 8… literally 8 halion sonic instrument tracks, all record armed, playing 3 note chords of the same heavy patch simultaneously. And Pro tools can only do 4 safely (this has blown my mind as i thought pro tools at 128 buffer always equalled cubase capability… this is a new finding for me, unless there is some problem with the aax of halion sonic 3).

at 32 or 64 buffer, i cannot even play one in Cubase without way more dropouts than even Pro tools. Just non stop dropout city.

To play 8 in pro tools and match cubase at 128 samples, i have to be at 256 buffer in pro tools.

At 128 samples using the macbook pro’s in built audio driver, Cubase reports a 8.231 ms output buffer
(I have no idea what it is in PT as PT doesn’t tell you anything like that).
At 64 samples, Cubase reports 5.329ms output latency
At 32 samples, Cubase reports 3.878ms

So we go to Logic 10.4.1:

  1. At 32 buffer, Logic reports 3.2ms output
    At 64 buffer, Logic reports 3.9ms
    At 128 buffer, Logic reports 5.3ms
    At 256 Logic reports 8.2ms, so basically the same as what Cubase reports at 128 buffer! Is Cubase adding some sort of extra buffer time cause of the asio bridge to core audio?

Now look at the Logic results…

At 32 buffer, using the same patch, “chord builder 124” and playing 3 note chord progressions, Logic has a very small amount of pops and crackles with one instance… like the other daws, not good, but way better than both of them as they are literally unplayable. At least logic is playable and bearable.

At 64 buffer, Logic can play 4 instances simultaneously.… Both other DAWs could not do one!

It doesn’t matter what you up the buffer to… at 128 buffer or 256 buffer, logic can still only play 4 instances simultaneously without pops and clicks, as Logic only uses 4 threads for live VI tracks at any one time (hey it’s much better than before when it used to stack them all on one core!). And yes, I am using the “multithreading for playback AND live tracks” setting.

So at 128 buffer, Cubase is literally getting DOUBLE the instances of Logic and PT… but… Logic says 5.3ms latency and Cubase is over 8ms (is this a reporting error on Cubase’s part or is it really that much higher output latency? if so, this would explain the performance difference to a degree!)

Of course once they are record disabled and logic shuffles them onto the higher internal playback buffer, it uses all 8 threads with ease and gets a similar amount to Cubase, perhaps even just a tad less…

But the point is, at 64 buffer, logic is 4x as powerful on what one can do in realtime. when it comes to VI.

I have received the same results when testing 32 audio tracks armed for record and putting FX at 32 or 64 buffer… logic is getting even sometimes 10x the effects of cubase or pro tools in that scenario.(logic uses ALL 8 threads for live armed audio tracks, unlike live armed VI’s). There is no DAW that can track through native effects at 32 buffer like Logic can. It’s crazy good at that. I mean, with say 4 presonus quantums, 96 inputs all being monitored through fx at 32 buffer in logic, is actually a realistic scenario on any modern desktop imac and up.

Since i would never have the need to play more than 4 Vi’s live, Logic wins this for me as i like to play my VI’s at a 64 buffer.(also, since i have a virus Ti through usb and that double the buffer, this means the virus is at 128… where as in Pro tools or cubase since i have to use 128 buffer in general, my virus is always at 256… too high).

But the way cubase is majestically spreading the load around the cores at 128 buffer for live armed tracks is amazing… it is thrashing everything in that regard.

Could Steinberg do something to improve the performance at 64 buffer? I certainly hope so… And why is the output latency of the same audio hardware so much higher in Cubase?

I don’t know the answer to everything but I found this all very interesting…

Ultimately, the issue is, that every DAW seems to have issues on both my macs at under 128 buffer for doing large amounts of “live” stuff, except for Logic. I know Avid will never fix it, but I am hoping that Steinberg at least will improve this.

Thanks for your time and reading :slight_smile: