Bitwig's cpu core allocation per track?

[antic604]

My 24 Thread AMD CPU results:

More CPU Cores = More Tracks with more Plugins.
tested (20x Arturia Pigments per track equal smooth play)

One Track with Instrument Layer is One Audio Signal Way which limited to One CPU Thread.
tested (4x Arturia Pigments in Instrument Layer and hear Jitters)

That's strange. I've run 10 instances of Pigments separately on each track and then 10 instances inside of Instrument Layer on singe track. In both cases CPU utilisation was the same (27-29% on average), but seemed a bit more stable (less variance around the mean) for the latter scenario.

I'm on Win10 / i7-8750h laptop.

Out of curiosity - does Bitwig run the AVX2 audio engine on AMD processors?

[ThomasHelzle]

AFAIK Bitwig optimises whatever it can, even inside a single track.
Layers are one area, but also it's own devices compile themselves on the fly optimised for the current machine (you may see that pop up quickly when you switch between Grid presets for instance).
I don't know the specifics, but it's definitely more involved than just 1 track = 1 core.

Cheers,

Tom

[pdxindy]

AFAIK Bitwig optimises whatever it can, even inside a single track.
Layers are one area, but also it's own devices compile themselves on the fly optimised for the current machine (you may see that pop up quickly when you switch between Grid presets for instance).
I don't know the specifics, but it's definitely more involved than just 1 track = 1 core.

Cheers,

Tom

The Bitwig instruments are multi-core. I wonder if Instrument Layer is too?

[WasteLand]

Out of curiosity - does Bitwig run the AVX2 audio engine on AMD processors?

yes, because AMD has AVX2..... well not all of course, i don't know when they introduced it.

so Zen 1/2/3 have them, for instance (and Zen3 is the first gen, that competed, or even performed better than intel procs, single core, well the 12th gen, gets over Zen3, but Zen3+ is in the making and Zen4, and well M1 Max, etc. they will jump over eachother the next years, till again one of the 3, or 2 of the 3, will be left behind. as it always does... well... prediction is like branch prediction, but branch prediction is a better prediction, but not for things that aren't anticipated).

EDIT: the most common advice is;

the highest IPC per core + the highest amount of cores, within your particular budget.
single core performance (not the CPU Benchmark, that does indeed benchmark one 'thread', the term 'thread' is now very known, but o well, it still a strange term.. but CPU Benchmark, only utilize ONE core for its benchmark, so you can never know how it performs when all cores are utilizied.
overclocking Zen3 is not interesting, and even the intel 12th gen. they can already do things better, without manual overclock.
don't know how it is called for intel now, but Zen3 delivers; with PBO2, curve optimizer per core, negative values (-15, -20 for the best cores, guidelines, you must test this), ánd PPT/TDC/EDT set to manual; especially EDT set it to a lower value, and the use of EDT drops! in percentage... if you set it higher, via AMD Master or Motherboard settings, it goes very fast to 100%, EDT...

so also balancing you CPU, for optimal results.

Zen3 is already outdated, as always, when you buy a CPU.... but it's IPC per core ánd multicore IPC per core, is when balanced, with PBO2, curve optimizer and PPT/TDC/EDT... can perform quite impressive at an acceptable power draw, 180/190 watt, full load).

that is the story, common, independt of DAW's.

how DAW's react, and it depends on many factors.

Cubase is performing like crazy, for Cubase. Bitwig also...

always user cases; what plugins are important to you, i have many plugins that sweep a CPU away, my old system could still manage it, but i to limit myself. limitations there always be.

and by the Pigments 3; it depends on the patch how Pigments 3 performs; it performs quite badly with MPE; high grain count, a lot of modulation. or badly, perhaps under the hood, it does keep the sound as its precious, against loading a full core....)

[Lutein79]

Hello again back,
so I must to take back my previous post about the Jitter's in Instrument Layer device.
I now tested again 8 Pigment Instances in Instrument Layers on One track and no CPU overhead, no Jitter's.
I Think that the issue was any temporary issue, dependent on previous work in da project.

[WasteLand]

Hello again back,
so I must to take back my previous post about the Jitter's in Instrument Layer device.
I now tested again 8 Pigment Instances in Instrument Layers on One track and no CPU overhead, no Jitter's.
I Think that the issue was any temporary issue, dependent on previous work in da project.

so instrument layers, are not one core, i already mentioned in my previous post, Pigments 3 can perform quite good, with simple patches, but it can outmax with 3 or 4 instances, the CPU, with complex MPE patches. also, don't forget; also the driver/interface of the soundinterface, the implementation has indeed influence on the performance; more plugins/instances.

i am not talking about latency, but at the same sample rate and sample buffer, with the same system, i could only have 2 instances of Pigments 3, with a very very demanding MPE patch, with two different soundinterfaces, and the third one; gave 4... double, still not a lot..

anecdotal evidence. by the way. but for me a personal test for my system + soundinterfaces. so the third i still have, the other 2, that owned for some time, are gone now...

[Lutein79]

Hello again back,
so I must to take back my previous post about the Jitter's in Instrument Layer device.
I now tested again 8 Pigment Instances in Instrument Layers on One track and no CPU overhead, no Jitter's.
I Think that the issue was any temporary issue, dependent on previous work in da project.

so instrument layers, are not one core, i already mentioned in my previous post, Pigments 3 can perform quite good, with simple patches, but it can outmax with 3 or 4 instances, the CPU, with complex MPE patches. also, don't forget; also the driver/interface of the soundinterface, the implementation has indeed influence on the performance; more plugins/instances.

i am not talking about latency, but at the same sample rate and sample buffer, with the same system, i could only have 2 instances of Pigments 3, with a very very demanding MPE patch, with two different soundinterfaces, and the third one; gave 4... double, still not a lot..

anecdotal evidence. by the way. but for me a personal test for my system + soundinterfaces. so the third i still have, the other 2, that owned for some time, are gone now...

Yes, it seems that Instrument Layer work in multithreaded mode too. Now rally tested. Perfect. Bitwig is liberation for me. After suffering the last few years with Ableton, Bitwig is a gift from heaven.

[WasteLand]

Yes, it seems that Instrument Layer work in multithreaded mode too. Now rally tested. Perfect.

great!