Did you test it and got the same behavior?antic604 wrote: ↑Tue Jan 07, 2020 7:52 am Interesting! Perhaps Diva somehow is aware there are several instances being run and randomises each one ever so slightly to sound "analog" (by using different random seeds), but once you've separated them - by switching to individual - it doesn't know that anymore and uses the same random seed for each instance?
Latest News: Bitwig updates Bitwig Studio to v5.1
Completely baffling and amazing CPU behavior with Bitwig 3!
-
- KVRist
- 497 posts since 1 Jul, 2009
-
- Banned
- 11467 posts since 4 Jan, 2017 from Warsaw, Poland
No, sorry. I've not even touched Bitwig this year (!) yet... Perhaps over the weekendanoise wrote: ↑Fri Jan 10, 2020 7:47 amDid you test it and got the same behavior?antic604 wrote: ↑Tue Jan 07, 2020 7:52 am Interesting! Perhaps Diva somehow is aware there are several instances being run and randomises each one ever so slightly to sound "analog" (by using different random seeds), but once you've separated them - by switching to individual - it doesn't know that anymore and uses the same random seed for each instance?
-
- KVRian
- 841 posts since 6 Nov, 2004 from UK
Hi guys. Was going to start a new thread about this but may as well keep it here as it's related to CPU performance in Bitwig too.
So recently i got a new Imac2019 i9 8core 5ghz, 40gig mem, SSD. I thought it was going to be amazingly powerful - but when i started testing against my previous machine (i7 quad core) i got some VERY CRAZY results. Crazy as in inconsistent. The difference between DAWs and Plugins is just unreal.
Here are the results:
The only real consistent factor is that bitwig (v3) is bad compared with Reaper. Infact i nearly switched recently because of this. Some of the difference is not small at all.
The other very odd thing is that in a lot of test cases, Bitwig is performing better on my previous machine than the new one. What's annoying is that the clicks & pops are occurring when the activity-monitor is showing that hardly any cores are being stressed at all, over half remain unused and the rest are all hovering around 30% max.
Just hope that the Devs of bitwig can look into this and improve things.
Best
Dale
So recently i got a new Imac2019 i9 8core 5ghz, 40gig mem, SSD. I thought it was going to be amazingly powerful - but when i started testing against my previous machine (i7 quad core) i got some VERY CRAZY results. Crazy as in inconsistent. The difference between DAWs and Plugins is just unreal.
Here are the results:
The only real consistent factor is that bitwig (v3) is bad compared with Reaper. Infact i nearly switched recently because of this. Some of the difference is not small at all.
The other very odd thing is that in a lot of test cases, Bitwig is performing better on my previous machine than the new one.
What's annoying is that the clicks & pops are occurring when the activity-monitor is showing that hardly any cores are being stressed at all, over half remain unused and the rest are all hovering around 30% max. Just hope that the Devs of bitwig can look into this and improve things.
Best
Dale
You do not have the required permissions to view the files attached to this post.
-
- KVRAF
- 1996 posts since 16 Jan, 2013 from USA
What version of macOS were you running?
-
- KVRian
- 841 posts since 6 Nov, 2004 from UK
Catalina on the new Mac and Mojave on the i7 mac.What version of macOS were you running?
-
- KVRian
- 1262 posts since 15 May, 2002 from Finland
I wasn't nowhere near CPU maxing out, but those sharp spikes happened a lot, probably that was either some hardware driver, some plain problem from my five years old Windows 7 installation, or something related to multithreading. I'd like to test this under Linux. Also my desktop is a HP Elitedesk G2 SFF 65W super small chassis with a lot of Intel and HP hardware level security shit running since it's a business computer, so it's not probably the optimal i7 6700 setup for realtime audio work (but I like it since it's tiny).
Did we all use the same Bitwig plugin hosting settings "by plugin"?
Did we all use the same Bitwig plugin hosting settings "by plugin"?
-
- KVRian
- 985 posts since 10 Sep, 2014
Most of the driver problems come from the graphic cards or the wifi card. Be sure you're running the latest driver for your graphic card and turn off wifi when using Bitwig. You could also dualboot into a barebones Win installation if you have enough free hdd space.Taika-Kim wrote: ↑Fri Jan 10, 2020 8:27 pm I wasn't nowhere near CPU maxing out, but those sharp spikes happened a lot, probably that was either some hardware driver, some plain problem from my five years old Windows 7 installation, or something related to multithreading. I'd like to test this under Linux. Also my desktop is a HP Elitedesk G2 SFF 65W super small chassis with a lot of Intel and HP hardware level security shit running since it's a business computer, so it's not probably the optimal i7 6700 setup for realtime audio work (but I like it since it's tiny).
Did we all use the same Bitwig plugin hosting settings "by plugin"?
I'm hosting plugins individually.
-
- KVRian
- 985 posts since 10 Sep, 2014
-
- KVRian
- 1262 posts since 15 May, 2002 from Finland
I don't have a graphics card, the case can't fit one 
Internal Intel HD graphics. I did clean actually old drivers by hand with that one command line tool that I forget... I had to do a lot of work to get my installation to boot when I did a complete system upgrade from AMD to Intel. Oh yes WiFi was enabled, through a hub, which should make it even worse 
But I was in a bit of hurry and had to do some stuff simultaneously.
And, oh, results with 64-bit VST3.
Also, now I got completely confused when I wrote earlier about "by plugin", if I actually tested "individually". Well, I can do this again, I think I'd like to test from an Ubuntu Studio USB live boot, in my earlier experience I got as much as 30% better performance under Linux than Win 7 because of much more even CPU load. Even to Windows 10 difference was 10-15% in favour of Linux.
Internal Intel HD graphics. I did clean actually old drivers by hand with that one command line tool that I forget... I had to do a lot of work to get my installation to boot when I did a complete system upgrade from AMD to Intel. Oh yes WiFi was enabled, through a hub, which should make it even worse But I was in a bit of hurry and had to do some stuff simultaneously. And, oh, results with 64-bit VST3.
Also, now I got completely confused when I wrote earlier about "by plugin", if I actually tested "individually". Well, I can do this again, I think I'd like to test from an Ubuntu Studio USB live boot, in my earlier experience I got as much as 30% better performance under Linux than Win 7 because of much more even CPU load. Even to Windows 10 difference was 10-15% in favour of Linux.
-
- KVRian
- 1262 posts since 15 May, 2002 from Finland
My Thinkpad X220 laptop with internal sound, 2-core i5 2320m CPU, Windows 10:
8-note arp: 24 instances.
8-note chord: 4 instances.
Less spikes than Win 7, with 8 note chords the DSP load line was almost a completely flat line.
8-note arp: 24 instances.
8-note chord: 4 instances.
Less spikes than Win 7, with 8 note chords the DSP load line was almost a completely flat line.
-
- KVRian
- 1249 posts since 17 Oct, 2018
Agreed. We need a baseline to compare with.
Studio One // Bitwig // Logic Pro X // Ableton 11 // Reason 11 // FLStudio // MPC // Force // Maschine
-
- KVRian
- 1262 posts since 15 May, 2002 from Finland
Welllll this is interesting, I just installed Ubuntu Studio, and I get 120-130 single voice instances and 14 instances playing 8 notes, so the same i7 6700 performs 100% better under Linux 
It is a matter of usage of the cores/threads. You can see here the load jump almost 100% when I got from 8 instances to nine (my CPU has 4 cores with 8 threads). So that ninth has to start using one of those already taxed cores, since those eight voices can't be divided amongst the separate threads since they need to be summed inside the same plugin before outputting. And all the rest of those threads need to wait for that one to finish before the sound can be summed for final output. Diva's own multicore settings don't make a difference here. But those single-voice instances take very little CPU, so the load is easy to spread along the available thread capacity.
This was not a fair test though, since now I have a clean install vs several years old one. Still, pretty interesting results.
Oh, and those sharp spikes that many people are complaining about: I was expriencing them under Windows with this very same benchmark, but at least for now, there seems to be no similar issue under Linux.
https://ibb.co/RcpDJ9P
It is a matter of usage of the cores/threads. You can see here the load jump almost 100% when I got from 8 instances to nine (my CPU has 4 cores with 8 threads). So that ninth has to start using one of those already taxed cores, since those eight voices can't be divided amongst the separate threads since they need to be summed inside the same plugin before outputting. And all the rest of those threads need to wait for that one to finish before the sound can be summed for final output. Diva's own multicore settings don't make a difference here. But those single-voice instances take very little CPU, so the load is easy to spread along the available thread capacity.This was not a fair test though, since now I have a clean install vs several years old one. Still, pretty interesting results.
Oh, and those sharp spikes that many people are complaining about: I was expriencing them under Windows with this very same benchmark, but at least for now, there seems to be no similar issue under Linux.
https://ibb.co/RcpDJ9P
