How does Multicore works on U-He Synths? still a mystery

[accessdune3]

So how does it work ? or doesnt work ?

Its still a mystery to me. For my own curiosity i have tried to create same type of patch on Obsession and Repro 5. So i created and played 8 voice patch. It seems all good and the CPU usage according FL Studio metrics fairly the same actually when playing 1 instance with MC setting on.

But what i noticed is really weird, when i add more instances of each, Repro 5 CPU usage scales quite alot and in the end i was able to play 11x8 voices of Repro where's on obsession i just lost the count how many because it was more then twice and i didnt even count

So im genuinly curious, what does 1 instance of Repro (also noticed this with diva) do that it can take as much/least CPU as other competitor vsts but when added more instances the usage becomes so much more higher vs the competitors.

This is not to drag anyone im just really am curious how this works

[Urs]

Multithreading costs more CPU in total even though it costs less on any single core. I think we communicate this rather clearly in the user guide.

[accessdune3]

Multithreading costs more CPU in total even though it costs less on any single core. I think we communicate this rather clearly in the user guide.

What about the metrics show in the DAW? as an example of Obsession taking same CPU as Repro when using one instance but later on multiplying. Does that mean that Obsession in this case should be non-multicore?

I also noticed this with Zebra and Hive, they can appear that they take alot of CPU by the metrics when 1 instance is playing, but when piled up they really dont take that much of CPU.

[Urs]

I don’t know about Obsession. Hive and Zebra are relatively low CPU, hence have no multicore option.

[accessdune3]

I don’t know about Obsession. Hive and Zebra are relatively low CPU, hence have no multicore option.

It also applied to zebra and hive. Sorry if I frustrated you but i don't think you understud my given point or i maybe did not explained that well

[Urs]

I get your point, but it seems convoluted: Multicore on one hand, multiple instances on the other. There's no metric to compare things for me.

If I had to boil it into a single answer, I'd say that expensive algorithms which suit themselves for multithreaded processing generally don't scale as easily as inexpensive algorithms which may make better use of resource sharing. It's like sculpting a sandcastle with sand vs. pebbles.

[claudedefaren]

...

[Funkybot's Evil Twin]

I don’t know about Obsession. Hive and Zebra are relatively low CPU, hence have no multicore option.

Hive, low CPU? Relative to what?

RePro-5 and Diva.

[claudedefaren]

...

Hey mods, why did you prevent me from editing my next post in this thread? You wanted me to remove it so bad, so why are you actively preventing that from happening? Lot of people talking rubbish about an "agenda" here — what's yours?

[Urs]

Well, that’s just like your opinion.

[david.beholder]

Haha, that makes most every plugin on the market low CPU. Hive is certainly not low CPU relative to Zebra though.

Have you been living under the rock for last 5 years?
Also I would like to see your comparisons between Zebra, Hive and say Serum under the same conditions.

[claudedefaren]

Haha, that makes most every plugin on the market low CPU. Hive is certainly not low CPU relative to Zebra though.

Have you been living under the rock for last 5 years?
Also I would like to see your comparisons between Zebra, Hive and say Serum under the same conditions.

Looks like you didn't do your due diligence. Such a video already exists. Which one of us is living under A rock?

[Urs]

LOL, convenient settings chosen to demonstrate the "purpose"

[claudedefaren]

...

[Urs]

Hive uses as much CPU for 1x unison as for 4x and so on. Omitting that is convenient. Likewise, comparing to Serum and not using Wavetable modulation with unison is also convenient, but not a everyday situation.

But yes, I’ll happily admit that Hive is not optimized for setting which hardly need optimization.

Good luck with your ongoing endeavors to figure this out.