Ram vs. CPU when it comes to Multi's - What is the biggest bottleneck?

MyStudioOne

I don't recall seeing a thread about this subject before...

My understanding has always been that the instrument patches are loaded into RAM and so having sufficient RAM makes the bigger difference if you want to load a multi. Negligible difference on CPU side since you are likely not playing all the instruments at the same time.

To the real technicians out there, is this correct? And if so, what would be the ideal minimum amount of RAM in order run multi's comfortably?

Perhaps most of us with towers have plenty of RAM these days but when using a laptop with limited RAM for instance, would it not be better to use single instances and avoid multi's? I am just curious...

JesterMgee

It depends on many factors really.

It helps to start with understanding the difference in sample plugins Vs synth plugins as this determines firstly if RAM or CPU will be the main factors. Next is understanding how modern DAWs work with modern multi-core processors.

A Sample Streaming plugin (like Kontakt for instance) is a RAM heavy plugin especially when you load up many layers as the instruments are all large multi-sample libraries that stream either from your HDD or loaded into RAM. Your CPU is not used much at all and this is why it is often easy to load many layers in Kontakt without hitting many CPU issues.

A Synth Plugin (like Massive for instance) has almost no RAM requirements but has to generate all sound on-the-fly and there are MANY factors that impact how this can be impacted. Since sound has to be generated and processed in Realtime there is a larger CPU load and effects such as Reverb/Delay and processes such as Unison and large voice counts can add a lot of additional processing and increase CPU load.

A DAW is usually designed to make the most of the available cores and threads and usually the way work is split up over cores is track by track so a track is assigned a core. If you have several synth plugins over several tracks then this is split over several cores. However if you load several plugins within a multi-host container on a single track then only 1 core may be used and this can cause issues with performance.

There are loads of other factors but this is exactly why it is almost impossible to answer those common questions like "how many tracks can I load using this setup".

So in some cases you need more RAM, some cases you need more cores and some cases it is much better having less cores but with more power per-core, it comes down to how each person needs to work and what kinds of instruments they usually use. This is why you can see some people have 300+ tracks in their project and it never skips a beat because they are mostly samples and samples instruments while you are sitting with just 4 instances of serum and your machine struggles to run.

Hopefully that helps understand things a bit, does get more complicated than that but that's the general gist.

JesterMgee

It depends on many factors really.

It helps to start with understanding the difference in sample plugins Vs synth plugins as this determines firstly if RAM or CPU will be the main factors. Next is understanding how modern DAWs work with modern multi-core processors.

A Sample Streaming plugin (like Kontakt for instance) is a RAM heavy plugin especially when you load up many layers as the instruments are all large multi-sample libraries that stream either from your HDD or loaded into RAM. Your CPU is not used much at all and this is why it is often easy to load many layers in Kontakt without hitting many CPU issues.

A Synth Plugin (like Massive for instance) has almost no RAM requirements but has to generate all sound on-the-fly and there are MANY factors that impact how this can be impacted. Since sound has to be generated and processed in Realtime there is a larger CPU load and effects such as Reverb/Delay and processes such as Unison and large voice counts can add a lot of additional processing and increase CPU load.

A DAW is usually designed to make the most of the available cores and threads and usually the way work is split up over cores is track by track so a track is assigned a core. If you have several synth plugins over several tracks then this is split over several cores. However if you load several plugins within a multi-host container on a single track then only 1 core may be used and this can cause issues with performance.

There are loads of other factors but this is exactly why it is almost impossible to answer those common questions like "how many tracks can I load using this setup".

So in some cases you need more RAM, some cases you need more cores and some cases it is much better having less cores but with more power per-core, it comes down to how each person needs to work and what kinds of instruments they usually use. This is why you can see some people have 300+ tracks in their project and it never skips a beat because they are mostly samples and samples instruments while you are sitting with just 4 instances of serum and your machine struggles to run.

Hopefully that helps understand things a bit, does get more complicated than that but that's the general gist.

LostInFoundation

Now we even start to see GPU plugins, helping to alleviate work load on CPU ☺️

MyStudioOne

https://community.native-instruments.com/discussion/comment/87871#Comment_87871

Thoughtfully answered. Thank you.