Why does Wave Lab behave better when 32 bit float vs. 64 bit float is used?

First off, the majority of my plugins are 32 bit, and granted, it would make more sense to setup WL for 32 bit float considering this. However, Cubase 12 responds better overall (using all the same 32 bit plugins) when I set it up for 64 bit float. And when I say better ‘overall’ this would be while gauging the CPU usage and while checking the usage meter in Cubase. (My PC is an i7 intel 8th Gen processor with 6 cores and running Win11 64 bit.)

Cubase does have a few things WL doesn’t have, notably ASIO Guard and the option for the use of multi processors, yes, and these boxes are checked in Cubase. However, is the answer this simple? Does WL not incorporate such things (ASIO Guard and multi processor use) because it’s not a benefit to Mac users? Is the WL motive here to cover all it’s bases (Mac and PC)? Or has somehow WL put something in place to accommodate both MAC and PC equally that I have not discovered?

Added Note: Now, having switched to 32 bit Floating, my 32 bit Ozone plugins open and close via the DBL click on the GUI header without a hitch. Interesting… :confused:

This meter does not represent the same thing in Cubase and WaveLab. In WaveLab, this shows the peak CPU. In Cubase, this value is divided by the number of cores, AFAIK.

WaveLab 11 has both ASIO Guard and multi processor support. This being said, Cubase is more capable of low-latency playback compared to WaveLab.

1 Like

Thanks for the responses, PG. I did have that Multi Core box checked but I did not recall the ‘ASIO-Guard’ header in the Connection dialog. Which is funny because I check this connection page often!

Now, if all my plugin creators ever comply with 64 bit architecture, then I will switch back to 64 bit plugin processing in WL. At this point I have no idea why Cubase 12 doesn’t seem to be bothered by this feature and actually seems to perform better with it setup this way. Oh well, very different programs on some levels.

The 64bit architecture has nothing to do with 64bit processing as you’re referring to it.

64bit processing means that the audio data is calculated with 64bit precision instead of 32bit or 24bit.
Maybe @PG1 can explain it in more detail.

If a plugin does not support the 64-bit resolution (“architecture” is not the correct term), then WaveLab uses 32-bit for this plugin, automatically. This being said, 64-bit processing is more CPU-consuming.

Thank you for explaining this, PG.

Q: If I were using all 64 bit plugins, would it follow then that - if I chose 32 bit resolution - WL would automatically convert the 64 bit plugins to the 32 bit float processing?

Yes. Actually, a plugin declares to the host (e.g., WaveLab) if it is capable of processing 32-bit or/and 64-bit samples. Typically, all plugins provide 32-bit sample support. According to the WaveLab setting, WaveLab adapts the internal sample bus.
Therefore, saying “64-bit” plugin is more like saying “Dual 64/32 bit plugins”.


Thanks for this response, PG. However, when you say ‘typically’ there must be some exceptions to the rule?

For example, I have plugins by SIR Audio Tools that - during the install process - list the option to include (OR NOT) VST3 32 bit versions of the plugin. I do not load these 32 bit versions but instead load only the VST3 64 bit versions. And actually, these plugins do not seem to have a problem operating in WL.

So, another Q: If a VST3 64 bit plugin can be operated correctly in a 32 bit float process environment by some type of a dual (64/32) bit ‘awareness’ by WL, why does the plugin designer offer the VST3 32 bit plugin versions? It seems this 32 bit offering would be unnecessary. Is this offering then purely for 32 bit Mac systems?

I know none. Since some host only support 32 bit processing, I hardly imagine some plugin not supporting 32 bit processing.

This is about the architecture. This is not about the sample resolution. Today, whose using a 32 bit operating system ??? Only use 64 bit plugins.

Again, don’t misunderstand the OS architecture and the sample processing resolution.

1 Like