Since Cubase is not working with 32 bit plug-ins anymore, why do you still have the option to choose 32 bit processing for the audio engine ?

I've read the articles from Steinberg about it, but it's still not clear why they didn't just take that whole 32-bit option out...

You're confused.

64 bit plugins may still use 32 bit floting point arithemetic. The 32-64 bit Cubase applies the the memory model used by the applications not the instruction set. Applications may still use 16,32 bit floating point (and 128 bit sse registers as well) it is appropriate as well as 8,16,32,and 64 bit integers. The audio engine does calculations using 32 or 64 bit floating point arithmetic regardless of weather it is using a 32 or 64 bit memory model. Not that many plugins may not do 64 bit audio processing yet and cubase will have to convert to 32 bit before passing them to those plugins. You will basically only benefit from 64 bit floating point if you use enough plugins that support it.

Yes i am, that's why i'm asking this

So, basically you're saying that 64-bit plug-ins and 64-bit audio processing are different things ?

When i scan for 64 bit compliant plug-ins in the plug-in manager, only a few plug-ins show up : DMG, Boz, Slate, Steinberg (ofcourse) and a few single plugs from other brands. I would have expected to see at least Fabfilter, UAD, TDR and Izotope (to name a few) in that list too...

But still, it's weird that the option is still there..

Yes, they're two different things.

.... and more to the point can you actually in a double blind ABX aural test hear the difference between 32 and 64 bit processing!

Don't just engage it becuase it's there - it will cost some CPU cycles so there has to be a tangible benefit on your system to make it worth using.

There seem to be different opinions about the whole thing. Saw some posts on GS from respected developers who basically say that the whole 64-bit floating thing is unnecessary.

Yet another developer i talked to stated that both things (64-bit plug-ins and 64-bit engine) were basically the same thing and 32-bit mix engine could be dropped because Cubase only works with 64-bit plug-ins anyway. That's why i started this topic in the first place...

I personally use a lot of plug-ins that Cubase rate as being full 64-bit. But there's also a lot of stuff not showing up in Cubase's list, which have been confirmed by developers as being full 64-bit (or 80).
So i also don't understand how that list is made and if it's been updated with a Cubase update or if it's using some technical specs in a search.

But yeah, i personally don't hear a difference, even when some 80% of all plug-ins used are showing up in Cubase's list. I just thought about it the same way as with using dithering : just use it as adviced by most and don't think about it too much

First, in a situation where you want dample accurrate historical recall....you will leave the engine option there so Cubase 12 can open Cubase 8 projects and be spot on. Same with LPX opening L9 and below.

Second, as has been covered, the audio path and the codebase bit depth are tow different things. You can have a doible precision word procesisng any time you want. And you can have 64bit codebase where the audio is only 32. The old Cubase I usually use is this.

Ihappen to have recently dont some session null testing, and if you dont use the cubase channle EQ....the 32 and 64 null. It might apply tto other “channel” functions—i just rarely use their channel processing—found it would null everything EXCEPT where id done somrthing with their channel EQ. Most everything done with plug ins.

So, they wont likely “drop it”. They may change it to be the default (currently it defaults to 32 still)....but, it will be in there a long tike, if not forever. Cant you still write 16bit as the file format? How many decades ago was that a viable limitation?

Some VST plug-ins will be "blacklisted' by the Cubase plug-in manager. In addition to 32-bit plugins, Certain Izotope plugins being recognized is a known problem, as are certain pirated or incorrectly installed VST's.

Ah yeah, but that's not what i meant. In the plug-in manager you can do a scan for full 64-bit compliant plug-ins, which according to Cubase will work better with/make full advantage of their 64-bit double precision engine.

Aren't those two different things?

I was under the impression that Steinberg no longer supports 32-bit plugins and that that's what the scanning is all about. It has nothing to do with processing of audio and all to do with being written as a 64-bit plugin.

It's just that it gets confusing when Steinberg then starts talking about 64-bit processing in the audio engine because it really has nothing to do with the above.

Post #2 in this thread remains relevant, 1 week later.

Well I think it's been explained three times so..

From the Cubase manual :

Processing Precision

Allows you to set the audio processing precision to 32 bit float or 64 bit float. Depending on this setting, all channels are processed and mixed in 32-bit floating-point or 64-bit floating-point format.

To show all plug-ins that support 64-bit float processing, open the VST Plug-in Manager and activate "Show Plug-ins That Support 64-Bit Float Processing" in the Display Options pop-up menu.


So, even more confusing

Cubase doesn't actually know what precision is used inside a plugin. All that Cubase knows is whether a plugin can receive/send 32 or 64 bit data. That is what it is reporting.

You can have a 32 bit plugin (32 bit memory addressing) processing at 64 bit float internally but only receiving and sending 32 bit data. This is what many plugins used to do. Or you could (theoretically) have a 64 bit (memory addressing) plugin that processes internally at 32 bit float but receives and sends data at 64 bit. This is highly unlikely though.

The point I am making is that there are really three different aspects to a plugin when it comes to bit depth: The memory addressing aspect, the internal processing precision, whether the plugin can receive/send data at 64 bit float.

The memory addressing aspect is now limited to 64 bit, 32 bit plugins simply won't work, so we can pretty much forget about this. It doesn't affect the sound at all. The internal precision aspect is up to the developers. They, the developers, could always use 64 bit precision (or higher) internally so nothing has really changed here. I.O.W., we can also pretty much forget about this aspect as long as the plugin sounds good.

So we are left only with whether a plugin can send/receive in 64 bit. If it does, that's nice. We probably won't ever hear any difference but I'm a fan of a bit of over engineering so why not... but all in all, nothing to worry about.

I just have my Cubase set to process internally at 64 bit (which just means the summing really) and don't worry about it.

Alistair

Nice explanation, thanx.

The part that wasn't clear to me was how Cubase determines which plug-ins are put in their 64-bit list, but you say that it is determined by whether or not the plug-in sends/receives 64-bit data. (And therefor the 64-bit engine can work with it).

That also explains why some developers are claiming that their plug-ins are full 64-bit but are not showing up in Cubase's list. They do the first 2 things you mention (memory addressing and internal processing) but miss the ability to send/receive the 64-bit data. If i got it right

Anyway, a lot of confusement about this thing (or rather 3 things) called 64-bit...

Try opening a new empty session in cubase. Set processinng to 32 bit. Add halion plugin with acoustic grand piano sound. Play some big chords monitor in headphones.

save the session

now set precison to 64bit. (it will ask you to close thats why its haard to A-B)

Open the same session w halion and listen to the piano in 64 bit in the same headphones.

I could tell a big huge GREAT difference right away. So much so that I never want to use 32 bit ever again. SCARY! I thought it was a fluke since it did this by accident playing with my CPU settings

You are imagining things.. That setting only affects the summing engine. It does not affect Halion's internal processing.

Alistair

Maybe so... from reading about the 32 vs 64 bit processing it would seem that I am imagining things, but I swear I heard a pretty noticeable difference. I will try it again to see.

edit: I tried again result here: #27

In theory Halion is audio going into the summing engine, so if I hold a bass note with the pedal and then play a chord those sounds all get summed, no?

The different notes get summed within Halion and are then handed over to the summing/mix engine. The 32/64 bit switch in Cubase will have no effect on the summing in Halion. Cubase just sees a stereo (or surround or whatever) signal from Halion. That's all.

Alistair

Well the thing is that your DA converter can most likely only handle a fixed-point audio stream, so regardless of whether you are using 32- or 64-bit internal summing it all gets converted to that bit rate anyway.

For the purpose of you finding out if there's a big difference this is good. You can basically create a test yourself where you use the same content, sum it at both bit rates internally, and export to the highest bit rate the interface can handle. Let's say you choose 24-bit fixed point to store both at. You can then import both of those into a clean 64-bit float summing project and play them back at the same time with the phase inverted on one.

What do you hear?
What do you see?

I once went to an AP (Audio precision) thing at a hotel. (no masks and free lunch). At lunch i sat next to a dude who was into high end IEC AC power cables ($200+) and would sell them to customers and religiously swore that they sounded better because he would get a better 60Hz power cycle from the wall.

What I realized (this is gearslutz blasphemy or maybe GS propaganda ) is that if you PERCEIVE the difference and that makes you happy then do it (good for you)

In this case the 64 bit precision button is free (theoretically) and will also sound better (theoretically unedr some circumstances).

So, I will leave my nintendo sounding 32bit summing for the old geezers and get into the new millenium; where we are getting just a little bit closer to infinity-bit summing (analog, but didnt the old geezers have that?)

I like the idea of bouncing the track at 64 bit and 32 bit then inveting the POLARITY of one channel and checking the sum but im not convinced it works like that

I would rather just get the dalek to run without glitches/dropouts/jitter at the best audio quality as possible with analog speed latency.

Sometimes good advice. Other times bad advice.

It absolutely works like that. Inverting polarity on one source and adding it to the other will yield a difference if there is one. If there is no difference in the output, i.e. it is all "zeroes", then the two sources are identical.

If there is a difference then you can look at what that difference is and sometimes determine what caused it and whether or not it's of value. Say for example that there's a tiny difference at below -100dBFS, what could that be? Well, it could be dither. So if that's all you see and that's all you hear then again there's no perceivable difference in the music, just in the actual dither.

Often people won't do these things because they just want to believe.

Just want to point out that we old geezers certainly never had infinity-bit summing. On the contrary - the noise floor in the stuff I used in the 70s and 80s was typically way worse than what is achieved by even the lowliest 16-bit DA today.

And side note: the guy selling a $200 power cable “forgot” to mention that just before his cable the power probably travelled in your walls through the cheapest available wires, didn’t he

good point

Finally got back to DAW after some power outages last week (none of this stuff works with no power )

I tried this

Yes you are right. I recorded MIDI notes to halion track. Set processing to 32-bit and did export > audio mixdown. Then Set processing to 64 -bit, restarted Cubase and did same export > audio mixdown (both files 48k, 64 bit float .wav),

Then imported them on separate tracks starting from 0, and inverted the polarity of one.

Master meter showed NOTHING. They were completely cancelled.

Good advice

I will try again with 24bit mixdown, but I assume there will be no difference.

Interestingly though, I let one audio mixdown unmuted and played it back with the Halion instrument track in inverted polarity. Mostly silence, BUT there were intermittent spikes on the meter and could noticeably discern piano sound transients in headphones. Tried this using several different buffer settings, same result. Jitter?


After doing the above test I sat back down at the piano with Halion opened and played some notes (which felt good after all this hacking and Gearsluting)
Then I restarted Cubase in 32 bit mode and 64 bit mode and switched several times.

I couldn't tell the difference.

The first time I did this, I was experimenting with other settings and different patches so it could be that a switch got flipped in between, or just that my ears were foggy after all the experimenting.




Also, I found this Halion GM 001 Acoustic Grand to be a great patch while playing and listening to yourself in headphones, however when summed down to mono, it sound terrible and there is a noticeable chorus modulation effect somewhere. I will post more about Piano sounds on a separate thread. HERE: Stock Virtual Instrument Acoustic Piano comparison