Nuendo?Cubase 64bit floating point

I was just on a PT 10 webinar and many questions were had about PT11 and it being 64bit fp, my question is this, since N1.6 you have always been 32bit fp, when will Steinberg move to 64bit fp. you cant let avid beat you guys to the punch!!!


As far as I can tell there is essentially little advantage to move to 64bit float. I’m sure if you do a search you’ll find some good info on why that is.

PT11 doesn’t exist, and PT10 is still a 32bit application. I’m still on PT9, so I don’t know about the rest.


64-bit float brings me nothing I could appreciate, so far, other than a busier PC and the need for bigger HHDs.

16-bit = 96 dB, 24-bit = 144 dB, 32-bit = 192 dB, 64-bit = 384 dB
As the range of human dynamic perception is approximately 120 dB from silence to the threshold of pain. Do we need a dynamic range greater than offered by 32-bit float? The gain in accuracy, some claim to be audible, is wasted to a large part on not usable dynamic. I doubt that anybody human can hear the difference compared to a 88,2 kHz / 32-bit float recording.

Big K


A 64bit summing engine might- in some specific cases- be more accurate than a 32bit summing engine, but in practice, there is no advantage. What you say about dynamic range is not really accurate. It’s more or less accurate when saying that summing is adding dynamic range, so the bigger the summer (dynamic range) the better accuracy before bringing the numbers back down to 16 or 24 bit. So it are the intermediate stored numbers that get bigger in 64bit. That’s all.

All critical processes do benefit from 64bit, but most of these processes (EQ, Dynamics, etc …) already work in double precision. So when we compare a 64bit with a 32 bit engine, we are actually only talking about the internal “adder”, the summing engine.


AFAIK 32 bit floating point gives more then 1500 dB…

I’ll say this too:

Nuendo (and the other DAWs) already have summing engines that sound fine.

I’d rather SB spend time fixing remaining bugs and ill-conceived functionalities, and then add proper track grouping (a la PT), VCA’s possibly, maybe punchers/streamers for recording or ADR specific stuff etc.

There’s a bunch of stuff I think is way more important than 64-bit internal summing…

I’m in complete agreement with this.
I’d also like to add greatly improved surround support for the bundled plugins too, as the current multimono implementation is hopeless.

I can’t help it. I’ll post anyway.

A 64 bit internal path for plugins and for data streams within Nuendo would take care of some of the not quite out of range anomalies that can occur (and usually happen only when concocting tests to reveal the problem) when truncating that summed 80 bit floating point number back down to the 32 bit fp data stream currently in use. Instead the numbers would be truncated back down to 64 bit.

Nika Aldrich (spelling?) said some things would likely be moot as well in his view by going to 64 bit fp.

I couldn’t see 64 bit fp files, although someone would.

I know for general real world usage the 32 bit fp is just fine, but the current line of processors and general hardware handles these 64 bit fp data widths without issue, the market is going there anyway, and some esoteric issues would simply vanish.

My vote though is for fixes that truly address customer concerns in the existing product.

I’m assuming that currently with a 32-bit float system, that there is no way for the mixer to properly dither from 32-bit float down to 24-bit for playback? Granted, the distortion would be extremely low level… Still, would a 64-bit system allow for true dithering to 24-bit?

1st question: The consensus here is that you don’t need to dither this, although I do anyway (and I haven’t been on this forum in quite a while so take my “consensus” offering as something that should be looked into). Just load up the dithering plug on the output stream and make appropriate settings. You will have dithered output. It is considered normally unnecessary here because the couple of low bytes when converting to int in THIS case are mostly noise anyway and most folks are mixing for a maximum peak of just less than 0dbFS. See Fredo explanation.

2nd question, answer with a question, why would 64 bit fp not allow for true dithering to 24 bit integer? At least in this context the dither is not quite so moot or debatable as in the 32 bit fp to 24 int situation. Yes you can dither to 24 bit integer with appropriate dithering plugins.

Isn’t the Nuendo engine still 32 bit fp even though it is a 64 bit application now? Many other floating point DAW’s have a 64 bit fp audio engine although that kind of gets tossed out the door if you still load up 32 bit fp plugins.

You are micing up two different things.
A 32 or 64 bit OS is not related in any way to the internal bussing and adding structure.

Secondly, Nuendo is 32bit float in both 64bit and 32bit (OS) versions.

Thirdly, in Floating Point, no dither is required.

Best regards