What is the advantage of 64-bit OS for audio?

As I understand it, a 64-bit OS (like Windows 7/64) uses 64-bit data words (memory chunks). Since the biggest audio bit depth used today is 24-bit, that means over half of every 64-bit audio data word is filled with zeros (i.e., “zero padded”, no useful data).

Even though a 64-bit machine can crunch numbers with more precision, it seems a lot of that power is wasted crunching zeros. And even though a 64-bit machine also has a greater data address capacity it seems it is also wasting half of it!

So, what is the advantage of 64-bit OS for audio?

OS64/Cubase64 - pros: unlimited memory cons: unusable bridge for 32bit plugins (not every 3rdparty plugin exists in 64bit version)
OS64/Cubase32 - pros: most of 32bit plugins available cons: 4GB limit, Rewire not working
OS32 pros: most of 32bit plugins available, cons: very low(!) 2GB memory for Cubase
Speed differencies OS64 vs OS32 - ± neglectable

First of all, you understand it wrong. The length of data words used in OS for addressing RAM has nothing to do with the audio bit depth. Vras has summed up the pros/cons correctly, although I think that the cons for the first option (64/64) are diminishing as time passes by. I also wouldn’t say that the internal Cubase bridge is unusable since it highly depends on the particular 32bit plugin you want to bridge. I personally bridge TC Electronic Fabriks (R/C) without any problems…

So, I recommend to go for the 64/64 option. This is the right direction :wink:

Ah, I see the “more is better” philosophy is still alive and well.

I certainly do understand the difference, but you don’t understand my point.

Audio data conversion has reached a human hearing upper limit of 24 bits which is very efficiently handled by 32 bit CPUs. We don’t have - or will ever need - 64 bit audio. Much of the processing power and memory used by a 64 bit CPU to process 32 bit data in this case is wasted. The technology has exceeded the NEED.

So, it seems that all we gain by running Cubase in 64 bit is the ability to create GIANT projects. If that’s the case perhaps we should spend some time re-addressing how we make music. :wink:

… and the ability to access more than 4 GB of RAM.

I’m afraid you still understand it wrong. Your idea is that one 24bit sample takes 64 zero-padded bits in a 64bit OS, but this idea is not correct (unless you use 64bit floating-point precision for the internal processing). Let’s think of DSD audio consisting of 1bit samples. Do you think that a 64bit processor processes DSD signals by 1bit chunks per tick? Certainly no!

The 64bit data word enables us to address practically unlimited memory space. This is the only (but huge!) improvement over 32bit systems. If you worked with big sample libraries, you know that the 2 GBs of usable RAM was really an annoying limitation in 32bit systems. So, there is really a practical need for 64bit systems. Nevertheless, 24bit audio is processed in the same way, regardless of whether you use a 32bit or 64bit OS.


The address bus and data bus/busses are not the same thing!

None that I’ve seen on Mac. On Windows, it lifts the MS restriction of 2gb (hackable to 3gb) of user mode memory…allowing you to run their ram sucking but much improved OS (Win7) and still have a comfortable windows of RAM.

I dare say I’ve never popped 1GB (for Cubase itself) for an audio production project of any size-and approaching that likely means there are drum samples loaded in somewhere-so not even strictly audio.

It’s of major use for virtual instruments.

People wonder why companies like Waves were so slow to get 64bit versions of their plug ins…audio recordings of real instruments are what are used by people who are not making demos and care what things sound like enough to put thousands into plug processors. So, Waves was being asked to recode their AUDIO plug ins-and the new versions would sound no different…effectively for people who use a lot of virtual instruments.

Now, in theory…once Cubase decides to implement a 64bit float ENGINE (like Sonar has I believe)…it should result in more “headroom” (sic)–akin to my old 56bit fixed Akai or 48bit fixed TDM mixers. But, at the end of the day, that’s mostly a workflow improvement-being more forgiving of noobs sending hot tracks to be mixed…although, that’s debatable–if an analog model expects -12 or -18 as Odb equivalent, you may still need to trim down the tracks pre fader. But, correct me if I’m wrong, Steiny hasn’t even implemented that with C7. So, for the purpose of this board discussion, there’s none.

This topic has apparently revealed more confusion among the audio community than I thought there was! Guys, I’m repeating myself: the 64bit floating point audio engine has nothing to do with the 64bit operating system. Sonar makes use of its 64bit audio engine even when you run it as a 32bit application on a 32bit operating system! In 64bit Windows you can use more RAM, that’s all! And 64bit plugins have to sound exactly the same as their 32bit versions. So, 64bit operating systems do not improve audio itself in any way. However, if you don’t want to be constrained by RAM limits, you should go for a 64bit OS.

Good point. Now what is the case with the latest generation of PCs running Windows 7/64, etc. - do they have 64 bit data busses or just 64 bit address busses? I assumed it was both.

It would actually be ideal, I think, (for music purposes) to have a 32 bit data bus and 64 bit address bus. That way you are using all of your memory for useful data and not number-crunching zeros.

Have you ever written anything in assembler, regardless of CPU, or at least looked at disassembled code?

Yes I have and that’s why I’m asking these questions. Many microcontrollers that I have written code for (ATMEL, Intel, etc.) have an 8 bit data bus and a 16 bit address bus - and they are referred to as “8 bit processors”.

But I don’t know what the latest and greatest “big league” PCs and OS systems are using. Is it 32 bit data, 64 bit address or is the whole thing 64 bit?