| Author
|
Watz up with 64 bit Processing ???
|
ird
Started Topics :
1
Posts :
16
Posted : Jan 30, 2004 00:13
|
"BUT, other than the mac high end platforms, there's no one standard for 64bit CPU processing"
I would assume that most of the todays software plugins are coded using 64-bit floating point arithmetics ("double" type in c), which has been a standard for over a decade, and is supported by all modern FPUs (even the old 486 can do that in hardware, if I'm not mistaken).
In fact, using floating point computations is a standard for VST, since VST plugins receive a stream of floating point numbers (32-bit floats). Each "sample" is represented as a floating point number from -1.0 to 1.0 (as opposed to the -32768 to 32767 integer representation used for the 16-bit values that are fed to the DACs).
All calculations are performed on floats
or doubles (after conversion), and then the resulting values are passed back to the VST host (for example, cubase). All internal processing and mixing is done on floating point numbers, and only at the final stage they are converted into the 16-bit output stream.
Things are a bit different in the hardware domain. Hardware devices (synths, FX) typically use specialized DSP processors, which often do not support floating point operations. For instance, the Access Virus synths use a Motorola 563xx series DSP, which is a fixed-point 24-bit device. This means that it operates on 24-bit integers, not floating point numbers.
My $0.02. |
|
|
Kaz
IsraTrance Full Member
Started Topics :
90
Posts :
2268
Posted : Jan 30, 2004 01:25
|
Superglue: because a 64bit platform that is 1.5x as fast as a 32bit one on regular operations is 2x (or more, since more instructions to utilize this structure are added). Windows does NOT work in 64 bit. It is happily 32bit, and it takes a lot of hard work to make a new 64bit OS while keeping compatability with all the old software. Just like when in win2k you work in an emulation mode for 16bit OS based software (which is much slower and a lot buggier).
If you want to work efficiently with 64bit software, you need 64bit hardware (which if coded specifically for can add a lot of complex operations that it does more than twice as fast as 32bit chipsets) and a 64bit OS (for high stability, standardizing the communication with the hardware and a lot of other bottlenecks). An EQ that works with a lot of information would run 3+ times faster if the entire system was at the same speed at 64bit. The problem is not what you can or cannot do, the problem is how fast you can do it   http://www.myspace.com/Hooloovoo222 |
|
|
|
Kaz
IsraTrance Full Member
Started Topics :
90
Posts :
2268
Posted : Jan 30, 2004 18:34
|
Superglue: it's about as 64bit as win95 was 32bit. That means it supports a few 64bit operations - probably in order to optimize server work and nothing more. I doubt that it's very relevant to soundwork, as opposed to the mac's 64bit OS which utilizes everything that the CPU has. And windows server 2003 is a VERY heavy OS, I doubt that the extra oomph given by the 64bit code is actually worth it.
As long as MS is coding based on older software it will NOT be true 64bit work.
http://www.myspace.com/Hooloovoo222 |
|
|
|
ZilDoggo
Started Topics :
4
Posts :
663
Posted : Jan 30, 2004 19:13
|
kaz,
actually, win2003 is the final developement of what already started in win nt.
win2003 is running in a virtual machine and this machine can be ported to almost any platform (including linux and mac)
It's a bit like a virual pc on which windows runs.,
Things like 32 or 64 bits only realy matter for the virtual machine not for windows itself.,
But then software that calculates on 64 bits data will run faster on a 64 bits machine.
superglue,
mac g5 is so expensive because it's got a lot of tricks up it's sleve. it can for instance do about 256 instructions SIMULTANIOUSLY under certain circumstances.,., that's bloody fast. and very good for music i can imagine..
unfortunately all software will have to be optimized to use such features and this process takes time.
The question is wether NOW is a good time to get one.
grts,.
aka.,,!., |
|
|
|
thockin
Started Topics :
1
Posts :
114
Posted : Feb 1, 2004 01:15
|
64 bit processing in audio software is called 'double precision'. There is a standard (IEEE 754) floating point format. That format can be 32 bit (single precsion) or 64 bit (double precision).
Single precision is fine for many uses, but can start to show loss when you use lots of very very small calculations (such as audio). In that case, going to 64 bit is a win. But ONLY INTERNALLY. There isn't any point in sending 64 bit data (double precision) between plugins.
You see, 64 bit is twice as large as 32 bit in memory space, so sending 64 bits where you could send 32 just makes everything SLOWER.
|
|
|
|
|
