[citation][nom]to-pse[/nom]> has zero latencyThat's bull - no system has zero latency - not even DSP-based ProTools.All soundcard-hardware has at least 32 samples of latency.[/citation]
I meant, zero latency caused by the OS - of COURSE you'll always have latency caused by various system elemens! However, harware latency is usually stable with little or no latency (don't forget to fine-tune your PCI bus latency in BIOS if your sound card is PCI-based, for example), so it can be compensated for (that's what fine tuning is about). However, while there's no way to completely remove hardware latency (if only for physical reasons), there's a way to remove software latency - by allowing real time access to the hardware.
The NT kernel is NOT real time capable: even XP, which allowed direct access to the hardware, didn't preclude kernel preemtion. Vista, which does not allow such access without dedicated drivers and APIs, introduces far more latency for purely software reasons, and also doesn't preclude kernel preemption.
But Linux can be made to PREVENT preemption on a process requiring real time access! Real time means that the process has full access to the CPU and other devices and won't be forced to relinquish these accesses even if other processes require CPU time (yes, this is inherently a kernel process' schedulerproblem, see); this is where a quad core actually makes sense, in that Linux cal allow a process to run in real time on a CPU and piece of hardware, and let the three other cores ready to deal with the rest:
- audio filters and processors
- memory and hard disk controller
- GUI
Now, hardware may have 32 samples of latency, but stop me if I'm wrong, 32 samples on 96,000 Hz single channel audio is a 1/3000th of a second, or a third of a millisecond.
Windows' kernel latency is around 13 millisecond, to be added to whatever latency the software running on top of it is.
Linux' -rt latency is zero.
While not all plugins run on Linux, there are enough professional-grade mixing software there that people actually use it professionally - if only because it has the lowest latency broadly available out there, without costing an arm and a leg: in fact, it's free.
One doesn't look at a given horse's mouth; but when said free horse is in perfect health and runs faster than a paid-for horse, then maybe one should.