Sample rates vs. latency

[filtersweep]

Reading the "problem with SPDIF" reminded me of a sample rate question that has been bugging me. When I first hooked AkoustiK Piano to my m-audio firewire 410 interface, I had the sample rate at 48K, yeilding at latency of 14ms. That's OK, but near the upper limits of where I would like latency to be. I thought that increasing the sample rate to 96K would mean the processor

(1.7GHZ Duo core 2) would have to work twice as hard and result in longer latency, but was happily surprised that the 96K rate actually reduced the latency by 50% to 7ms (which is great). What is the logic behind this? My guess is that the latency logjam (at least in this case) is not caused by processor load but may be instead related to how long it takes to get enough sample into the buffer before it can be played (i.e.,sampling faster loads the buffer more quickly, therefore less lag to playback???) Am I close?

[Cloacal-X]

That's pretty much it. when you set the latency, you're actually setting the buffer size, which is defined in terms of samples. The only thing to note is that you are working the processor twice as hard, so that may limit the number of plugins you can load, voices you can play, tracks you can have simultaneously, etc.

[filtersweep]

I guess I shouldn't have surprised me that working the processor harder would result in better performance. Instead, I was thinking that working the processor would push it past it's limits, but apparently I just haven't reached those performance limits with this processor. I take your point that working the processor hard for the sample rate might cost me in terms of numer of plug-ins, etc. Not really a problem for me because I use AP, B4, etc as stand alone apps and record to a Korg D1600. I want to experiment with Ableton or Sonar, but my plan is to just import audio to those apps from the Korg.