I have a theory as to what is going on, and SOME actual measurements to back it up.
The theory is that what is most important is main memory access, NOT processor load. Think about it for a minute, the processor is connected to the main memory chips with a bus that is hundreds of bits wide, when a memory access occurs a large percentage of these switch all at the same time (actually within 50 pico-seconds of each other). The capacitance on a wire going from one chip to another is orders of magnitude greater than the capacitance inside a chip, thus every time that external bus switches huge currents flow in order to charge and discharge those capactitances. This produces huge currents in the groundplane and PS traces as well as EMI emitted from the signal wires themselves. My analysis of groundplanes indicates that noise caused by the memory interface is much larger than that from the processor itself doing its thing. Thus the most critical thing you can do is to cut down on main memory accesses. The easiest way to do this is to make sure that as much as possible stays in the processor cache. This I think is where most of these tweaks actual work. Decreasing the buffer size means that frequently the buffer can stay in cache and the data doesn't have to go in and out of main memory as much. Same for processes, every time the OS switches to a new process the cache is cleared and code and data have to come from main memory. This can also help explain things like flac or PCM decoding. The flac decoding may not take all that much more processing power, but the code is much more complex, the PCM code will have a higher probability of staying in cache. This may not be everything that is going on, but I think it is a major contributor to what is happening. Unfortunately there is no tool you can run that gives you a graph of processor cache misses, it would take a logic analyzer looking at the memory bus. I think that focusing on processor utilization is a red herring and not worth worrying about as long as it doesn't get too high. John S. -- JohnSwenson ------------------------------------------------------------------------ JohnSwenson's Profile: http://forums.slimdevices.com/member.php?userid=5974 View this thread: http://forums.slimdevices.com/showthread.php?t=91322 _______________________________________________ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/audiophiles
