[hpsdr] computer memory added question

[michael taylor]

On 7/15/07, stoskopf at tri.net <stoskopf at tri.net> wrote:
> I am using my test bench computer for a bunch of stuff and decided to add
> more memory.  Now using the Pentium D 3.3 GHz with 4 Gb of  800 MHz RAM.
> I had noticed with the SDR-1000/Ozy/Janus combo that using 2048 buffer and
> 19K sampling that I had quite frequent "Pops" in the audio.  Doubling the
> RAM took out almost all of that effect, but didn't change the CPU
> utilization.

That is not strange at all. The additional RAM gets uses as disk cache
for the hard disk and the operating system spends far less time
swapping in and out pages of virtual memory to disk.

So there are less interruptions (pun not intended) for the computer,
and the CPU can focus on decoding the audio, not doing
"house-cleaning" tasks like virtual memory management or waiting for
data to be loaded from the hard drive. The decoding of the audio still
uses the same amount of numeric computations, so the CPU utilization
should remain basically unchanged.

This is a good example why we weren't playing with SDR twenty years
ago, it was not just the cost/availability of high performance A/D
converters, but also the availability of low cost yet sufficient
computing resources necessary to decode signals in "soft" real-time.

-Michael, VE3TIX

 1184588812.0