Re: [Discuss-gnuradio] processing power

[Eric Blossom]

On Wed, Mar 20, 2002 at 08:34:47PM -0700, Stephen Nichols wrote:
> 
> Ok, another question. What about processing power? Again, I know it would
> depend on the signal and the amount of sample data you are dealing with but
> are there any ballpark figures ? Take that microtune board at 20Ms/s as an
> example. What would be the low end of computer processing power you would
> need to work with that amount of data ?

You could probably look at all the data with a 450 MHz Pentium III,
maybe even run a single FM demod without GUI.  800 MHz allows for more
stuff, including at least the single channel broadcast FM demodulator
with GUI enabled.

On our end, we're currently using dual processor Athlon 1800+ MPs
(entire system about $1400 from local shop with a pile of memory, big
disk, etc), although for most stuff any high end single processor
should work fine.  (The 10.76M symbol / second rate of ATSC is pretty
demanding, hence all the horsepower we're throwing at the problem.)
The dual channel broadcast FM demodulator runs fine on a single
processor 1.7 GHz P4, and that's without the MMX stuff enabled.  
A 1.5 GHz PIII would probably be roughly equivalent.

FWIW, at this stage of the game, we've got our attention on making it
work.  We're not to the making it fast part.  We figured that a $1400
investment in fast hardware allows us to dodge the performance bullet
for a while.  Programmer time vs hardware cost... YMMV.

Eric