[asterisk-users] How accurate is show translation?

[Leo Ann Boon]

Vicky wrote:
> I tried it on a intel 3 ghz p4 box and a athlon 3000 768 mb ram  
> running vista and host for centos 4 ( vmware ) considering  the load 
> on athlon running asterisk ( that too under vista plus vmware ) while 
> intel 3 ghz p4 1 GB ram box was sitting idle with centos , there was 
> hardly a 1 ms difference in show translation on both machines . 
> Besides i just compared my p4's results to ur D930 results and there 
> is no difference ( infact my g729 results are better than ) .. But 
> this doesnt mean both are same  dual core cpu's will definitely give 
> much higher number of channel transcoding then lower p4's . Put both 
> the box under some cpu load by other programs and then use show 
> translation recalc 30 and you will see performance difference between 
> them ;)
Vicky,

The point of the exercise is that you should run 'show translation' with 
no load to get the baseline value. Your results confirmed my suspicion 
that the value is not tied to the number of CPUs - which indicates that 
the test was run on only 1 CPU. My concern is why the performance 
plateau. It makes no sense that a 3GHz CPU should take the same amount 
of time as a 1.3GHz CPU - that is unless there's something else is 
holding back the transcoder. It's like those graphics benchmarks - at 
some point, all the CPUs show the same FPS because the refresh rate is 
the one holding up the CPU.

At this point, I don't feel that 'show translation' is a useful 
indicator of actual transcoding performance. It's OK for relative 
comparisons but utterly useless if you need the figures for sizing 
purposes.

Leo