I thought I would run the same benchmarks on my PC that I have been running
on my Pi. This is a very different beast as it is a i7-4770 with 4 cores (8
with hyperthreading) and 16GB RAM.
This is a bit academic as any x86 PC in the last 10 years can encode very
quickly and unless you are converting a huge catalogue you will not notice
a small change in speed.
That being said I wanted to see if the speed changes I saw on the Pi were
reflected on the PC as well. In short they seem to be. It is a mixed bag
for 1.1 which is faster at some bitrates only on some complexities. Overall
it is slower than 1.0.3 or even 0.9.14.
Again on a PC it is probably worth it for the new features and quality
improvement but just thought I would show it.
As it went so fast I had to change how I got the speed. I now measure
elapsed time in seconds to complete. This flips the graphs from my other
benchmarks as lower is now faster. I am re-running this on the PI as well
so they are consistent but it will take another 24 hours to complete.
As the PC is so fast I ran at 6, 16, 24, 32, 64, 96 and 128kbps.
The charts can all be seen here:
https://docs.google.com/spreadsheet/ccc?key=0AmUOhhaYBtrKdDZlUThEUWdJRG9zM2VPNTFTZ29PWkE&usp=sharing<https://docs.google.com/spreadsheet/ccc?key=0AmUOhhaYBtrKdDZlUThEUWdJRG9zM2VPNTFTZ29PWkE&usp=sharing>
Next music and then I will try fixed point.
Best Regards,
Stuart Marsden
Tactical Communications Consultant
FinMars Consulting Ltd
UK: +441865589833
Finland: +358453046287
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.xiph.org/pipermail/opus/attachments/20131222/4e8bf7c4/attachment.htm
