[FFmpeg-user] Not using all resources

Henk D. Schoneveld belcampo at zonnet.nl
Fri Sep 11 15:41:53 CEST 2015


First tried to upload to incoming but ii’s presented to me as RO

Uploaded a 20MB file for someone interested in investigating at:
http://we.tl/CpG3aaamaO

On 08 Sep 2015, at 02:10, Andy Furniss <adf.lists at gmail.com> wrote:

> Henk D. Schoneveld wrote:
>> Hi all,
>> 
>> I’m encoding from several DVB-S sources, satellite, with always the
>> same parameters but recordings from some sources CPU-load is about
>> 50% of what it could be, others use max. available as attachments
>> show.
>> 
>> The 1st inserts a srt file while the 2nd doesn’t but because there’s
>> no IO-bottleneck that should be a problem I think.
>> 
>> Any idea what could be the cause of it ? It’s not really problematic,
>> just curious.
> 
> May not apply in this case/your CPU but -
> 
> Cpufreq on_demand is sometimes stupid and top doesn't always show the
> full picture especially if simd is involved (may depend on h/w) and I
> guess if cpufreq on_demand is in use % figures are meaningless without
> polling the current frequency anyway (accepting that if they are > 90%
> freq should have risen).
> 
> It could be my observations are just because I have an old AMD Phenom -
> but I once came across an old ffmpeg-devel post where some one made some
> C version of CSC code that beat the SIMD ASM. Shame it was old or I
> could have posted - yes I see exactly the same - but that's because
> cpufreq on_demand doesn't kick in in the SIMD case - set the CPUs to
> perf and the SIMD will beat the C.
> 
> Above just to illustrate that "just looking" with top could mislead -
> not saying it does for you, but always when benchmarking/measuring
> consider setting CPUs to perf.
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-user



More information about the ffmpeg-user mailing list