I am using Volumio to with a RPI and a Hifiberry Digi+ card and I do 24/96 resampling (Fastest Sinc). This uses around 40% of my cpu power.
When I connect to the device with the browser and start scrolling through songs, the cpu usage of www-data increases and I start to get stuttering audio. To avoid this I increased the priority of mpd to -10. There are no issues.
I wonder if there is a way to make the priority change permanent and if it’s interesting for the developers to include this in the next release.
Thanks in advance and I have to say, I really like my Volumio with its 24/96 bit perfect output
I have an edit for the post above. It turns out that the nice setting was not enough to avoid the stutter. When I load large number of songs to the playlist or use unison to sync my music to the hdd on Raspberry, I still get stutters.
Why upsample? You said you liked the bit perfect output, but you do not use it as it should. Bit perfect is meant to be identical to the source. So when ripping a CD for example to Flac, you should rip it at 16/44.1 AND play it back at the same rate. When you upsample to 24/96, the bits that are not in the original source have to be ‘guessed’ by MPD. Most cases the sound will not improve and for sure NOT bit perfect anymore to the source.
The high CPU load is caused by this ‘guessing’ of MPD. The Fastest method is the least CPU intensive but also the poorest one, giving you the least change of any improvement of the original source. Using a better method will increase CPU load.
My suggestion, really use actual bit perfect playback and turn off up/resampling. Focus on getting good source material, preferable lossless instead of MP3 or other lossy formats.
Hi,
I agree with PatrickB 's comment. Without resampling, playing a 16b/44 flac file eats less than <5% the cpu. And even through a network, with a wifi dongle, I’ve got no dropout or lag.
The “bit-perfect” scenario is not possible when you want to use software volume control, which is my situation here. I stream to an amplifier with fixed gain and I do the volume control with Raspberry. If I ever feel very audiophilish I may set the Raspberry to 100% and expect bit-perfect output, and then do the volume control in the amplifier, but that never happened
I want to use upsampling for many reasons, some of them almost superstitious and none proven. But the basic idea is that I have a 24/96 capable DAC and it is most probably upsampling the 16/44.1 signal to its native rate of 96k anyway. So if this will be done, it should better be done as soon as possible, because I will modify the signal with the volume control and I prefer to do it on signal with more samples. I believe the volume control will introduce less quantization errors when the signal has more samples. I just believe this because it is not proven to me otherwise.
