I spent a lot of time on optimizing my Totaldac/Cubox-i4Pro/Armbian Roon/Squeezelite image. For those interested here’s some output on how this turned out. As you can see there’s hardly any load to speak of (all zero’s) when playing a Qobuz FLAC 9624 file. And we are looking at a quad core 1GHz i.MX6, same thing you’ll find in your avarege fridge… The moral is that big processors for endpoints is a waste of money and energy. Now that’s green audiophiliac…
That is well optimized. It would be interesting to see the load with a convolution filter of some type applied and maybe playing a DSD128 file. Both seem to add a lot of load to a Roon core. I assume you meant Roon core and not Roon endpoint?
Negative Baldy, this is the endpoint… That’s what I like about Roon (and LMS for that matter) is that processing and buffering is done on the “central” server where the power and the speed is. In order to leave the “playing” to the delicate ultra optimized endpoint… My Roon core is a HP DL380 G8 rackmount full blown MF with all the bells & whistles on it. And if I do all that stuff you mentioned it is working for sure…!
Cool. Thanks for the quick reply!
I could not agree more.
Finally someone agrees with me, thanks @djsamibop
sooo, how does it sound?
have you tried to bind (pin) Roon to 1 specific CPU core, while other processes (and irq interrupts) to the rest?
i was doing something similar with rpi3 (heavily underclocked and undervoltaged cpu, ram etc) based on tinycore linux
one is able to actually hear impact of the optimizations , it’s just sometimes not clear which one is actually better
Yes @Maniac, as you can read in my initial post I did change processor affinity for all Roon related tasks from 0-3 to 2-3, all but USB to 0 and USB to 1… I must say I don’t hear the difference, but others (say they) do… The interesting part is the fact that we here in this forum are so close to the edge of possibilities that every bit starts to count.