21 Sep
2008
21 Sep
'08
7:24 a.m.
hollunder(a)gmx.at wrote:
On Sun, 21 Sep 2008 01:06:21 -0400
"Hector Centeno" <hcengar(a)gmail.com> wrote:
This is partially true.
The fact that the transit has only a channel count of 2in/4out while the
FA-101 has 10in/10out + midi makes that it's not really a fair comparison.
However the main issue is that the kernel-space firewire implementation
is not very CPU efficient. There are some issues with how DMA memory
coherence is implemented that make things CPU intensive. Messing with
the kernel level implementation to improve CPU consumption is not
considered a priority for the freebob/ffado developers ATM. The 'new'
firewire kernel drivers will allow us to implement a more efficient
scheme reducing CPU. But let's first get the current FFADO out.
FreeBoB/FFADO themselves are fairly CPU efficient, although things can
always be improved.
Greets,
Pieter
PS: you don't by any chance use a freebob/ffado debug build?
Hello,
I have two computers (desktop and laptop) with Ubuntu Hardy (using
kernel RT on both) and a Edirol FA-101 firewire interface. I was
comparing Jack's CPU usage (no other audio app running) using the
Edirol against a USB M-Audio Transit. I noticed on both computer the
CPU usage to be much higher with the firewire interface than with the
USB. On my laptop goes up to 15% or more and 8% on my desktop
(Centrino Duo @ 1.7GHz and Core 2 Duo @ 2.1 GHz respectively) while
using the USB interface the usage remains below 1%. Is this a normal
behaviour of firewire interfaces? Is this a Jack related issue? Is
there a recent and more efficient version of Freebob (ffado maybe?)
than the one from the Ubuntu repos?
Cheers,
Hector
Are you sure that you are comparing the same latency settings? Just
a thought because Firewire is capable of lower latencies than USB and
lower latencies also mean higher cpu usage.
As said, just a thought, I never used FW.