AlexM wrote: 
> About a year ago I replied to the TT3 thread with some measurements
> comparing CPU utilisation in the Touch at 16/44 and 24/96 with and
> without server-side FLAC decoding. For my tests, I used a wired Ethernet
> connection. CPU utilisation was measured simply by connecting via SSH
> and running 'top'
> 
> I don't have the exact figures to hand, but IIRC at 16/44 the difference
> was something like 10% (via FLAC) vs. 30% (PCM to SBT), and at 24/96 it
> was (55-65% vs >90%). I couldn't detect any beneficial effect of
> streaming PCM, but this did establish that the theory supporting
> improved SQ with at least this component of TT3 mods was a bust.
> Unsuprisingly nobody could think of a reason why more CPU utilisation
> would be better, especially since the theory was based on the opposite
> being true. Suprising that nobody bothered to check before, especially
> as it was so easy to do.
> 
> Ah.. here we go.. found my tests on a different board -
> http://theartofsound.net/forum/showthread.php?t=15594&highlight=Touch+CPU&page=8
> 
> Archimango - excellent work on the tests BTW. I also have a Emu 0404
> USB, and have long been telling people what a cracking interface it is,
> especially for the money!. Your figures do show it to be a very
> transparent device for A/D. You have inspired me to do some testing of
> my own to optimise the noise performance of my system.
> 
> Cheers,
> Alex

Thanks for the note Alex! Glad to see another 0404USB user :-).  Let me
know if in your testing you find a good way to measure JITTER!

Good point about checking on the Touch's CPU utilization. This AM I
logged into the Touch and ran 'top -d1' to check CPU usage every second
while playing music. I chose Prince's 'Batdance' since I had a good
24/96 vinyl rip as well as the original 16/44; I figure a dynamic track
like this could give the Touch's FLAC decoding a nice work out. This is
with my Touch running default EDO firmware settings, ethernet
connected.

For MOST of the time, the CPU use looks like this:
24/96:
FLAC level 8 compression: 20-30% default view, 70-85% peak with spectrum
analyzer

16/44:
FLAC level 8 compression: 15-25% default view, 65-75% peak with spectrum
analyzer

Clearly turning on the spectrum analyzer "Now Playing" view eats up CPU
cycles significantly more than just the playback itself from what I
see.

About a couple of times over the 4 minute song, the jive_alsa process
wakes up in a big way and this pushes CPU use to 90-100% for about 5-10
seconds. I believe jive_alsa runs at realtime priority level and there's
a discussion about it elsewhere. Whether this changes sound quality, I
dunno.

SO even though I couldn't find any sonic change going the way of TT3
with turning off the screen and everything, I wonder if doing the
opposite - stressing out the CPU with 24/96 FLAC + spectrum analyzer
view running during playback lead to stuff like increasing noise
floor...  Maybe I'll give this a try tonight.


------------------------------------------------------------------------
Archimago's Profile: http://forums.slimdevices.com/member.php?userid=2207
View this thread: http://forums.slimdevices.com/showthread.php?t=97950

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/audiophiles

Reply via email to