[ardour-dev] Ardour lab

[Doug]

First, thanks to Kjetil for mentioning the bus speed issue, that was the 
problem exactly.  The CPU dip switches are all set to AUTO by default, 
but there is a single bit dip sw behind one of the PCI slots that is set 
to ON for a 200Mhz lock, by default.  I'm now at 2400+ (2Ghz).  That'll 
learn me for not reading the manual...

Sorry about the lack of info, I'm going to put all the specs on my 
website then post results to the list, when I get so motivated :)

Here's more info:

Both PC's are running Slackware 9.1.  All software installed is from 
source.  I try to keep up with latest ardour/jackd cvs builds,and as of 
last night I was:

Ardour/GTK 0.454.0 running with libardour 0.729.0
jackd 0.95.3

tapedeck is linux-2.4.25 w/ low-lat patch and capability.h mod
athlon1 is ditto except linux-2.4.23 (soon to be .25)

They are both dual boot to 2.6.3 as well, booted into each one to get 
everything working , then went back to 2.4.x till I hear better news 
about realtime stuff.

And, now after correcting the CPU issue, tapedeck appears to be 
performing better.  I now get pretty similar performance between 
tapedeck and athlon1:

Using two projects, both 48Khz 8 track PCM:  2003-01-03 is 46min/4.1GB, 
2003-02-14 is 55 min/5.1GB

The projects load well on both machines, navigation thru project is 
great on athlon1, on tapedeck its a little slow till I carefully scroll 
from end to end once, then its great.  The bigger one requires more care 
not to get a disk stream error on tapedeck until scrolling end to end, 
then its fine.  I'm thinking there arent any real *problems* here, these 
are big projects, and I think most are accustomed to working with single 
songs at a time or?  I am adding 1GB ram to both machines (for a total 
of 1.5GB) this week, along with a seperate OS / data hard drive scheme.

Later today I will re attempt recording on tapedeck.  I get 0 xruns on 
athlon1 at -p1024, while browsing, chatting, emailing, etc.  Shooting 
for similar performance from tapedeck.

Question:  I ran top on both machines while recording, and the CPU usage 
is always 50-95%+ for ardour (the value gets split pretty evenly with X 
until ardour is minimized, then ardour takes it all), but qjackctl and 
ardour both show a common, very low value, always under 10%.  Is there 
any rhyme or reason here?

Paul Winkler wrote:
> Hi Doug,
> 
> This message seems more suited for ardour-users list, but here's
> my impressions...
> 
> On Sat, Mar 13, 2004 at 07:41:29AM -0500, Doug wrote:
> 
>>/usr/local/bin/jackstart -R -t500 -dalsa -dhw:1 -r48000 -p1024 -n2 -m -H
> 
> 
> 
>>Mostly I tried to get a handle on whats happening with the I/O 
>>performance on tapedeck.  I reversed roles and recorded into athlon1. 
>>Tapedeck was spitting and sputtering and snap crackle pop, burping and 
>>farting its way thru the same cubase project, while ardour, running on 
>>athlon1, proudly and flawlessly captured every embarrasing moment. 
>>Everything afterwards is pretty much inconclusive i think, till i 
>>determine whats up with that CPU thing.  I was vey happy with ardour's 
>>performance on athlon1 though.  Worst case I'll just go Nforce2 on all 
>>my machines.  I'm sure that wont be necessary though.
> 
> 
> Odd. You don't say what kernel version(s) you're using or
> how you set up your audio workstation (e.g. planet CCRMA 
> or Agnula or did everything "from scratch" or what)...
> This certainly sounds like hardware that is either bad or
> something is misconfigured.
> Have you been through EVERYTHING in the usual checklists?
> this is a good one:
> http://myweb.cableone.net/eviltwin69/ALSA_JACK_ARDOUR.html
>  

-- 
http://nostar.isa-geek.com/