[Asterisk-Users] quad fxo

[Michael Sandee]

Steven Critchfield wrote:

>On Mon, 2004-05-03 at 12:39, Michael Sandee wrote:
>  
>
>>The bus isn't wrong... debian is wrong. Like everything in debian... it 
>>ships with an old pci.ids
>>(No flames intended... but still :P )
>>    
>>
>
>If that was all, then it would have still showed up in the PCI bus just
>like you mention below. 
>
It did as "generic communications device".

>It also means he is probably running the "STABLE" version. Debian named
>it stable only because they don't change it very often. If you want
>something as new as the other distros, you have to go with "Testing", or
>"Unstable", possibly even "Experimental". Just be glad they properly
>mark their releases unlike others whose x.0 releases shouldn't ever be
>trusted in a production environment.
>
>(Not flaming either, just filling in the picture) 
>
Well... tell me what is unstable about putting a new pci chip 
identification database into your distro?
I run debian stable on my workstation... Some things are ok to be 
stable... but things like this are...well not so nice.

>>replace yours with one from:
>>http://pciids.sourceforge.net/
>>
>>And It *should* report it better... (Didn't verify)
>>    
>>
>Same should work with an upgrade in debian.
>
I seriously advice against that... vividly remebering the NPTL debacle 
in unstable... and loads of other glibc problems you can read about in 
the bugtracker.

>>Not that any of this matters... Just load the driver and get on with it.
>>    
>>
>
>While this may be true, the kernel should probably be recompiled for
>best performance on the new hardware.
>
Agreed. But that wasn't my point ;)

While we are at the subject anyway I can put some on-topic info here 
aswell. Recent benchmarks with 6 QuadBRI's on a Pentium-4 2.8Ghz 
resulted in a almost 100% improvement in load under Linux 2.6(.4) over 
Linux 2.4(.25). This ofcourse resulted in less (no) quirks in the sound 
under 2.6 than under 2.4.
The loadtest was done looping the BRI's to each other and in such way 
that we used all 24 BRI's (2 channels) resulting in 48 active channels.

http://voidptr.astmaster.org/loadtest1.jpg
http://voidptr.astmaster.org/loadtest2.jpg

The audio quality was easily monitored in contrary to often proposed 
"test suites"... Two phones at each end...

Regards