asterisk users - Jan 2006 - How does the PCI bus effect latency and echo?

Hello All,
I was wondering, how does the PCI bus effect echo and latency?

I know a large part of the echo issues out there have to do with the pci
bus latency, but are there any suggestions on the hardware side to
minimize this?  For example, a 533mhz vs 800mhz fsb, would it have any
effect on communications?  Even on a low load server, I am trying to
decide if getting a dell poweredge server is worth the effect, versus
using an intel board or asus board.

Regards,
Greg

> I was wondering, how does the PCI bus effect echo and latency?
> 
> I know a large part of the echo issues out there have to do with the pci
> bus latency, but are there any suggestions on the hardware side to
> minimize this?  For example, a 533mhz vs 800mhz fsb, would it have any
> effect on communications?  Even on a low load server, I am trying to
> decide if getting a dell poweredge server is worth the effect, versus
> using an intel board or asus board.
You should be able to find hundreds of postings relative to this over
the last two/three years.

The issues with pci bus seem to mostly relate to the pci North and South
bridge chips, and deficiencies that have been identified _after_ those
chips were manufacturered. Those lists involved with audio/music have
spent far more time analyzing the issues, and seem to have a much better
handle on quantifying the problems compared to the * list. The physical
bus (and its speed) is not the issue.

The pci components involve a chip set on the motherboard and an associated 
chip set on the pci card. Mucking with one piece of that relationship has 
not been proven to have any effect on the problem. Therefore, it would 
appear the issue with digium analog cards is more oriented towards the 
TigerJet chip, but even that has not been proven.

One of the more recent true tests (that I've heard about) are those 
individuals that have purchased x100p compatible winmodem cards (that
use a different chipset) and have successfully made the spandsp fax
software work reliably. In other words, if one can get spandsp to function
with an analog card, the card improves other audio issues as well. That
seems to futher support the statement that the TigerJet pci chip is at
least part of the problem.

Kevin was going to obtain a logic analyzer several months ago to identify
the root issues with the digium analog cards, but apparently that effort 
has been side-tracked or back-burnered as well.

I've played around with several motherboards attempting to find something
that might impact the issue. Haven't found one yet that actually improves
the latency issues with the digium analog cards. Dell is not the answer
for sure.