Packet loss and stuttering audio

For discussion of the WDM and ASIO Axia Drivers

Moderators: AXIA_milos, Maciej

Post Reply
tzeiger
Posts: 1
Joined: Mon Nov 15, 2010 11:26 pm
Location: Berlin

Packet loss and stuttering audio

Post by tzeiger »

Hello everyone.

For quite some time we are experiencing packet loss on two of our machines which are used to encode various webradio channels. This results in frequent stutter of the audio.

On the same livewire-network there is an almost identical machine that does not show this behaviour. The only difference to the other machines is the NIC used for the internet access.
All NICs used for Livewire are the same.

With driver version 2.5.2.3 the statistics look like this (version 2.4.8.12 had no RxOr but was the same otherwise):

Image

In the screenshot channel 1&2 are standard stereo while channel 3&4 are live stereo.

Did anybody experience something like this and knows a solution of any kind?

Specs:

HP Compaq dc7800 Convertible Minitower-PC
Windows XP Pro SP3 full patched
Axia IP Audio Driver 2.5.2.3

NIC1: Intel 82566DM-2 (driver 9.13.41.0) used for livewire
NIC2: Intel PRO/1000 GT 82541 (driver 8.10.3.0) used for internet access


We ran DPCLAT on the machine but it is in the yellow around ~1000 so this should be "ok" at least. The bar never goes to the red.

Anybody has ideas, thoughts or advice? Please share! :roll:

Thanks alot in advance, this has really been nagging on us for quite some time now.

Tim

Andrew_KOOP
Posts: 47
Joined: Thu Aug 23, 2007 7:25 pm
Location: Austin, TX
Contact:

Post by Andrew_KOOP »

Hello,

Does your Intel 82566DM-2 card have any "TCP offload engine" features? If so, try disabling them. You should see them listed when you pull up the network card properties in Windows.

We have two HP servers whose internal NICs have very robust TCP OE features. Those are great for normal traffic but they sometimes the extra intellegence leads to unusual behavior. We disabled these features and the cards performed fine.

We meant to do a design of experiment to figure out which features caused the confusion but clearly this hasn't made it off the backburner of things we'd like to try.

Post Reply