[mythtv-users] Latency timings...

Brian Wood beww at beww.org
Fri Apr 14 23:08:39 UTC 2006


On Apr 14, 2006, at 2:47 PM, Steven Adeff wrote:

> Couple questions....
>
> 1)
> Is there a reason setpci wouldn't work? I run:
> setpci -v -s 00:0a.0 latency_timer=80
> or
> setpci -v -s 00:09.0 latency_timer=80
> for my two SATA devices  and it shows:
> 00:09.0:0d 80
> but when I check with lspci -v they're both still set to zero?

Some chipsets seem to report a latency of zero no matter what you do.  
In those cases you can adjust the values for things capture cards and  
video cards but can't adjust the disk systems. Playing with latency  
values is something that sometimes helps some people but is not a  
panacea.


>
> 2)
> The wiki page doesn't give much help in the advice area as to latency
> values or how to "calculate" them for Myth use (ie DVB cards, etc),
> except to say that your harddrives should have a higher value than
> your PVR card. Anyone have any advice in this area?
>

It's pretty much a trial and error thing at this point. The mobo  
manufacturers don't seem to give out much info, they just assume that  
you will be running Windows and be happy.

The "optimum" values are so subject to so many variables that any  
attempt to give advice is doomed failure, you just want to understand  
the general principles well enough to make somewhat intelligent  
adjustments, but the exact correct value will depend on what type and  
how many video streams are being delivered, what sort of storage  
system they are being delivered to, the buffer capacity of everything  
involved in the system, what other devices are competing for time on  
the PCI bus, what interrupts might be shared and many more that I  
have not even thought of I'm sure.

The purpose of the WiKi entry was basically to let people know what  
the parameter is, the fact that it can affect performance of a Myth  
system and what that error you sometimes get from the IVTV drivers in  
your boot log might mean.


More information about the mythtv-users mailing list