[mythtv-users] SOLVED System Load/Performance Question

[Brian Wood]

On Feb 24, 2006, at 5:20 PM, Bob Cottingham wrote:

> On 2/24/06, Brian Wood <beww at beww.org> wrote:
>> On Feb 24, 2006, at 3:00 PM, mrwester wrote:
>>> This sounded like it could be related to problems I've been seeing
>>> with ivtv corruption and prebuffering pauses, however with my  
>>> nforce2
>>> chipset I can't adjust the pci latency for the IDE, audio or a few
>>> other controllers, which are all set to 0.  I don't have a problem
>>> changing the latency for the build-in video (which was set to  
>>> 248) and
>>> the ivtv cards.  I found this to be the case for both nforce2 based
>>> systems I tried.  I didn't find anything through google that really
>>> addressed this issue.  Does anyone else see this?  Is this a  
>>> potential
>>> issue?  Is there any reason to have the video card latency so  
>>> high?  I
>>> lowered the video latency and I'm going to see what effect it has on
>>> performance.
>>>
>>> Hi Bob-
>>>
>>> Did you come to any  resolution with this?  I've got an nforce2
>>> system as well (chaintech) , and I cannot adjust pci latency for
>>> IDE either...  I see a systematic stuttering (1 every 1.5 seconds
>>> or so?) of CNN tickertape with my XP2900, FX5200 system, that I
>>> can't help think it has something to do with PCI latency.  I get
>>> occasional glitches in HDTV recordings as well, for example if
>>> someone accesses mythweb while the recording is happening.  I think
>>> this is all related, but like you, I'm not finding any answers by
>>> google.  Thanks for any insight,
>>>
>>
>> Every system I have checked (albeit only 4 of them) has had the
>> graphics card set at 248. I've seen a couple of boxes (like my Dell
>> P4) that had IDE set to 0, but I'm not running Myth on that one so I
>> haven't played with it yet.
>>
>> The "0" value may be something special in the PCI/DMA setup, and may
>> not mean what it seems to mean.I'm trying to learn more as well.
>>
>> You say you can't change the latency for the IDE system, what happens
>> when you run setpci? Does it seem to work and just not set the value?
>
> I tried setpci -v -d "*:* latency_timer=b0 as discussed in
> http://www-128.ibm.com/developerworks/library/l-hw2.html, and it
> responded as though everything had been set to b0, however when I
> reran lspci -v the only thing that had changed was the encoder cards
> and the graphics card.  Everything else was still 0.  I think this may
> mean that they aren't writeable and I don't know if the relatively
> large values of 64 for each of my two encoder cards is significant or
> if the 0 value simply means some special setting as you mentioned and
> therefore isn't can't be directly compared.  I'm wonder because it
> would apply to soundcard, LAN port and IDE channel.


I'd be interested to know what host bridge chipset you have, I'm  
trying to accumulate data on as many as I can.

My Dell with an Intel Tehama chipset (82850 850) behaves the same as  
yours, everything seems defaulted to 64 except for the graphics card  
(248) and the IDE interfaces and the bridges, which are zero. I think  
it has something to do with the way the bridges interact with the PCI  
system.

Clearly we need more information on this. Like so many of the "fixes"  
we find, they work for some folks and not others. I think there are  
more variables than we are currently aware of that affect this matter.