[ATrpms-users] nVIDIA drivers
DM - ATRPMS
atrpms at tazman.net
Thu Aug 10 04:20:46 CEST 2006
I started the system and tried to simulate Myth like usage without using
myth by playing back a DVD (using mplayer) and transferring huge files
over the network (using SCP). Ran fine for a little over 30 minutes and
3GB transferred, so I stopped it and tried again with Myth. The box
froze again after less than 5 minutes.
Notably, mplayer would not work using -vo xvmc (which I think would most
closely simulate Myth playback), but did generate similar CPU overhead
to Myth when run without any -vo option. Overall CPU was higher during
the simulation (due to the SCP), and network throughput also appeared to
be higher than a normal Myth playback, so it does not appear to be load
related.
All this got me thinking and I couldn't remember if I had run through
the MPEG2 decoder settings or if I had just concentrated on getting XvMC
working (since everyone seems to think it's the way to go). Switching
to Standard/ffmpeg or libmpeg2 seems to take care of the lockup problem,
but at a substantial processor penalty (from roughly 40% util to 85% and
load jumping from about 1 to 3 or more). Playback isn't bad
considering, but still stutters too much to bring satisfaction.
I found the nvidia-settings utility and temperature never went above 37C
during playback. When I tried to go to the "OpenGL/GLX Information"
screen during playback, it locked up however (only during playback,
non-playback worked fine). In fact, starting nvidia-settings at anytime
other than during playback worked, but locked during playback.
Playing with some more Myth settings, it appeared that unchecking the
"Enable OpenGL vertical sync for timing" option takes care of the lockup
problem and allows the use of XvMC. It's been going about 20 minutes
and appears that it does not adversely affect playback or CPU usage.
(I'm watching a baseball game tho, so I can't tell if video/audio sync
is right tho... IIRC that's why I turned it on to begin with.)
Therefore, it appears that it has something to do with some combination
of the following:
- XvMC interface/support with Myth
- OpenGL
- nVIDIA 6200 chipset
- PCI interface bus
At least for the time being, I'm satisfied with this results, but am
willing to continue to troubleshoot if someone wants to point me in the
right direction. (I don't know how much developers of the respective
software / libraries / drivers monitor this list tho, and I suspect that
would be where it needs to go next.)
Thanks again!
Dan
DM - ATRPMS wrote:
> Keith C wrote:
>> Any way you can check the card in a Windows box? Any overheating issues
>> or seating (with the AGP slot) issues?
>
> Not easily... I'd have to do a load.
>
> I doubt it's overheating... the heat sink doesn't get that hot and it is
> fan cooled. I did pop the heat sink off to see what kind of thermal
> compound was underneath, and it was using grease (not the adhesive pads)
> so it should be fine. I put a little more in just for good measure tho.
> CPU, bridge, everything else is relatively cool to the touch.
>
> I also doubt it's seating. It's been in and out of the machine probably
> a dozen times as part of troubleshooting. (And it's PCI, not AGP, so
> should have fewer seating issues.)
>> This really sounds like a bad card. I'm running a 6200 series card and
>> it works fantastically.
>>
> That's what I'm fearing... unfortunately it's mail order, so it's not
> that easy to swap out. I just realized that I haven't really stressed
> it that much other than through Myth, so maybe I'll throw it back in and
> start glxgears on it for a while tonight. (Any other suggestions for
> working a bus and video card?)
>
> Dan
>
>
> _______________________________________________
> atrpms-users mailing list
> atrpms-users at atrpms.net
> http://lists.atrpms.net/mailman/listinfo/atrpms-users
More information about the atrpms-users
mailing list