<div> </div>
<div><span class="gmail_quote">On 2/19/08, <b class="gmail_sendername">Scott D. Davilla</b> <<a href="mailto:davilla@4pi.com">davilla@4pi.com</a>> wrote:</span>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">>Figuring out the hang and video corruption issue during xvmc decode<br>>are the biggest issues right now. Once I solve those, all the other<br>
>issues are trivial. Testing the various changes takes a lot of time<br>>as the two issues are random and might take several hours to occur.<br>>For example, I made a change last night and run 4 hours straight<br>
>before I got a hang. Before going to bed, I made another change (send<br>>audio to /dev/null) and it was hung again in the morning but a<br>>difference type of hang, I could escape back to the main menu. There<br>
>are many variables in play.<br>><br><br>Just a quick report, I switched to mplayer to simplify debugging the<br>xvmc hang/video corruption and I'm happy to say that after much<br>investigation, I think I might have it solved.<br>
<br>So far, mplayer has been running continuously for over 12 hours doing<br>xvmc 1080i mpeg2 decode from a HDHomeRun. Not a sign of any glitches.<br><br>The trick seems to be underclocking the nvidia gpu. The "standard"<br>
clocks reported by nvidia-settings is gpu = 360Mhz and vram = 720MHz.<br>I'm running at gpu at 200MHz and vram at 800MHz. Interestingly, the<br>vram speed effects xvmc decode performance more than gpu speed.<br><br>I have to backout of some kernel config changes and patches and<br>
retest under myth to be sure but I've never seen 1080i xvmc run this<br>long without glitching.</blockquote>
<div> </div>
<div> </div>
<div>Scott,</div>
<div> </div>
<div>This is great news. A low cost, small, low-noise HD capable frontend is like the holy grail. Out of curiousity, how extensive are the changes you made to a base kernel in order to get to this point? Meaning can someone with reasonable linux experience be able to get a box to the point you have?</div>
<div> </div>
<div>thanks,</div>
<div>matt</div>
<div> </div>
<div> </div><br> </div>