ATSC PCI cards: Difference between revisions
No edit summary |
(Further report on performance with hardware xvmc acceleration) |
||
Line 12: | Line 12: | ||
My experience with capturing using the pcHDTV is that it uses 1.3% of CPU (azap + cat + cx88[0] dvb on an athlon64 3500). Actually watching live or captured streams does take a lot of CPU, though, especially without xvmc. --[[User:Mitchskin|Mitch]] 21:50, 14 May 2005 (CEST) |
My experience with capturing using the pcHDTV is that it uses 1.3% of CPU (azap + cat + cx88[0] dvb on an athlon64 3500). Actually watching live or captured streams does take a lot of CPU, though, especially without xvmc. --[[User:Mitchskin|Mitch]] 21:50, 14 May 2005 (CEST) |
||
Using a HD-3000 on an Athlon XP 1800+ and GeForce 6600GT: watching live 720p using mplayer -vo xvmc takes 45-60% cpu while 1080i causes frame drops and pretty much pegs the cpu at 100%. Capturing either takes very little CPU as posted by Mitch above. |
Using a HD-3000 on an Athlon XP 1800+ and GeForce 6600GT: watching live 720p using mplayer -vo xvmc takes 45-60% cpu while 1080i causes frame drops and pretty much pegs the cpu at 100%. Capturing either takes very little CPU as posted by Mitch above. |
||
Another data point: using 866MHz Pentium III with GeForce FX 5200, recorded 720p is smooth with mplayer -vo xvmc but pegs CPU. Live 720p is watchable if jerky with -hardframedrop. 1080i is another story; it takes 10 seconds to play 9 seconds' worth of video. |
|||
The ATSC frontend of the DViCO cards has been tested with 8-VSB (OTA) and QAM-256 (Cable) in the US. Source code is in video4linux + dvb-kernel CVS and kernel sources 2.6.13 and later. |
The ATSC frontend of the DViCO cards has been tested with 8-VSB (OTA) and QAM-256 (Cable) in the US. Source code is in video4linux + dvb-kernel CVS and kernel sources 2.6.13 and later. |
Revision as of 21:15, 18 October 2005
As of right now there are 6 confirmed working cards.
- The PCHDTV card
- The air2pc card
- The DViCO FusionHDTV 3 GOLD-Q
- The DViCO FusionHDTV 3 GOLD-T
- The DViCO FusionHDTV 5 GOLD
- The DViCO FusionHDTV 5 LITE
The air2pc seems to take some work off the processor while I have read the PCHDTV uses 80-90% on an athalon64 3000.
My experience with capturing using the pcHDTV is that it uses 1.3% of CPU (azap + cat + cx88[0] dvb on an athlon64 3500). Actually watching live or captured streams does take a lot of CPU, though, especially without xvmc. --Mitch 21:50, 14 May 2005 (CEST)
Using a HD-3000 on an Athlon XP 1800+ and GeForce 6600GT: watching live 720p using mplayer -vo xvmc takes 45-60% cpu while 1080i causes frame drops and pretty much pegs the cpu at 100%. Capturing either takes very little CPU as posted by Mitch above.
Another data point: using 866MHz Pentium III with GeForce FX 5200, recorded 720p is smooth with mplayer -vo xvmc but pegs CPU. Live 720p is watchable if jerky with -hardframedrop. 1080i is another story; it takes 10 seconds to play 9 seconds' worth of video.
The ATSC frontend of the DViCO cards has been tested with 8-VSB (OTA) and QAM-256 (Cable) in the US. Source code is in video4linux + dvb-kernel CVS and kernel sources 2.6.13 and later.
Here is a feature matrix to help keep track of what card does what:
pcHDTV HD-3000 | Air2PC HD5000 | HDTV3 Gold-Q | HDTV3 Gold-T | HDTV5 Lite | HDTV5 Gold | |
NTSC | yes | no | yes | yes | yes | yes |
Comp/S-video | ? | no | ? | yes | yes[1] | yes |
Analog CC | ? | no | ? | ? | yes | no[2] |
PID filtering[3] | ? | hw | sw | sw | sw | sw |
- [1] Appears to have a working comp/s-video port hidden behind the card-plate.
- [2] Hardware should allow it, but the cx88 driver has no support yet.
- [3] 'hw' = hardware, 'sw' = software. Hardware PID filtering allows the card to discard unwanted packets, which typically amounts to half the 19.3 Mbps bitrate of an ATSC broadcast.