PDA

View Full Version : Latest XviD DShow Decoder (v-1.0RC3) Higher CPU Utilization than DivX 5.1.1?


therealjoeblow
9th March 2004, 18:52
Has anyone else noticed that the latest XviD decoder (1.0RC3) uses (significantly?) more CPU horsepower than the DivX 5.1.1 decoder? On my Celeron 1.3A, Asus TUV4X mb, what I notice is that with XviD as the decoder, I have to turn off *all* of the postprocessing settings, and then 640x272 widescreen encodes will generally play back with CPU utilization ranging from 70~90%. If I turn on the either of deblocking or deringing features (independently of eachother), CPU Utilization goes to 100%, however decoding continues mostly OK, but the whole system responds slugishly since the decoder is sucking up all of the processor cycles trying it's best to not drop frames. If I turn them both on, then playback drops many frames as CPU is pinned to 100% and can't keep up. With 512x384 (4:3) encodes, even with all of the postprocessing settings turned off, I get the 100% pinned utilization with dropped frames, so I can't really use the XviD decoder at all on these.

With the DivX 5.1.1 decoder, I can set post processing settings to partial control (which enables deblocking I believe) for 640x272 encodes, and have processor utilization only at 50~70%; turn on all postprocessing full and have the 100% utilization, but where everything still plays back without dropping frames as described above. The 512x384 encodes also play back fine with partial postprocessing enabled, which I can't do with the XviD decoder at all.

I've done these encodes with AutoGK 0.9x~1.0 using the bundled alpha release of the XviD encoder, so there should be no exclusive, incompatible encoder settings that cause the difference.


Others have similar experience or comments?

Koepi
9th March 2004, 19:14
It's some issue with your vga card.

Did it ever occur to you that you could try to turn on that "compatibility renderer" option while you were playing around with the PP settings already? (Close the palyer afterwards and restart it with the file you want to play to make it work, you can't switch renderers on the fly.)

It might be the solution to your vga hardware disabilities.

Koepi

therealjoeblow
9th March 2004, 20:56
Originally posted by Koepi
It's some issue with your vga card.

Did it ever occur to you that you could try to turn on that "compatibility renderer" option while you were playing around with the PP settings already? (Close the palyer afterwards and restart it with the file you want to play to make it work, you can't switch renderers on the fly.)

It might be the solution to your vga hardware disabilities.

Koepi

Thanks, Koepi, I'll give that a try tonight at home, but I doubt that it will be the source of the problem unless relatively current ATI Radeons 's all have some generic bug - the machine described above has an ATI AIW Radeon 9000pro (retail ati card) using Catalyst 3.9 drivers, and I also have another machine with a Celeron 633 on a Soyo (Intel BX) motherboard with an ATI Radeon 9200 (retail ati) using catalyst 4.1 drivers, it too has similar results (ie that the divx decoder appears to be more efficient), except that it cannot play with the XviD decoder with any of the PP settings enabled because the processor is much less capable.

therealjoeblow
11th March 2004, 23:05
Well, I tried compatibility mode, no difference. On both of my machines the XviD decoder has about 10~20% higher CPU utilization than the DivX 5.1.1 decoder (which is still higher than ffdshow but I gave up on that until they stabilize it on the newer xvid features).