Advanced Video Playback

The third element to the Winter Update of Avivo is concerning that of video playback quality, possibly one of the most important elements of Avivo currently, at least where standard definition content is concerned. Due to the nature of the issues surrounding R520's development and release a concerted effort on the software side had to be put off until quite late prior to its release, which left ATI prioritising what needed to go in immediately and what had to wait - obviously for R520's release the 3D performance and compatibility was the primary issue for software support, so the video elements had to be left out for the interim. However, with Radeon X1300 and X1600 also being targeted towards OEM desktop, media and notebook solutions the video quality and performance is highly important, and this was addressed with the Catalyst 5.13 release and onwards.

For the purposes of testing the video quality we'll take a look at the HQV DVD benchmark. The benchmark is a DVD with a series of clips that covers many facets of decoding that need to be done given the different natures of source material available, with a booklet that describes each test, what should be seen and what score should be given depending on the displayed output - a short description of the benchmark is available here, along with a link to the scoring guide itself. The benchmark itself was designed to test the quality of consumer DVD players as well as displays, with its top scoring representing the most advanced processing available on high end consumer players, however it can equally be applied PC DVD playback as well.

The Avivo video decoding technologies are the same across the entire X1000 line of boards, so we'll take a look at the quality via the X1600 XT, using drivers pre and post the Catalyst 5.13 Avivo Winter update, with the initial X1000 beta 5.9 drivers and the more recent 6.2 drivers.

Update 26-Feb-2006: ATI have made further video quality updates with the Catalyst 6.4 drivers, which should be available for download in April 2006. We've added the scores from the 6.4 beta driver into the HQV scoring table below.

If we take a look at the scoring alone, its evident that the X1600 (hence all the X1000 series) have improved considerably according to this test, going from (in our judgement) 35 to 103 - bearing in mind that that top score is 130 for the benchmark, there's only a few places left to improve. If we look at the results a little closer though we can basically break the tests down to de-interlacing tests, quality enhancements & noise reduction and cadence testing.

Even though we are moving into an HD era with more digital displays, and much more HD content (broadcast or physical), interlaced content will still remain important, hence being able to effectively decode it, and display it without artefacts is important for good quality output. In this area we see that, according the the HQV test, ATI have made significant strides since the initial driver release, with the video processing engine being much better at detecting what needs to be done and applying it on the given scenarios. The 6.4 driver also adds vector adaptive de-interlacing to High Definition MPEG-2 playback.

In terms of detail enhancement and noise reduction, or removing image graining, ATI appear not to have made much in the way of changes here, with apparently no image enhancement capabilities in place, and no apparent noise reduction changes, up to the Catalyst 6.2 drivers. Two of the main areas that ATI have stated they have addressed for Catalyst 6.4 and onwards is with Edge Enhancement (which would show up on the detail test on HQV) and De-Noise filters. With the Catalyst 6.4 drivers the noise filters have evidently been improved, however we couldn't discern any Edge Enhancement in the test.

The main group of tests are looking at the detection and support of different cadences. Cadence is basically the method that needs to be applied for converting different types of recordings, not natively translatable to consumer playback / display devices, to such devices. Probably the most popular instance of this is the the fact that movies are filmed at 24 FPS, but an NTSC video will be playing at 60 FPS - in order for things to be effectively encoded upwards from 24 FPS to 60 FP extra frames need to be inserted, and in this instance a single extra frame is repeated for every 4 frames. During playback the decoder can analyse the frames and determine the patterns of the fields to see how they originated  and should best be played back  - in the case of film content this is known as 3:2 pulldown.

It's with the various types of cadence detection that ATI have clearly made the biggest leap in regards to this test, with the X1000 boards basically detecting any but the most basic cadence types prior to Catalyst 5.13, but all types used in the HQV benchmark being detected and correctly supported post 5.13.

* Note: In later Catalyst drivers there is a "3:2 pulldown" option under the de-interlacing section of the video controls, enable this to get the best quality out of film playback as this appears to enable the cadence detection.