Video Quality Tested: GeForce Vs. Radeon In HQV 2.0

Test Class 1: Video Conversion, Cont’d.

Chapter 2: Film Resolution Tests

The tests in Chapters 1, 4, and 5 focus on cadence, and whether the video processor can perform appropriate pull-down detection. Cadences are used to stretch or compress video in different formats to fit the 60 interlaced fields/second NTSC standard. The concept may not be intuitive, but if a video processor cannot properly detect cadence, the resulting video may suffer from interpolation artifacts, such as moiré patterns. The benchmark includes indicator boxes to help the reviewer assess whether the cadence has been detected. Scoring is based on the video processor’s ability to lock, and how quickly it locks onto the correct cadence. Five points are awarded for a lock within one half of a second, three points are awarded if the lock is achieved under a second, and if the lock is not achieved or is intermittent throughout the clip, the score is zero.

This is an objective test. Our observations reveal that all of the tested Radeon cards are able to detect 2:2 and 3:2 cadences. All of the Nvidia cards detect the 3:2 cadence, but the GeForce 210, GT 430, and GeForce GTX 460 do not detect the 2:2 cadence in our tests.

Arguably, the 2:2 cadence is somewhat less important than 3:2. Most films are recorded at 24 FPS and this is converted with the 3:2 pulldown cadence, while the 2:2 cadence is only used in countries following the PAL and SECAM standards that shoot film destined for television at 25 FPS. As such, I question the wisdom of assigning these two cadences the same point value. The 3:2 cadence should be worth more points.

[EDIT: I'm going to clarify this a little. I'm not saying that PAL countries should be ignored, I'm saying 24 FPS film is probably more important in the big scheme of things. We're well aware that these graphics cards are to be used internationally.]

Chapter 3: Overlay on Film Tests

When text is overlaid on film, it’s possible that it may have a different cadence from the underlying video. If the video processor doesn’t detect the different cadences, the text or background may display shredding artifacts. If no shredding artifacts are seen, a perfect score of five is earned. If artifacts are seen at first but are quickly corrected, the score is lowered to three, and zero points are awarded if shredding is observed throughout the majority of the sequence. There are two tests in this chapter and each is scored separately: once for horizontal moving text and once for vertical moving text.

We found this one difficult to judge. Certainly, the text looks good across all graphics cards, but it doesn’t seem quite perfect, either, especially when it comes to the vertical moving text. Based on the good overall performance, we’re giving all graphics cards a full five points for both tests.

Chapter 4: Response Time Tests

This test assesses the speed at which the two most prolific cadences are detected and locked in. A circling globe helps assess how quickly cadence is locked by moving by a gauge. A full five points are awarded if the cadence is locked under one half of a second, and this is lowered to two points if it takes between one half and one second to achieve lock. If it takes more than one second to lock cadence, no points are awarded, and, of course, no points are given if the graphics card does not detect the cadence at all.

This is an objective test but it can be difficult to judge the time it takes to lock cadence because of the speed of the moving indicator ball. It appears to us that all of the graphics cards we’re testing can achieve a lock under one half of a second.

In the case of some GeForce cards, the 2:2 cadence is not detected and points are lost here. It’s interesting that response time is already a factor in the Chapter 2 cadence tests, and this test may be included as a way to assign more points to the notable 2:2 and 3:2 film cadences relative to the less-common ratios we look at next.

Chapter 5: Multi-Cadence Tests

These tests are identical to the tests in Chapter 2, except that the cadences are different.

Once again, the GeForce 210, GT 430, and GTX 460 are left behind, while the rest of the GeForce family and all the Radeon cards we test achieve full marks.

It’s arguable that these cadences should be assigned far fewer points than 3:2 and 2:2 film cadences. There are some relatively obscure ratios here that many viewers will never see. So, in our opinion, the weighting of these tests on the final score (30 points out of 210) may unrealistically indicate all-around video performance.

Chapter 6: Color Upsampling Error Tests

The tests in this chapter expose how the video processor handles downsampled video. If it cannot up-scale the video properly, color artifacts may result. The ICP test focuses on interlaced video while the CUE test examines progressive video. The full score of five is assigned if no color artifacts can be observed. This is reduced to two points if faint color halos or mild stair stepping is seen. A score of zero is the result of highly-observable color artifacts.

We find that this test can be difficult to score when extremely subtle artifacts are seen. We’ll err on the side of generosity and give five points to the graphics cards that appear to produce very good results.

While we hand out the full five marks to all of the test cards for CUE progressive video output tests, all of the GeForce cards appear to produce some artifacts in the ICP interlaced video test, except for the GeForce GTX 470, a card that earns full marks.