As we reported earlier, Intel showed off their integrated G965 graphics chip in a side-by-side demo with another computer that was powered by an AMD/ATI X1600 discrete card. Both computers were running Windows Vista and had 2 GB of RAM. Nick Knupffer, PR Manager for Intel, told TG Daily that both computers were identical except for the graphics cards and obviously the drivers.
The test in question is the flapping American flag video from the HQV benchmark DVD from Silicon Optix. The DVD contains several other tests which are used to gauge how well television sets and graphics cards render video. In the flag video, good graphics cards can render the waving flag with little or no interlacing artifacts, more commonly called "jaggies" or "stair casing". In Intel's comparison, the AMD-powered computer appeared to render out huge 10 by 10 pixel jaggies during the full-screen flag video playback under the Cyberlink player. The video also had a much lower framerate and appeared to stutter at several points.
The Intel G965 computer rendered the video smoothly and we didn't notice any jaggies in between the stripes of the flag. Knupffer told TG Daily that their engineers did everything to make it a fair test between the G965 and AMD's X1600. In an email he writes, "Both machines were identically configured. Both Core 2 Duo's, both had the same memory, 965 chipsets, the same Windows Vista Ultimate build. Both had the latest drivers from the web, Catalyst 7.2 and Intel 15.2." Knupffer adds that his staff has contacted AMD engineers through back-channels to figure out if anything was incorrectly configured.
Since the video results appeared very lopsided, in Intel's favor, we wanted to verify that both machines were configured optimally. Nick Knupffer, PR Manager for Intel, did let us take a quick peek at the driver settings and the video files, but we found nothing wrong. The video files on both machines were the same size and resolution. We did not physically open any of the computers to verify the components.
We played back the video files on both machines with regular Windows Media Player because we wanted to test if the supposed lopsided performance was solely related to the Cyberlink playback software. With Windows Media Player, both computers appeared to play back the video at the same speed and with significant jaggies. Knupffer explained that this was because of Cyberlink's integrated with the G965's ClearVideo acceleration and smoothing routines.
You would expect that a discrete graphics card would have beaten the pants off of the integrated solutions, especially in video playback routines. There's no denying that Intel's upcoming driver for the G965 will make tremendous improvements, but we were puzzled by how bad the AMD fare. In fact, our personal experiences have shown that the X1600 card is very capable of video playback, when configured correctly. The graphics guys at Tom's Hardware Guide are now testing out their X1600 cards and will try to duplicate Intel's results.
But we aren't the only ones who were puzzled at the side-by-side comparison because Intel showed the same demo to other journalists, journalists who have told us that they will also run their own tests. Scott Wasson, editor in chief of The Tech Report, emailed us saying that his lab guys are testing out the card and will publish the results in a future article. Another editor, Charlie Demerjian from the Inquirer, said that the demo looked strange and told us that they may do their own testing. As a side note, Charlie told us that he offered to bring an AMD to the demo to help Intel correctly configure the computers, but Intel, as you can imagine, declined.
While Intel representatives insist that everything was configured correctly, AMD officials say the opposite. Will Willis, AMD's Senior Public Relations Manager, told TG Daily, "The X1600 is unquestionably a much better card for video playback." He added that it's easy to cherry-pick tests that make one product look worse than the other. "We want to let the world know that this one test is not the whole story," said Willis.
To prove his point, AMD cranked up their internal testing labs and didn't use just the flag demo, but all the tests in the HQV benchmark suite. Unlike the Intel demo though, they played back video from the built-in player/DVD decoder of the 32-bit version of Windows Vista instead of the Cyberlink player. According to Willis, the G965 was thoroughly trounced, in HQV benchmark scores, by the X1600. He also stresses that the scores come from drivers that are currently available as compared to Intel's upcoming driver.
"We ran G965 with Intel's 15.2 driver release (the current driver that's publicly available), and ran it against an ATI Radeon X1600 using Catalyst 7.2 (our current driver), and an AMD 690. Make sure that "Pull-down detection" in Catalyst Control Center is enabled. The final scores look like this: AMD/ATI Radeon X1600=100, Intel G965=40," Willis wrote in an email.
Willis told us that there could be several reasons why the X1600-powerd computer performed poorly during Intel's test. He says that basic deinterlacing could have been set in the Catalyst control panel or that hardware acceleration could have been turned off. The consensus among the journalists that we talked to and our internal graphics staff is that the driver could have been misconfigured or Avivo wasn't installed properly. Of course the problem could be as simple as a malfunctioning AMD card.
So is Intel correct or should the AMD-powered computer have performed much better? Intel's performance really isn't at issue here. Their reps showed off an upcoming driver that looked quite decent for video playback and some light gaming and you would expect a company to show their own products in the best possible light. But why did the AMD card do so poorly, at this point it's anyone's guess. You've got the labs from both sides saying that their respective chips do better so it will probably take an independent lab to fully decide the issue.