Microsoft Windows 7 Test Drive, Part 2
Last issue, I wrote about my experience installing Microsoft Windows 7 on a Windows XP 64-bit system. In this issue, I'll compare the rendering performance of Windows 7 with that of Windows Vista, and ultimately Windows XP. I say ultimately because I wasn't able to test my Windows XP system until past the deadline for shipping this article. So we shipped on time with Windows 7 versus Vista, and we'll add the XP component in a day or two.
I was fortunate to have an extraordinarily convenient test bed for comparing the operating systems. Specifically, custom computer manufacturer Puget Systems provided a 3.33GHz dual-processor, quad-core Intel Nehalem-based Xeon Genesis II system with 12GB of RAM, Asetek-based liquid cooling, and more importantly, three swappable system drives that let me change the computer's operating system by pulling out one drive and inserting another. This let me quickly and easily go back and forth between the operating systems, which I did multiple times during the tests.
My original test plan was to compare Windows Vista (SP2) with Windows 7, and Puget shipped both of these system drives with the original computer. Then it struck me that many folksmyself includedwere also interested in how Windows 7 compared to Windows XP. Because I asked so late, and because of holiday-related delays, I wasn't able to get the XP system drive in time to include the results with the original article, so I'll add those later.
I tested the Adobe Creative Suite using both synthetic and real-world tests. The synthetic tests started with a 3:30 DV project with a range of effects, including chroma key, slow motion, and color correction, along with a superimposed logo and timecode. Outputting to the presets shown in Table 1, Windows 7 proved, on average, about 8 percent faster than Windows Vista.
Next, I tested with a range of HD source formats. The HDV-source video was a 1-minute multicam project with one base track and a half-sized picture-in-picture (PIP) floating from upper right to lower left with one 360-degree rotation. I output all HD formats to the H.264-Apple TV 720p preset. As you can see in Table 2, rendering with Windows 7 was 5 percent faster than that with Vista.
The AVCHD and DVCPRO HD-source projects were both 1-minute long with three clips on the timeline, a single base clip, a static half-sized PIP, and another half-sized PIP panning from upper right to lower left with one 360-degree rotation. Rendering times in the two operating systems were virtually identical for both formats.
The Red Digital Cinema Red One source project was a variation on the same theme, with one static background track and a spinning, panning PIP. I rendered a 34-second chunk of the project, and Windows 7 proved 11 percent faster.
Next I rendered two real-world projects with the two operating systems. The first was the 53-minute second act of a ballet production shot with two HDV camcorders and edited using Premiere Pro's multicam feature. Effects included color and gamma correction, with about 2 minutes of rolling credits at the end of the video. I output this to widescreen MPEG-2, and Windows 7 proved 3 percent slower than Vista.
The second project was a 10-minute, single-camera widescreen DV concert shoot rendered to YouTube widescreen SD format. Here, Windows 7 proved 5 percent faster than Vista.
I tested the 64-bit version of Vegas Pro with an AVCHD-based synthetic test about 6 minutes long. As with the Premiere Pro projects, the project had two tracks: one static base track and a half-sized PIP that panned from the lower left hand corner to the upper right hand corner with a 360-degree rotation thrown in for good measure. Effects applied to the base clips included color correction, Gaussian blur, light rays, and film grain, with a millimeter logo overlay and timecode insertion atop.
I rendered using the output presets shown in Table 4. As you can see, Windows 7 was actually slower in the first two tests, though not by much, while proving 12 percent faster in the Windows Media tests. Unless you're a heavy Windows Media producer, there's little performance-based reason to prompt a Vegas user to upgrade to Windows 7.
The last time I benchmarked Vista performance, it was with Microsoft's initial release, and it proved 25 percent slower than XP. Here, I tested with Vista SP2. I have no idea whether it's faster than the original release, though you would suspect that it would have to be. Either way, it appears that Microsoft has avoided the same type of performance drop with the Vista-to-Windows 7 migration, though producers looking for a significant, across-the-board performance boost will be disappointed.