DivX 6.7 Encoding

Website: Divx.com

We tested video encoding performance using VirtualDub-MPEG version 1.6.15 and DivX 6.7 with multi-threading enabled and SSE2 or SSE4 enabled where appropriate. We did a two-pass encode of a 15-minute 276MB MPEG-2 digital TV recording with a target file size of 100MB.

While this benchmark does use SSE4 which is currently unique to Intel's 45nm processors, the latest DivX 6.7 has enhanced Multi-threading support and also uses SSE2 enhancements, which applies to all of the processors tested.

While Intel has admitted to bit-tech that it throws out some "best case scenario" numbers to the press, we took our usual MPEG-2 digital TV recording that we use for our XviD tests and DivXed it instead using the same VirtualDub we always do.




As expected the QX9650 performs trumps here by nearly 40 seconds (18 percent) over that of the same clocked QX6850, however some/most of this can be directly attributed to the extra 4MB of cache it has because the SSE4 is only directly responsible for the level of motion detection, not a more universal function like compression.

However, we're not complaining either way - it's "free" speed and if you're an encoding guru a cheaper baseline Yorkfield, overclocked, will make you a more productive person with money left over. If you can't wait and have money to burn - then a QX9650 is your friend, or even look at the new 1,600MHz FSB "Harpertown" Xeons in a new "Stoakley" setup.

AutoMKV x264 Encoding

Website: Doom9

We tested x264 compression using AutoMKV version 0.90 and x264 to to compress a 1.1GB DVD VOB file into 350MB MP4 file using a two-pass encode and we used a 112kbps LAME encoder to compress the audio. The whole process is dependent on both single and multi-core performance and the entire encoding time was recorded.

There's quite a shift to using MKV or MP4 wrappers for x264 content now, especially for movie content and those in the large anime fansubbing community. x264 doesn't have the same SSE enhancements as DivX 6.7 but the benefits of extra cache and better memory performance should still show notable improvements.


The QX9650 again churns through the process a minute faster than the same clocked QX6850 and more cores make for lighter work over and above that of general front side bus performance. The AMD Athlon 64 X2 6400+ even outperforms the E6750, however the E6850 is still over five minutes faster than it and doesn't take a lot of effort to overclock from 2.67GHz to 3GHz with a Core 2 Duo.

MP3 Encoding:

We used LAME MT for our audio encoding test - it's the multi-threaded version of the popular LAME MP3 encoder. We ran tests with both Intel's and Microsoft's compilers - naturally, the Intel compiler resulted in some performance increases on Intel's processors. We converted all 18 tracks from Moby's Play album to a 192kbps variable bit-rate MP3.



The notable difference is less with MP3 encoding, however the QX9650 is still faster than the QX6850 by seven or eight seconds even in such a short test. Even without using the Intel-optimised compiler, the AMD processors are still outperformed by the slowest Intel processor with a gap of 20 seconds.

Windows Media Encoder x64 Edition 9

Website: Microsoft.com

Windows Media Encoder x64 Edition 9 comes as a free extra for 64-bit versions of Microsoft Windows. We encoded a 1440x880 Taxi 3 h264 trailer into 1280x720p VBR Windows Media Video file with CBR six channel audio in two passes, measuring the length of time it took for both.


Again the QX9650 comes out slightly ahead but overall more cores makes for better performance, more so than simple memory bandwidth.