G.Skill DDR3-2200 C7 - Transcoding Video

Transcoding over the CPU or GPU

We recently added another benchmark to the test-suite. It's MediaShow Espresso. The fun thing about this video transcoder is that it can utilize the GPU to assist it with the transcending process.  However, you can also solely use the CPU, making this a very interesting benchmark and you can check out behavior of CPU transcoding AND GPU transcoding all in one test.

Below you can find the first results of this new test. In this test we transcode a 200 MB AVCHD 1920x1080i media file to a 1280x720P MP4 binary.

Here's where we change the benchmark test method a little. What we'll do is test with the versy same PC and the same clock frequency and then test at 1333, 1667 and 2000 MHz and hunt down any performance differences.

For competitive scaling i inserted a Core i5 750 with 1333 MHz C9 memory in there, the equal AMD 890FX with 1090T processor and 1333 C9 memory and then to top it off, that 1000 bucks costing Core i7 980X setup on X58 with it's 1333 MHz clocked memory in a blazing triple-channel configuration.

Make no mistake, you can not compare the platforms against each other that way. But they are inserted for scaling purposes only. The AMD setup with 1090T processor was 1 second faster @ 2000 MHz memory over a 1333 MHz configuration.

Multi-threaded Video Transcoding H.264

x.264 is a free library for encoding H.264/MPEG-4 AVC video streams. Encoding/transcoding to that format is one of the most intensive tasks a processor can perform. As such this is probably the best test in the entire review. We encode a h.264 Dolby Digital 1080P trailer of 150 MB to Matroska x.264 with 5.1 channels AC3. It's compressed in such a way you can play it back with Haali media splitter and/or FFDSHOW codecs. We use Handbrake software which is multi-core aware... the more processor cores it sees, the faster it can and will transcode.

This software is a perfect benchmark for CPU and memory testing as it is very sensitive to multiple cores and memory frequency. The displayed number is the number of frames rendered per second averaged out over the encoding process. The higher the number, the faster the performance is.

Again if we focus solely on the G.Skill memory, we do measure a performance difference in-between the tested timings, but it's really close to NIL.