This is a test of different bitrate settings using Magic Lantern v0.1.9 with Canon's v2.0.8 firmware. It should also be noted that I was using a 32GB 60MB/s UDMA Sandisk Extreme CompactFlash card.
I did this test because I found that most of the other tests I've seen have been very inconclusive. I was looking to discover what exactly higher bitrates effect. Also, what effects higher bitrates.
While the qscale does effect bitrates directly it is not the sole contributor. ISO settings will also greatly effect your bitrate. So if you are using Magic Lantern be aware that you have to adjust your qscale accordingly with your ISO if you want the optimum quality. In this test you can see that qscale -16 runs at 146 Mbps with 6400 ISO and at 58 Mbps with 100 ISO. My card allowed me to run a bitrate of 55-60Mbps before filling the buffer which means that my ideal settings for 100 ISO should be qscale -14 (-15 & -16 would sometimes fill up after 20 seconds or so) and 6400 ISO should be qscale -7 (-8 would give me two buffer bars). These are the settings that worked well for me but I would suggest running your own tests with your set-up. I kept an excel sheet recording how my qscale and ISO effect my bitrate so that I am able to refer to it if unsure what qscale to use with my ISO settings.
As far as picture quality is concerned I think it is fairly obvious that the higher bitrate - the better. But in which way? If you look at the comparison between the -16 and -6 flames you can see how boxy the flames are on the -6 compared to the -16. This is, I think, going to be the biggest difference between your bitrates. It is also a difference that is especially noticed when zoomed in 1000%. Note that I was only able to sustain -16 with 6400 ISO for 28 frames so it could never be used for any practical purpose. So is this a game-changer? That's up for debate but I think we can all agree that it's a step in the right direction.
Here is a direct link to this post on my blog: tylerjclements.com/2011/05/bitrate-test/
Loading more stuff…
Hmm…it looks like things are taking a while to load. Try again?