Technicolor just released a Picture Profile called CineStyle (http://bit.ly/lTApYK) for Canon DSLR's. It's similar to the SuperFlat style most users have defined for themselves, but seems to be providing HUGELY better latitude in the shadows. This obviously creates a wonderful image to work with when colour correcting.
Anyway, in their support document, they suggest, along with specific settings for Contrast and Saturation, that ISO be used in multiples of 160. I wanted to see if this was true, so I ended up doing a little home experiment...
Shutter Speed: 1/50th
Frame Rate: 24
Music by: Andrew Schär - "Outnumbered" (Yip, it's one of my songs. Check out about.me/andrewschar for more info, or to download my music for free, go straight to: http://bit.ly/ajs-mu)
The results as you can see in the video, are pretty much spot on with what Technicolor has said. Interestingly enough, it seems to me that ISO 640 has about as much noise as ISO 100.
It also seems that ISO multiples of 125 are specifically bad. ISO 1250 is pretty much comparable to IS0 125!
NOTE* For the purposes of this test, I boosted Curves to make the noise more observable, so it won't be this bad in the real world! I've also added two pictures of screenshots from low to high noise.
It clearly illustrates that best ISO's to use for video, starting from lowest noise level to highest are as follows:
160, 320, 640, 100, 200, 400, 800, 1250, 125, 250, 500, 1000, 1600, 2500, 2000, 3200, 4000, 5000, 6400.
Of course, if you're shooting low-light and you need a high ISO, the best ones to go for would be: 1600, 2500, 3200 & 5000.
If there's plenty light, keep things on ISO 160.
Multiples of 160 are best, then multiples of 100 and then multiples of 125.
Vincent Laforet says that ISO in Still Photography is different compared to when shooting video. I did a quick test using the exact settings from this test.
The result? It seems it is different - there's WAY more RGB colour noise, and from what I can see, ISO 160 still somehow manages to sneak in as king of low noise.