Here are results from a low-light, high ISO video test I conducted in 1080p24 with the following three HDSLRs:
Panasonic GH13 - 100Mbps Max Latitude Native 24p Patch @ ISO 1600 Panasonic GH2 - Unhacked 1080p24 @ ISO 2000 Nikon D5100 - 1080p24 @ ISO 4500
These tests were all done with the same lens mounted on a tripod: a Tokina 20-35mm f2.8 in Nikon mount, set at f2.8. I shot the D5100 at about 28mm focal length, and to produce a similar field of view, shot the GH1 and GH2 at about 22mm. Shutter speed was set to 1/60 second.
The wide range of ISO settings resulted from my efforts to achieve a reasonable match in the exposure levels of the three cameras. While it's well known that the GH1's ISO calibration is set lower than the GH2, I was surprised by how inflated the D5100's ISO settings turned out to be. This test really changed my opinion of the Nikon's reputation for high ISO performance.
I've updated the video sample files to identify each camera, see which one you prefer:
Well my 2 cents worth.. B is the better overall in terms of contrast and black level.. A is a little softer and C is pretty flat with a lot of visible noise noise.. My guess is: A: GH13 B: GH2 C: Nikon 5100
@humpman - Congrats on correctly identifying each camera!
The Nikon D5100 looks best to my eyes, but only after I set its film mode sharpness to 7 (on a scale from 0-9). I also have Nikon's Active D-Lighting set to Normal, and High ISO NR set to Normal as well. The D5100's H.264 encoder is limited to about a 20Mbps bitrate, using B-frames. At this low a bitrate, I think it's better to keep the noise down to avoid overtaxing the encoder.
Here's a sample from the GH2 hacked with cbrandin's 66Mbps AQ2 settings:
Thanks for testing! Would be interested in 800 and 1000 ISO. Don't like the banding in the GH13, but the grain is pretty chunky in the upper part of the fan in the GH2. Astonishing what the Nikon uses for ISO--but good results.