The build from 2013/05/19 for 5D2
a75c4a74d4b9
- I got some more frames vs previous build
Thanks a.d. !!
Test conditions:
- Komputerbay 32GB 1000x
- CF card in camera formatted prior to each setting
- 3 consequtive runs
- black image (lens cap on)
Results:
- @ 1880x840 25fps hacked on: 1104 (64.1 MB/s), 4GB limit (65.2 MB/s), 1426 (64.6 MB/s)
- @ 1880x840 24fps hacked on: 4GB limit (63.1 MB/s), 4GB limit (63.1 MB/s), 4GB limt (63.1 MB/s) <- stable!
- @ 1880x840 25fps hacked off: 1280 (64.5 MB/s), 1288 (64.5 MB/s), 1380 (64.5 MB/s)
- @ 1880x840 24fps hacked off: 4GB limit (63.1 MB/s), 4GB limit (63.1 MB/s), 4GB limt (63.1 MB/s) <- stable!
- @ 1880x960 25fps hacked on: 181 (64.1 MB/s), 189 (64.2 MB/s), 172 (64.2 MB/s)
- @ 1880x960 24fps hacked on: 246 (65.4 MB/s), 246 (65.1 MB/s), 246 (65.1 MB/s)
- @ 1880x960 25fps hacked off: 181 (64.5 MB/s), 181 (64.6 MB/s), 181 (64.3 MB/s)
- @ 1880x960 24fps hacked off: 246 (65.1 MB/s), 246 (64.8 MB/s), 246 (64.9 MB/s)
I have the impression while "hacked on" gives higher peak rates, "hacked off" results in more stable write speeds.
I also did a couple of tests with actual footage (not just lens cap on black images) and that resulted in considerably lower frame counts. How is that possible with uncompressed RAW footage? Is DNG somehow compressed? (fast runlength compression or something like that?)