I'm curious as to how accurate the reported speed is during recording. I'm also curious as to wether some frame sizes write to a particular model of card better (sort of like how the cards often perform better in different parts of the benchmark based on write size).
I'm using a Sandisk 16GB Extreme Pro (60 MB/s).
I'm having difficulty producing results with any consistency but using raw_rec with global draw off I have seen writes of up to 59.6MB/s on this card (yay!)
However I'm not really sure how to go about achieving this on a regular basis. Often a resolution and frame rate that should fall below the maximum performs worse. For instance if I try to record at 57 MB/s the card will only do 55 but if I ask for 62 it will do up to 59. Obviously 62 is not possible

I'm just wondering if there is anything obvious I am missing here while I am testing.
I've also noticed that buffer fill method is available in mlv_rec but not raw-rec. Is this setting unneeded in raw_rec?
If anyone with the same card is interested:
raw_rec.mo
1664x936 (16:9)
Global Draw: Off
Preview: Auto
Card Warm-up: 256 MB
Extra Hacks: On
Framerate: 24
FPS Override: Off
This is 1x from the .bin
Sometimes it will only reach about 54MB/sec but if I stop it and start again immediately it will go to 59.x (Perhaps this is working similarly to a warmup?)
On some runs it appears to hit about 59.5 and stay there, other times it seems to go between 57.5 and 59. It will often start at 54 or so and slowly work its way up to 59.
Like I said I don't know what I should do in my testing to try and get some consistency so any advice would be appreciated.
I just ran another "test" and with the same settings as above but mlv_rec.mo peaked at 59.8MB/sec using Buffer fill method 3.
EDIT: Added more info.