This test is a bit flawed, i own both a Epic-W and have Sony cameras, RAW as well, tools for the job in hand. I love the Sony RAW images and the Red too.
You can't rate the helium at 800 iso, you need to rate both cameras at their native ISO otherwise it's just not a valid comparison. The helium is more sensitive, if you want to match exposure stop the lens down.
Then with the footage keep it RAW and then into linear space and actually compare image to image, data to data - keep it simple. Using LogC to 709 LUTs is another step where differences can happen. That 3D LUT could even be bringing highlights back (i.e. if Arri knows the clip point of a channel then it can use a 3D LUT to mix back other channels - i'm not saying they are but they could and it would be valid). But you could do that with Red too. There are so many points in a pipeline where clipping can happen erroneously.
So to make sweeping comments you really need to test bare bones like for like otherwise it's not fair on either camera.
It's also valid to test a camera in the way that you personally choose to use it, but the results aren't definitive then, they're a combination of camera, taste and workflow and everyone works differently. It's not quantative.
I appreciate these tests usual degenerate into fan-boism, but it is possible to test side by side in linear - which is the only valid quantitive result, assuming point of capture is fair for both.
This test is a bit flawed, i own both a Epic-W and have Sony cameras, RAW as well, tools for the job in hand. I love the Sony RAW images and the Red too.
You can't rate the helium at 800 iso, you need to rate both cameras at their native ISO otherwise it's just not a valid comparison. The helium is more sensitive, if you want to match exposure stop the lens down.
Then with the footage keep it RAW and then into linear space and actually compare image to image, data to data - keep it simple. Using LogC to 709 LUTs is another step where differences can happen. That 3D LUT could even be bringing highlights back (i.e. if Arri knows the clip point of a channel then it can use a 3D LUT to mix back other channels - i'm not saying they are but they could and it would be valid). But you could do that with Red too. There are so many points in a pipeline where clipping can happen erroneously.
So to make sweeping comments you really need to test bare bones like for like otherwise it's not fair on either camera.
It's also valid to test a camera in the way that you personally choose to use it, but the results aren't definitive then, they're a combination of camera, taste and workflow and everyone works differently. It's not quantative.
I appreciate these tests usual degenerate into fan-boism, but it is possible to test side by side in linear - which is the only valid quantitive result, assuming point of capture is fair for both.