Remember how I performance tested my fancy new MacBook Pro in February and it was slower at AI-based noise reduction in DXO PureRAW than my previous-generation fewer-cored M1 MacBook Air? Remember how I discovered that it wasn’t actually engaging the Neural Engine of my Ferrari of a computer and I reported it as a bug to DXO? Well it’s finally fixed and the performance numbers are in: my MacBook Pro is faster than my MacBook Air. Phew!
DXO released version 3.0 of PureRAW in late March, and hidden among all of its actual feature changes is support for the M2 Pro’s Neural Engine. If you recall from my post, the Neural Engine hardware on the M2 should be almost 50% faster than the one in the M1. So is it?
I retested with my 18 RAW files, and discovered another hidden feature: Pure RAW 3 is a bit faster than version 2 overall. I retested my MacBook Air and completed the test about 18 seconds faster (about 7%) than in version 2. This brings its total time to 3:50. My Pro, as expected, is even faster – completing the test in 2:49. That’s a 27% improvement on the Air. It’s not 50%, but it’s still faster and it doesn’t use the GPUs at all. An updated graph is below.
I was so excited by the improvements that I purchased Pure RAW 3 and have been using it in my recent posts.
Relatively Ambitious 