Finally some numbers for the new Apple TV 4K that don't come from a keynote slide. An app developer (he makes I Call Window Seat, among other things) put a benchmark suite together over a weekend and ran it on the 2nd gen box with the A12, the 3rd gen with the A15 and the new one with the A19. He had Fable 5.2 write the suite, which will make some people roll their eyes, but the part that matters is that the method, the charts and the raw CSV/JSON are all public. That is more than most hardware reviews hand you.
I haven't gone through every chart, so I'm not going to quote figures from memory. What I'd look at is the gap between the A12 and the A15 versus the gap between the A15 and the A19, because that second gap decides whether a 3rd gen owner has any reason to move. For playing a 4K remux in Infuse the A15 was never the bottleneck. Extra headroom could show up in games, in heavier interface work, and in whatever Apple ends up running on-device for Siri, and synthetic tests only tell you so much about that last one.
The other thing worth watching is sustained load. These boxes usually live in a closed TV cabinet, and a short burst benchmark won't tell you whether anything slows down after twenty minutes of a game. If the dataset has repeated runs, that is the column I'd sort by first.
2nd gen owners will look at this and feel the age. The 3rd gen crowd is who this data is really for, and theirs are the replies I'd want to read.
I haven't gone through every chart, so I'm not going to quote figures from memory. What I'd look at is the gap between the A12 and the A15 versus the gap between the A15 and the A19, because that second gap decides whether a 3rd gen owner has any reason to move. For playing a 4K remux in Infuse the A15 was never the bottleneck. Extra headroom could show up in games, in heavier interface work, and in whatever Apple ends up running on-device for Siri, and synthetic tests only tell you so much about that last one.
The other thing worth watching is sustained load. These boxes usually live in a closed TV cabinet, and a short burst benchmark won't tell you whether anything slows down after twenty minutes of a game. If the dataset has repeated runs, that is the column I'd sort by first.
2nd gen owners will look at this and feel the age. The 3rd gen crowd is who this data is really for, and theirs are the replies I'd want to read.