I mean, even on the laptop front Apple's performance-per-watt is in question. In Blender, the ~80w 2070 MaxQ performs roughly the same as the ~90w M2 Max. That'

I mean, even on the laptop front Apple's performance-per-watt is in question. In Blender, the ~80w 2070 MaxQ performs roughly the same as the ~90w M2 Max. That's a pretty impressive showing for a 12nm dGPU versus a 5nm SOC, and things only get worse when you bring in the newest laptop chips Nvidia has. Bit of a moot point, but it's worth mentioning. I don't think its unreasonable, though. Apple has never been about putting the highest specs in a box, Apple users were prepared for M1 to not dominate every scoreboard. The problem is more that their SOC-style approach isn't very effective for desktops. It helps you make cute workstations like the Mac Studio (which looks straight off Minisforum), but it's obviously hamstrung for larger tower-style distributed boxes. It's even evident in their lineup - there were a lot of rumors about an M1 Extreme that got cancelled for various reasons. It honestly just looks like they're hitting the limits of phone-style SOC manufacturing, which is good in some places (latency, integration, power management) but bad in others (cramped die, bottlenecked IO, thermal distribution).