@djsumdog @tk @icedquinn @iceloops @sun >I think storage is too because of the encryption chip.
Mac Studios use the m.2 connector for flash modules, it's superficially similar to m.2 SSDs but most of the controller functionality is on the M* SoC and naturally they are encrypted and paired to the device.

>Once x86 pulls ahead on power consumption
I'm still waiting for an x86 SoC to not lose to M1 on any efficiency-related metric by my testing. They've had years at this point, and node advantages now, but when I grabbed a Ryzen 7840U, set it to match observed M1 MBA power, and threw some of my leetcode-tier solvers at it, it was considerably (~1.2x runtime) slower.

Yes, the 7840U is much faster (~0.5x runtime) at matched power on AVX-style workloads.
No, x86 is not competitive on CPU-style CPU workloads. It's ridiculous to not have a clean sweep against chips made by the same foundry 3 years prior on a less-refined node, or at least be equal on the worst cases.