What stands out about these tests to me, apart from Apple Silicon remaining highly impressive, is that AMD in both and Intel at least in multicore are beating the Snapdragon on performance per watt. Many people boil Apple’s efficiency down to an ARM vs x86 thing, but that’s a tiny part of the chip by now, it’s more in Apple’s bespoke architecture around it.

Here we see AMD even on x86 beating an ARM chip on Perf/watt, Intel on multicore, and Meteor Lake launching within a month will drop power use by 50%. It’s not just or even mostly the ISA, it’s everything else around it that matters more.

8-Core CPU:

• Up to 35% faster than M1
• Up to 20% faster than M2

1. 4-Cores (Performance) Top = 4.056 GHz (or ~3.6 GHz when all cores are loaded)
2. 4-Cores (Efficiency) Top = 2.748 GHz

10-Core GPU:

• Up to 60% faster than M1
• Up to 20% faster than M2

1. Dynamic caching, where cache and memory are allocated dynamically based on the actual requirements of applications.
2. Hardware-based ray tracing, mesh shading and AV1 decoding.
3. However, only two displays can be used at a time.

The other interesting summary figure is more-or-less maintaining power efficiency as before on lower power with comparable performance to other chips at higher power usage.

=

Either way it is spun (pun intended), it’s very impressive performance boost and fast cadence of release by Apple. notebook.check is a handy website, thanks for posting OP.

I have a mid 2019 i9 MacBook Pro, would you say the M3 is a good replacement for Final Cut editing and Blender animating? It’s not a pro processor, but it seems to have higher benchmarks, but what’s the catch?

But can it pull off the most impressive feat of all. Shipping with 16GB RAM standard.

“Analyzed”, man do I miss the real analysis that Ian Cutress did on these things. The actual architectural deep dives.