There is something interesting going on with those Javascript GeekBench benchmarks worth bearing in mind.

Despite Apple now for years destroying everything but the latest high clocked Intel processors running Chrome (all other browsers get hammered), this doesn't really translate into much real world.

Sure, the JavaScript performance is probably excellent, the best maybe, but I feel like they've only optimized for that.

I don't see video encoding or 3d rendering being a strength of Apple processors any time soon, but I guess even on Desktops that's moving and moved to GPUs.

I still have a hard time believing those benchmarks. Are they really comparing the same things - just compiled for different architecture? Or is there more behind it - one is a "mobile" benchmark, the other something else?

> Are they really comparing the same things - just compiled for different architecture?

IIRC they were just Geekbench 4 and 5, which are standard benchmark suites for the MacOs and iOS ecosystems, and IIRC only the subsets of the benchmark suite that ran on both the Macbook Air and the iPad Pro where compared.

The only thing that I know about Geekbench is that it benchmarks application performance, e.g., by running an application and performing a task (e.g. Cinema, Adobe, etc.).

I don't know if they use the same applications and the same tasks on every OS/arch combination. If they don't, the scores cannot be compared, and being able to compare the scores is the only purpose of these tools... so I hope so (the article I read seemed ok, so I had no reason to suspect).

Does anybody know for sure?

They don't benchmark apps, but test on a set of standard workloads and algorithms.

https://www.geekbench.com/doc/geekbench5-cpu-workloads.pdf

I'm with you. Benchmarks should always be looked at with extra-care. The environment of a tablet is quite different, and there's not much things that are benchmarked. Not saying that those cpus aren't great, but benchmarking those in the exact same conditions than an AMD or Intel CPU on a laptop, desktop or server might give some results that are more contrasted.

Apple as a company is all about integration. It’s amazing how many shared parts exist between the iPhone, iPad, and yes, even the Mac to some extent with the T2 chip.

A larger, thermally unconstrained ARM octacore, with SMT? Maybe with 20 hours battery life (but realistically they’ll just make the device thinner lol).

I think it’s entirely predictable for Apple to go ARM on Mac.

There will always be a question of backward compatibility due to the nature of the Mac ecosystem that you don't have in iOS. Then you have the issue of getting developers to migrate any existing arch specific optimisations. This can be mitigated somewhat and Apple has done this a few times before, but it's still a massive undertaking.

It may be easier for them to play the CPU vendors against each other if there is real competition in the AMD64 space. There are rumours of Apple going ARM on Macs, so it may be that that path is set, but if it isn't, there is sense in holding out the transition to ARM at this time.

Honestly, Apple has killed backward compatibility multiple times, first from PowerPC to x86-32 and x86-64, and now, killing x86-32 in Catalina.

Apple also has a good solution to this problem (the AppStore solves it), and a big interest into locking developers into this solution...

So Apple doesn't need to restrict themselves to any hardware vendor when it comes to CPUs, and due to Metal, they don't have to do that either when it comes to GPUs.

I wonder how they'll get past the RISC/out of order execution issue that still is an issue with ARM chips.

Sorry, can you elaborate? Apple's "A" series CPUs have been out-of-order for years and at least a few years ago had a similar superscalar/ooo structure to Intel's Haswell chips. (Apple poached a bunch of Intel processor designers…)