Agreed with the above general points. As always, buy what you need when you need it and don’t fret/obsess over a future that may never come. Plus, there’s both the opportunity cost of waiting an indeterminate time (ie., think of all the amazing stuff you could have built/done on the currently available product while you were waiting) and the slippery slope of release estimates (e.g., M2 going from mid 2022, to Oct 2022, to Nov 2022, to Q1 2023 to….?)
That said, I’d also like to add something that I feel is often lost in these discussions:
1. This is one benchmark and an unverified one at that, meaning it could be from an actual M2 Max that had other processes running or it could be a forged result (for the lulz, etc.) or it could be a pristine M2 Max running with no background processes.
2. Based on the M1 to M2 comparisons, the biggest gains we’d expect to see are on the GPU side, which this does not cover.
3. Additionally, unless it starts adding new measurements, GeekBench is going to become a less meaningful benchmark over time. For example, we know that the M2 added a ProRes encode and decode engine and added hardware acceleration for ProRes RAW, neither of which (to my knowledge) are measured by GeekBench. It’s reasonable to assume that an M2 Max would add some multiple of these capabilities (e.g., 1 to 2 ProRes encoders/decoders), which could significantly impact some workloads. Plus, I expect Apple will continue to add these dedicated capabilities at a hardware level (just as they did with the Neural Engine) because they’re in a unique position to do so, differentiating them in the marketplace.
4. This and likely most other benchmarks don’t matter to you since they only serve as a rough proxy measurement of potential performance. What does matter is your specific workload (apps and data) and use case (where/how you’ll use it). For example, I have an M1 Max MacBook Pro and while I‘m often tempted by the M1 Ultra or what the future may bring, it’s helpful to bring it back down to reality for me by looking at measuring what I actually do. For example, I currently have a small, hobby Xcode project (Swift/SwiftUI) that builds and runs all of its tests in a whopping 7s. Say it grows 100x and takes 700s. Even if the Ultra scaled linearly (it doesn’t) and cut it down to 350s (~6 minutes vs ~12 minutes), that doesn’t really change my experience. I’d still likely check email, get a cup of coffee, stretch my legs, whatever during that time. For me, I generally wait until something offers a 4x improvement or until it actually starts to bottleneck my productivity. (Or, let’s be honest, when I just really, really want one.)
As for comparisons to AMD/Intel, since I’m not tempted to maintain a Hackintosh or run something other than macOS, I just don’t focus on them, particularly for my mobility focused workload. Plus, Intel’s past shenanigans have adeed them to my “do not support“ list. However, there are valid scenarios for considering them (e.g., gaming, large ML workloads, etc.), so if that’s one of your primary concerns, by all means consider them, or cloud-based resources if appropriate.
