It's relatively simple to choose a laptop model to buy from Apple, whether your biggest constraint is budget (that's what the MacBook Neo is for), weight (MacBook Air) or you need the MacBook Pro's power and connections. Once you've narrowed it down that far, though, figuring out how to configure it can become baffling. 

As far as I can tell, there are over 50 combinations of viable options for just CPUs, memory and storage across both the 14- and 16-inch MacBook Pro models, 33 for the 14-inch and 21 for the 16-inch. And your buying decision may be complicated by constraints imposed by the SoC design, notably with respect to memory, as well as the typical configuration limitations, like having to bump up a processor class from what you might otherwise need -- or want to pay for -- if you want 8TB of storage or 64GB of memory.

Watch this: New MacBooks Arrive With M5, M5 Pro and M5 Max chips

04:17

Both sizes have the same design and generally share the same features -- the screen, connections, wireless -- but the different processors support different generations of some of the features in addition to the expected performance differences. More specifically, only the 14-inch model has M5 configuration options, in addition to the M5 Pro and M5 Max, and the M5 is a generation behind the other two for its wireless and Thunderbolt. Those can be important: In my experience, Wi-Fi 7 is a lot more stable than 6E when connecting to 6GHz channels, and Thunderbolt 5 supports DisplayPort 2.1 as well as higher data transfer speeds compared to the earlier generation.

There are a couple of other things to consider when pricing your system. First, the $150 nano-texture option for the display, which makes it matte and nonreflective, is worth the extra money if reflections are an issue in your environment. I find it easier on the eyes. Second, they all come with different capacity power adapters -- 70W for the M5 and the M5 Pro (96W optional) and 96W for the M5 Max -- but the fast charge requires the 96W if that's your concern, a $20 upgrade if it's not included.

Matching needs to specs can be difficult, not just because you need to know how applications use the various components (and guess which applications you'll be using three years from now), but because performance is all about balance. For example, when we refer to an application or task as "GPU-intensive," we mean most of the work is performed by the GPU -- but that doesn't necessarily mean the GPU is, or should be, your primary concern. 

Maybe 90% of what the GPU does is render elements to the screen in your 3D design application, and that the final high-quality render is performed by the CPU. A powerful GPU but low-core CPU might make your designing and previewing experience fluid, but it may also make you wait irritatingly long for the final render -- especially if you have to iterate through different variations.

Go Pro (or Max)?

The M5 trails significantly behind the rest of the line, according to my testing, predominantly because CPU and GPU performance is mostly about multipliers: The individual cores in the M5 may perform the same throughout the line, but there are still a lot more of them in the higher-end chips, which makes the chips faster overall for tasks that use them.

Single-core performance is roughly the same across all three, because all the chips have the same super cores -- formerly known as "performance" cores. The same goes for the Neural Engine, which is used for basic on-device AI such as document summaries or generative writing, and is identical for the three chips.

As with previous generations, more GPU cores can make the difference between middling (or unplayable) gaming frame rates and good-to-excellent rates: When I tested Shadow of the Tomb Raider at 1,920x1,200, frame rates increased from 56 frames per second for the M5 to 96fps (10 cores) on the M5 Pro (20 cores) to 150fps on the M5 Max (40 cores). The new on-GPU neural accelerators can be a performance boon for boosting precision AI-driven operations such as generative imaging and video, machine learning and coding. One potential weakness of the entire M5 line, though, may be its lack of native support for new arithmetic data types, such as FP8 and FP4, which offer improved speed over higher-precision types but better precision than the faster integer.

Multicore performance widens the CPU gap between the M5 and the other two not just because of the multiplier, though, but because of the mix of cores in the Pro and Max. They exchange the low-power efficiency cores of the M5 for new, more balanced-tradeoff performance cores, which give the M5 Pro and Max a boost when running at max power. 

One advantage the M5 Max specifically has for high-resolution video editing, beyond just the number of GPU cores, is its two Pro Res accelerators. They help speed encoding and decoding.

Memory and Storage

Because you can't upgrade memory post-purchase but you can always add external storage, getting as much memory as you can afford and think that you'll need eventually is key. In fact, if you think you'll need more than 32GB then the amount will determine the minimum processor you can get: 48GB requires at least an M5 Pro, 64GB requires the higher-end M5 Pro and only the top-end M5 Max supports 128GB.

Determining how much memory you need is easier said than done, though, because it depends on the mix of applications you use, the size and complexity of the files or AI models you work with and how many at a time, and more. Apple's unified memory architecture, which shares system memory between the CPU and GPU, complicates that calculation as well.

How much MacOS can swap or compress matters, too. For instance, at the moment I've got only Chrome, Slack and Notes open right now, but 14GB of my 16GB memory is in use -- lots of Chrome tabs and big, multitab Google Sheets will do that. It's not affecting performance, but during CES I had so many Chrome windows and tabs open that my system slowed noticeably.

Storage, on the other hand, is easier to judge, not just because you can add more but because you obviously know whether the amount you have now is sufficient. If it's not, jump to double the capacity. If it's I-can't-function insufficient, consider the next tier up from double.

Keep in mind, though, that external storage -- especially external storage that's fast, like Thunderbolt 5-connected drives -- isn't necessarily a cheaper solution than internal SSD, and internal will always be faster, which is important for tasks like ML and video editing. Similarly, as you head up the line of chips, memory bandwidth grows as well, which provides a boost for memory-intensive tasks like ML and inferencing and high-res video editing.