M5 Max: 32-core or 40-core GPU, which to choose?

Choose the 32-core GPU if 36GB of memory is enough for your work, and the 40-core GPU if it is not. That is the real decision, because on the MacBook Pro the GPU option on M5 Max is not only about graphics speed. It also decides how much memory you can order and how fast the chip can read it.

If you do not already know that you need more than 36GB, the 32-core version is the sensible default. If you run large local AI models, build heavy 3D scenes, cut video timelines that fill memory, or simply want the most this MacBook Pro offers, the 40-core is the one that opens the door.

What the choice changes

The figures below are from Apple's tech specs for both sizes of the M5 Max MacBook Pro.

So the two share their CPU, Neural Engine, media engines, ports, screen and battery. What differs is eight more graphics cores, more memory bandwidth and, above all, which memory sizes you are allowed to pick. A 32-core configuration is fixed at 36GB. For 48GB, 64GB or 128GB you must take the 40-core GPU.

That tie is the part people miss. Someone who wants 64GB for virtual machines and decides the 32-core GPU is plenty will find the 64GB option is not there.

Memory first, graphics second

For most buyers of this chip, memory decides it. Work out how much you need and the answer follows.

• 36GB is enough: the 32-core GPU is the match.
• You need 48GB or 64GB: you are buying the 40-core GPU anyway.
• You need 128GB: the same, and it is the only way to get 128GB in a mini-LED MacBook Pro.

You choose memory once, when you order, so this is the moment to be honest about the next few years, not just this one. How much RAM you need in a MacBook Pro helps you find the number, including how to read the Memory Pressure graph in Activity Monitor on the Mac you have now.

Bandwidth: what the extra 154GB/s is for

Memory bandwidth is how fast the chip reads its unified memory. The 40-core configuration has 614GB/s against 460GB/s, about a third more.

Bandwidth matters most for work that streams a lot of data through the GPU: very large AI models, big 3D scenes, heavy video effects and some scientific work. For ordinary creative and development work it is rarely the bottleneck.

Apple's own claims compare M5 Max as a whole with M4 Max: up to 4x faster LLM prompt processing, up to 3.8x faster AI image generation and up to 3x faster video effects rendering, Apple says. Apple does not split those results by GPU option. So there is no Apple figure for how much the 40-core GPU adds over the 32-core in a given job, and any number you see for that gap is someone else's test, on someone else's work.

Who should take the 32-core

• Video editors with 4K projects and a normal amount of effects. Both options have two video encode engines and two ProRes engines, so exports are not held back by the smaller GPU.
• Developers. Compiling is mostly CPU work, and both options have the same 18-core CPU. For virtual machines and containers the question is whether 36GB is enough.
• People who chose Max for the screens. Both options run up to four external displays, so a desk of monitors alone is no reason for the bigger GPU.
• Anyone who wants a Max and does not run work that needs more memory.

Who should take the 40-core

• People who run local AI models. Larger models need more memory, and the 40-core is the only route to 48GB, 64GB or 128GB. The extra bandwidth helps here too, since these jobs read memory constantly.
• 3D artists whose scenes fill memory. The same goes for heavy ray-traced scenes and simulation.
• Editors of very long, layered or high-resolution timelines. If the project is big enough that memory is the limit, the 40-core's memory options are the reason.
• Anyone who would rather not revisit the choice. If the work is your income and you keep a Mac for years, headroom is a fair thing to want.

What the choice does not change

It does not change the screen, the ports or the camera. Apple lists up to four external displays for M5 Max whichever GPU you pick, over any mix of Thunderbolt 5 and HDMI.

Battery figures are given per chip, not per GPU option. Apple's footnotes say its M5 Max battery tests, up to 20 hours of video streaming on the 14-inch and 22 on the 16-inch, were run on the 32-core GPU with 36GB. Apple publishes no separate figure for the 40-core. Weight is listed per chip too: 3.6 lb for the 14-inch M5 Max and 4.7 lb for the 16-inch. The charger is the same either way, 96W on the 14-inch and 140W on the 16-inch.

For the wider choice between this chip and the one below it, see M5 Pro or M5 Max.

A focus session for the long jobs

A 40-core GPU usually belongs to someone doing long, deep work: a scene, a model, a grade. Notchpals, whose characters live in the notch of every mini-LED MacBook Pro, has a mode for that stretch. Open its menu in the menu bar, pick Start a focus session, and choose 25 or 50 minutes. While it runs, only Focus moments, a critical battery warning and a meeting about to start can visit. It needs a Focus character in your crew, such as Debug, the rubber duck.

DebugBegin. If you get stuck, explain it to me slowly. Usually that's enough.

When the session ends, Debug turns up again to say so. The rest of your crew goes back to its usual gentle pace, which is set by one daily allowance for all repeating reminders.