Apple's latest M3 chipsets were announced at the company's "Scary Fast" launch event on the evening of Oct. 30, alongside new MacBooks and a new iMac. The M3, M3 Pro, and M3 Max all arrived at the same time for the first time and are very likely to be fabricated on TSMC's N3B 3nm process, a first for the computing market. With the M2 for example, Apple announced the base-level M SoC first followed by the Pro and the Max at another, later event, though that's not the case here.

As for what's different, we'll first break down the differences between the M3, M3 Pro, and M3 Max, and then compare them to their adjacent M2 counterparts. Apple's MacBooks are some of the best laptops around, but the M3 series doesn't do much to move the needle any further.

Apple is doing things differently with M3

If I were to guess what's happening here, TSMC has reportedly suffered from a pretty low yield when it comes to N3B. Some reports suggest that TSMC has achieved a yield of only 55%, and Wccftech has said that Apple is paying for only the good chips, rather than the entire wafer itself. What this means though is that almost half of all produced chips are considered "bad" and can't be used, which reduces how many TSMC can actually pump out for Apple.

The MacBook Pro and iMac are expected to be devices that shift fewer quantities when compared to the likes of the MacBook Air, and this may be why Apple has launched all three at once and only in both the MacBook Pro and the iMac. It's possible that a MacBook Air may come later should yield improve, but for now, we can only speculate.

What's especially interesting is that Apple is absolutely engaging in a crazy amount of flexing right now with the M3 series. The 92 billion transistor number for the M3 Max is a massive number of transistors that sets the record for the most in a consumer laptop SoC. The M2 Max has 67 billion for reference, so this marks a 37% increase. While it's clear that there are definitely some process advantages thanks to TSMC's N3B, it's also possible that this makes the M3 Max an even bigger chip than the M2 Max already was.

Apple's M3 differences to the M2 series

CPU: Incremental improvements

The improvements over M2 are fairly small, but Apple had a few facts and figures to share when it came to performance. We have no real way of knowing if these cores are based on the A17 Pro cores or the A16 cores. The M1 chip was based on A14 cores and the M2 was based on A15 cores, so while it may seem like it makes sense for the A16 to be the basis of M3, there's more to it than that. The A17 cores are made with TSMC N3B in mind (so it would make sense that they would be used instead) and there's not enough information to definitively state that these cores are either based on A16 or A17 cores. We still don't even know the A17 core codenames, aside from the SoC itself being called "Coil."

In terms of tangible differences when compared to the M2 series, the M3 is said to have a 15% performance uplift in its performance cores and a 30% uplift in its efficiency cores. We don't know how these figures were obtained as Apple does not supply that information, so it may not be accurate in all situations and may even be as a result of a higher clock speed rather than any architectural improvements.

Interestingly, there are some changes to the number of P-cores when comparing the M2 Pro and M2 Max. Apple has dropped the number of P-cores in the M3 Pro from 8 P-cores to 6, and has increased the number of P-cores in the M3 Max from 8 P-cores to 12. The M3 Pro has also received two extra E-cores, meaning that Apple has somewhat rebalanced this chip. There may not be big performance improvements with the M3 Pro this year over the M2 Pro, but it should have better battery life.

GPU: Maybe a big step forward?

The GPU of M3 though seems like it can be a big step forward. Firstly, it supports hardware ray-tracing, a first for Apple Silicon. Ray-tracing was introduced with Apple's A17 Pro in the iPhone 15 series, and we would expect given the close relationship between Apple's M series of chips and A series that this is almost certainly just a larger implementation of the same hardware.

The above graph, shared by Apple, is all that we have to go off of when it comes to the performance of the M3's GPU. Even the claim of a 2.5x faster GPU is misleading, as Apple states the following:

"Render performance in Maxon Redshift is up to 5.3x faster than the fastest Intel-based MacBook Pro, and up to 2.5x faster than the 16‑inch MacBook Pro with M1 Max."

Apple is comparing it to the M1's GPU (not the M2) and is comparing it to a GPU that does not have ray tracing. While it's not clear if the Maxon Redshift test used hardware acceleration where available, if it did, then that would explain such a large leap in performance. When compared to the M1 Max, it was only 65% faster at rendering. The M3 GPU is undoubtedly the fastest Apple Silicon GPU yet, but where it lies in the plane of performance isn't exactly clear.

Referencing again back to the graph above, the best we can see is that the M3 consumes half the power to achieve the same performance of the M1. However, we can also see that it has a much higher power limit, so while it's definitely more powerful, it uses a lot more energy to achieve that output, too. In other words, until we get testing actual devices, it's hard to say how the M3 really fares.

Apple also introduced a feature called "Dynamic Caching" though was, characteristically, scarce on details. Apple describes it as a feature allowing M3 chips to only use the exact amount of memory that a graphical task requires. Apple describes it as "dramatically" increasing GPU utilization and says that it "significantly increases performance" for demanding apps and games.

GPUs run a lot of computation in parallel, and what Apple's describing appears to be a way to essentially pre-calculate the right amount of memory for more complex shader calculations. We'll need to wait for more details on this, but it could well be the case that this ends up being a genuinely important feature for Macs when it comes to gaming going forward.

Memory: Some weird changes

From poring through the data shared by Apple thus far about the M3 series, there is a notable change in the M3 Pro versus the M2 Pro. Apple has dropped the memory bus width from 256 bits to 192 bits, decreasing memory bandwidth to 150GB/s.

Why Apple has done this to the M3 Pro is anyone's guess, and it's not as if Apple will ever answer any questions about that either. However, Apple's M3 Max will support 128GB of RAM at maximum, making it a boon for hardcore enthusiast users. It has a 512-bit memory bus at maximum, but the cut-down version of the M3 Max supports only 96GB of RAM.

If you have an Apple Silicon Mac, it's probably not worth an upgrade

Apple went hard on comparing to Intel MacBooks this year, and when comparing to Apple Silicon, it was really only compared to the original M1 SoCs. Apple Silicon's biggest leap was when it jumped to the first M1 chip, and after that, those yearly improvements have simply been incremental. The M2 series and the M3 series are still not massive jumps forward for someone who has an M1 device, and the M3 Pro even appears to be more of a sidegrade from those who may be coming from an M2 Pro.

These new Macs are great, and the new M3 series is pretty good, but the changes are minute and until we get our hands on devices, it'll be hard to say if it's really worth upgrading just yet from any of the other best Macs around.