Computer

Apple stops using toy SoCs in its new M1 Max and M1 Pro

In every new node, they start manufacturing small chips before making the leap to the more complex ones, this is the reason why TSMC’s 5 nm node has been creating small chips for Apple and now those for Cupertino. have announced their M1 Pro and M1 Max, vitaminized versions of their M1 processor from a year ago that benefit from the ability of the world’s largest foundry to offer its N5 node to make larger chips. Let’s not forget that AMD’s RDNA 3 architecture GPUs and NVIDIA’s RTX 40 will use the same manufacturing process shortly, so this is the first confirmation of the Taiwanese advance.

On the other hand, we cannot rule out the fact that in a laptop for years they have used APUs of sizes very similar to the recently presented M1 Max, and in Apple we had already seen versions with a greater number of cores and memory interface of the iPhone chips. for the iPad, taking advantage of the greater consumption and available space.

Apple M1 Pro, your first real PC APU

Do not worry, since we have not been confused, we know that the M1 exists, but since this is nothing more than a rebranding of the A14X and therefore it was not originally designed to be used on a PC, the M1 Pro for its part itself has been designed to be used in a laptop like the new MacBook Pro.

The architecture of its components derives from that of the M1 which in turn derives from the Apple A14, so if we compare the CPU and GPU architecture with the also recently presented A15, not in number of cores, we will see that that of the latest iPhone is something more advanced. And what do we highlight compared to the standard M1? The fact that they have gone from a configuration of 4 performance cores and 4 of efficiency to a configuration of 8 + 2 cores and this is what represents the most remarkable change of all in terms of the raw power that this APU of Manzana.

Apple M1 Pro

The GPU on the other hand has gone from being 8 cores to 16 cores, which by eye would place it at an equivalent to an NVIDIA GTX 1660. We cannot say that it is a high-caliber graphics processor compared to dedicated PCs. like NVIDIA’s RTX 30 and AMD’s RX 6000, but no worse architecturally than the integrated Intel Xe-LPs from Intel’s Tiger Lake and AMD Vega we’ve seen so far in Ryzen APUs. You can easily say that in Cupertino they have an advantage in this regard, as they have the most powerful integrated GPU to date.

The other elements to highlight are a 256-bit LPDDR5 bus that gives it a higher bandwidth, 200 GB / s approximately, and an amount of RAM of 32 GB. It should be noted that again the RAM is soldered in the same interposer as the processor as the M1, thus preventing the ability to expand this section. Nor can we forget the Neural Engine in the face of tasks related to AI, which is at the moment the true spearhead against Intel and AMD as none of their processors for laptops have an equivalent unit on the market.

M1 Max, there is a bit of SoC in my GPU

Apple M1 MAX

The other APU that Apple has introduced is the M1 MAX, which is an even more vitaminized version of the M1 architecture, but it does not increase the number of CPU cores above those in the M1 Pro, especially because of the complications in infrastructure and interconnection, let’s not forget that neither Intel nor AMD are daring to unify more than 8 cores per cache group in their CPUs.

As you can see, the M1 MAX is dominated by a huge 32-core GPU or shader units and its memory interface. LPDDR5 has been expanded to 512 bits thereby doubling the bandwidth to the 400 GB / s, which means doubling the number of chips and reaching 64 GB of capacity. We have to emphasize that although it is not seen in the indicated lithography, it uses the same Neural Engine as the M1 Pro and therefore the changes are limited to doubling the bandwidth and the GPU.

To finish we have to say that we are surprised that Apple makes use of LPDDR5 memory for its M1 Max, especially due to the fact that they have patents where they talk about 2.5DIC configurations and therefore can use HBM type memory. Although we suppose that Apple would be reserving its use for an eventual M2 Max that we will see next year or who knows if in a couple of years.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *