New heights of performance and efficiency

Last week, Apple unveiled its new generation of MacBook Pro laptops, a new range of leading devices that bring with them significant updates to the company’s professional and advanced user base. The new devices are particularly different in that they are now powered by two new additions to Apple’s own silicone line, the M1 Pro and M1 Max. We covered the initial discovery in last week’s review article of two new chips, and today we get the first insights into the performance we expect from the new silicone.

M1 Pro: 10-core CPU, 16-core GPU, 33.7 billion transistors

Starting with the M1 Pro, the smaller brother of the two, the design seems like a new implementation of the first generation M1 chip, but this time designed from the ground up to increase larger size and better performance. In our opinion, the M1 Pro is more interesting than these two designs, because it offers basically everything that advanced users will consider generationally important in terms of upgrades.

At the heart of the SoC is a new 10-core CPU configuration, in an 8 + 2 configuration, with 8 Firestorm performance cores and 2 efficient Icestorm cores. We indicated in our initial reporting that Apple’s new M1 Pro and Max chips appear to use a similar, if not the same generation of CPU IP as on the M1, instead of updating things to the newer-generation cores used in the A15. It seems that we can confirm this, because we do not see any obvious changes in the core compared to what we found on M1 chips.

CPU cores run up to 3228 MHz, however, they vary in frequency depending on how many cores are active within the cluster, dropping to 3132 at 2 and 3036 MHz at 3 and 4 active cores. I say “per cluster” because the 8 performance cores in the M1 Pro and M1 Max really consist of two 4-core clusters, both with their own 12MB of L2 cache, and each can clock its CPUs independently of each other, so it is actually possible to have four active cores in one cluster at 3036MHz and one active core in another cluster running at 3.23GHz.

The two E-cores in the system run at 2064MHz, and unlike the M1, this time there are only two, however, Apple still gives them its full 4MB of L2 cache, just like on the M1 and A-derivative chips.

One big feature of both chips is their significantly increased memory bandwidth and interfaces – the M1 Pro has 256-bit LPDDR5 memory at 6400MT / s speeds, which corresponds to a bandwidth of 204GB / s. This is significantly more than the M1 at 68GB / s, and also generally more than competing laptop platforms that still rely on 128-bit interfaces.

We were able to identify “SLC,” or system-level cache as we call it, to drop to 24 MB for the M1 Pro and 48 MB for the M1 Max, a little less than we initially speculated, but makes sense given the SRAM matrix area – which represents an increase of 50% compared to SLC per block on M1.

M1 Max: 32-core Monstrosity GPU with 57 billion transistors

The M1 Max is the bigger brother of the two designs. From a fundamental perspective, it is essentially identical to the M1 Pro except in terms of GPU, as well as Apple’s advertising, they have doubled the media encoders.

The GPU and memory interfaces of the chip are by far the most diverse aspects of the chip, instead of a 16-core GPU, Apple doubles things down to a 32-core unit. On the M1 Max we tested for today, the GPU runs at up to 1296MHz – pretty fast for what we consider a mobile IP, but still significantly slower than what we saw on a conventional PC and console space where GPUs can now operate up to about 2.5 GHz.

Apple is also doubling the memory interfaces, using a huge 512-bit wide LPDDR5 memory subsystem – unheard of in SoC, and even rare among historically discrete GPU designs. This gives the chip a massive 408 GB / s of bandwidth – how this bandwidth is available to the various IP blocks on the chip is one of the things we will explore today.

The cache memory of the memory controller is at 48MB in this chip, which allows theoretically increased memory bandwidth for different SoC blocks, as well as reducing DRAM traffic outside the chip, thus reducing both the power consumption and the power of the chip.

Apple’s footage of the M1 Max was a bit odd at first because we weren’t sure if it actually represented physical reality – especially on the bottom of the chip that we noticed appeared to have a duplicate NPU – something Apple doesn’t do officially reveal. The doubled media mechanism makes sense because it’s part of the chip’s characteristics, however, until we get a third-party image to confirm that this is what the chip really looks like, we’ll refrain from further speculation about it.

Source link

Naveen Kumar

Friendly communicator. Music maven. Explorer. Pop culture trailblazer. Social media practitioner.

Related Articles

Leave a Reply

Your email address will not be published.

Back to top button