After months of rumours and speculation, Apple has finally revealed the specs of its in-house custom ARM processor that will power future Macs — well, at least most of the important ones, like core count and number of transistors. Unfortunately, those specs don’t paint a complete picture of how the M1’s performance will actually compare to other laptop processors currently on the market today, especially without hard clock speed numbers. But there are a few things we can gleam from Apple’s event today, and the M1 is shaping up to be an interesting chip that could hold its own against Intel and AMD.
The M1 is a system-on-chip (SoC) processor, which means the CPU, GPU, and RAM are all contained on a single chip, just like Intel’s 11th-gen Tiger Lake mobile processors and Apple’s A14 in its iPhone 12. This allows each component to communicate more efficiently for faster processing times and lower latency, and also reduces power consumption. The M1 chip is optimised for low-power systems, so that’s a good thing, especially since some of the first products that will have the new M1 chips are the new MacBook Air and Mac Mini — two computers that need to rely on lower power consumption because there isn’t much room for cooling, not to mention the fact that the new MacBook Air doesn’t even have a fan.
The rumours were correct about the M1 being built on a 5nm node, which is a smaller node than AMD’s 7nm chips and Intel’s 10nm and 14nm chips. In addition to SoCs being more power-efficient, smaller nodes (transistors) are more power-efficient as well. They can do more calculations per clock without getting too hot, which is one of the main factors in limiting CPU performance. The more transistors on a CPU, the faster things go — twice the amount of data, or more, can be processed at the same clock rate with more transistors. The one in the M1 has 16 billion transistors; by comparison, AMD’s Zen 2 7nm architecture has 3.9 billion, and Tiger Lake has 5 billion.
Another thing that would increase the M1’s performance is raw core count. The M1 was rumoured to have 12 total CPU cores, with eight dedicated to high-performance workloads and four slower cores for tasks that don’t push the system nearly as much (think video transcoding versus typing in Word). Those rumours weren’t entirely correct. The M1 will have a total of eight cores, but four will be for performance workloads and the other four will be for high-efficiency workloads. But Apple didn’t reveal the frequencies of said cores, just claimed it has much better performance and lower power consumption than the “latest PC laptop chip.”
If we’re going by transistors alone, then sure, the M1 has AMD and Intel beat. But ultimately the number of transistors is related to the overall objectives and design style of the CPU itself, so a chip with less transistors could still be more powerful depending on the rest of its design. So what in the heck PC laptop chip is Apple comparing the M1 to?
[referenced id=”1234012″ url=”https://gizmodo.com.au/2020/07/these-new-apple-arm-benchmarks-might-not-be-what-people-think-they-are/” thumb=”https://gizmodo.com.au/wp-content/uploads/2020/07/25/cgxtmczxl2ywhfpoqlvg-300×168.jpg” title=”These New Apple ARM Benchmarks Might Not Be What People Think They Are” excerpt=”New leaked benchmarks that supposedly show Apple’s developer transition kit running Geekbench 5 Pro natively on the Mac mini are starting to make the rounds — meaning unlike the benchmarks that had leaked previously, which showed Geekbench 5 running virtually via Rosetta 2 on the transition kit, someone has supposedly…”]
If we look at the latest versions of the Mac Mini, MacBook Air, and MacBook Pro, both the Mini and Air were both Intel 8th-gen. The Pro was 10th-gen. I’m going to assume Apple was talking about something like the Intel Core i7-1068NG7 in its 13-inch MacBook Pro, one of its mid-2020 models. In this processor, Intel uses something called hyper-threading, which doubles the thread count in comparison to the total number of cores. This increases the performance so each core can run two instruction streams at once instead of only one, so it processes twice as much information. Intel also implements something called clock scaling on all its cores, which automatically adjusts to the workload being performed.
The M1 does this, too, but since it’s an ARM SoC, typically one half of the chip will be active depending on what kind of process the computer is running, and it will switch between halves on the fly. So instead of taking advantage of all eight cores like the i7-1068NG7 does with core scaling, the M1 will be switching back and forth between the four “big” cores and the four “little” cores to adjust for changing workloads and conversing power when needed. They’re different approaches to solving the same problem, but without knowing the clock speeds of the M1 it’s still tough to make a definitive comparison between the M1 and the “latest PC laptop chip.”
The M1’s neural engine might give it an advantage, though. Like Apple’s A14 Bionic chip for iPhones, the M1 will have a 16-core neural engine on the chip itself, which makes neural networks and machine learning more energy-efficient than running those sort of tasks on the CPU or GPU alone. But not every task needs that sort of thing.
We now have a better idea of how powerful the M1 will be, but we still don’t have all the details. Only real-life testing will tell if the M1 lives up to all the claims Apple made today. This is a huge moment for Apple, and for the CPU world in general. We’re all holding our breath to see what happens.