This month, Apple launched the M2 chip – the second-generation customized silicon chip for the Mac. After the M1 chip revolutionized the Mac beginning in November 2020, how a lot better actually is the M2?
Apple says that the M2 chip takes the efficiency per watt of the M1 even additional with an 18 p.c sooner CPU, a 35 p.c extra highly effective GPU, and a 40 p.c sooner Neural Engine. There are additionally different important enhancements corresponding to extra reminiscence bandwidth and assist for as much as 24GB of unified reminiscence.
Because the second main iteration of Apple silicon, it will not be clear how important of an enchancment the M2 is over the M1, particularly because it sticks with a 5nm fabrication course of, opposite to some rumors that prompt a larger efficiency and effectivity leap could be in retailer. The M2 chip debuts within the 13-inch MacBook Professional and the redesigned MacBook Air, so the extent to which the M2 is healthier than its predecessor might be an vital consideration when buying a brand new Mac. Learn on to study extra concerning the variations between the 2 chips.
- Made utilizing TSMC’s 5nm course of (N5)
- 16 billion transistors
- 4 high-performance “Firestorm” cores
- 4 energy-efficient “Icestorm” cores
- 3.2GHz CPU clock velocity
- CPU cores first seen within the iPhone 12 lineup’s A14 Bionic chip
- 8-core GPU
- Assist for 8GB or 16GB unified reminiscence
- 68.25GB / s reminiscence bandwidth
- Neural Engine
- Media engine for hardware-accelerated H.264 and HEVC
- Video decode engine
- Video encode engine
- Picture sign processor (ISP)
- Made with TSMC’s enhanced 5nm course of (N5P)
- 20 billion transistors
- 4 high-performance “Avalanche” cores
- 4 energy-efficient “Blizzard” cores
- 3.49GHz CPU clock velocity
- CPU cores first seen within the iPhone 13 lineup’s A15 Bionic chip
- 10-core GPU
- Assist for 8GB, 16GB, or 24GB unified reminiscence
- 100GB / s reminiscence bandwidth
- 40 p.c sooner Neural Engine
- Media engine for hardware-accelerated H.264, HEVC, ProRes, and ProRes RAW
- Greater-bandwidth video decode engine
- Video encode engine
- ProRes encode and decode engine
- “New” picture sign processor (ISP)
Fabrication Course of and Transistors
Just like the A14 Bionic, the M1 chip is constructed utilizing TSMC’s first-generation 5nm fabrication course of. Then again, the M2 makes use of TSMC’s second-generation 5nm course of just like the A15 Bionic chip. The M2 provides 4 billion further transistors for a complete of 20 billion – 25 p.c greater than M1. The improved 5nm fabrication course of is on the coronary heart of lots of M2’s efficiency and effectivity enhancements.
The M1 and M2 each have 4 high-performance and 4 energy-efficient cores, however whereas the M1 options “Firestorm” and “Icestorm” cores from the A14 Bionic chip, the M2 presents “Avalanche” and “Blizzard” from the A15 Bionic chip. In line with Apple, this ends in 18 p.c larger multithreaded efficiency than M1.
In early Geekbench benchmarks, the M2, which runs at 3.49GHz in comparison with 3.2GHz for the M1, earned a single-core rating of 1,919, which is roughly 12 p.c sooner than the 1,707 single-core rating of the M1 13-inch MacBook Professional. The M2 earned a multi-core rating of 8,928, up about 20 p.c from the 7,419 rating of the M1 mannequin. That is proper according to Apple’s declare that the M2 chip is as much as 18 p.c sooner than the M1.
Each chips have high-performance cores with 192KB of L1 instruction cache and 128KB of L1 knowledge cache. The energy-efficient cores have a 128KB L1 instruction cache, 64KB L1 knowledge cache, and a shared 4MB L2 cache. The one distinction right here is that the shared L2 cache is bigger on the M2 chip – 16MB as a substitute of 12MB on the M1.
The M2 options two extra GPU cores over the M1, leading to a reasonable increase in graphics efficiency. Apple says that the M2 has as much as 25 p.c larger graphics efficiency than M1 on the identical energy degree, and as much as 35 p.c higher efficiency at its max energy. Within the early Geekbench Steel benchmarks, the M2 chip scored 30,627, a notable enchancment over the 21,001 rating earned by the M1.
Each the M1 and the M2 have devoted video encode and decode engines for hardware-accelerated H.264 and HEVC, however the M2’s video engines are additionally capable of speed up ProRes and ProRes RAW to allow playback of a number of streams of 4K and 8K video. As well as, the M2’s media engine features a higher-bandwidth video decoder, supporting 8K H.264 and HEVC video.
The M1 and the M2 are available in configurations with 8GB or 16GB of unified reminiscence, however the M2 provides an extra, top-tier 24GB configuration. The M2’s reminiscence controller also can ship 100GB / s of unified reminiscence bandwidth, a considerable enchancment over the M1’s 68.25GB / s reminiscence bandwidth.
M2 options enhancements to a number of of Apple’s customized silicon applied sciences. For instance, the Neural Engine can course of as much as 15.8 trillion operations per second – over 40 p.c greater than M1. The M2 additionally comprise Apple’s newest Safe Enclave and a brand new picture sign processor (ISP) for higher picture noise discount.
Total, the M2 chip presents reasonable enhancements over the M1, even when most M1 customers are unlikely to note important enhancements when upgrading to the M2. The M2’s developments are typically worthwhile, if not transformative, and the chip is for certain to supply a extra up-to-date expertise with lower-spec Apple silicon machines – particularly for these coming from an Intel-based machine.
Whereas the M2 presents enhancements throughout the board due to its enhanced 5nm course of, up to date cores, and extra GPU cores, the foremost upgrades come to customers who must work with video, in addition to those that have memory-intensive workflows. The M2’s higher-bandwidth video decoder and devoted engine for ProRes and ProRes RAW video provide significant enhancements for video editors, whereas the 24GB unified reminiscence tier and 100GB / s of reminiscence bandwidth considerably increase the capability of M2 machines to cope with memory-hungry functions and intense multitasking.