Science Technology

Without the launch of cars and VR, did Apple encounter the failure of Moore’s Law?

In the early morning of June 7, Apple WWDC2022 started on time. There is no long-awaited VR, no latest progress in car building, and only the M2 chip and the new CarPlay are the highlights.
On the one hand, Apple has picked up CarPlay, which had a low sense of existence before, and built it into a “mobile center console” that can be compatible with most cars. It seems to have become Apple’s stopgap in the smart car circuit.
On the other hand, the most interesting is the M2 chip, which uses 8-core CPU+10-core GPU, integrates more than 20 billion transistors, which is 25% higher than the M1 chip, increases the CPU speed by 18%, and improves the GPU performance by 35%.
Looking at the data alone, the performance of the M2 chip is also a surprise. Two computers with the latest chips, MacBook Air and MacBook Pro, also appeared one after another. With the further improvement of the computing power of the M2 chip, Apple began to turn its attention to the game book business, which has been neglected for years.
However, in terms of the increment of chip energy efficiency, the learning ability of M1 chip neural network is 15 times that of the previous chip; 3.5 times the performance of the previous chip; In contrast, the growth of M2 chips appears to be slow.
It seems that Apple has also fallen into the barrier of the failure of Moore’s Law. In this case, how can Apple face the smart car with higher computational power requirements?
87% performance with 1/4 energy consumption
After half a year’s delay, the M2 chip still came.
As expected, the M2 chip still uses TSMC’s 5nm process technology, and the 4nm chip that has been expected has not been realized. However, the enhanced M2 performance has improved a lot.
Without the launch of cars and VR, did Apple encounter the failure of Moore’s Law?
First, M2 chip integrates more than 20 billion transistors, 25% more than M1 chip; CPU performance is 18% higher than M1; On the memory controller, the M2 chip can provide a unified memory bandwidth of 100GB/s, 50% more than the M1 chip.
In addition, due to the small memory of M1 before, there will occasionally be bottlenecks when running high-performance work. This time, M2 chip is equipped with 24GB of unified memory, which can support processing larger and more complex workloads.
Figure: Apple M1/M2 chip comparison
On the chip path, Apple has always attached great importance to “energy consumption ratio”, that is, to minimize power consumption while maximizing performance. This time, the M2 chip has also achieved this point to the extreme.
According to the data disclosed by Apple at the conference, compared with the latest 10-core PC notebook chip, the CPU of M2 chip can improve the performance by 1.9 times under the same power consumption.
Compared with the latest 12-core chip, M2 chip can provide 87% peak performance using only a quarter of the power consumption of PC chip. In contrast, the latest 12-core PC notebook chips require a significant increase in power consumption to improve performance.
These data show the unique advantages of M2 chip in energy consumption, faster performance core speed and larger cache; The energy efficiency core has been significantly enhanced and can bring greater performance improvement. Together, they can easily complete tasks requiring heavy CPU use with very little power consumption.
Based on this advantage, Apple computers with M2 chips can be thinner, quieter, less prone to burn, and have longer battery life than those with ordinary PC chips.
In addition, M2 chip is also equipped with Apple’s next-generation graphics processor, which has up to 10 cores, two more than M1 chip.
The 10-core graphics processor integrates larger cache and higher memory bandwidth, which can improve graphics performance. At the same power consumption level as M1 chip, the graphics performance of M2 chip can be improved by up to 25%. At the maximum power consumption level, the performance can be improved up to 35%.
By contrast, under the same power consumption, the graphics processor of the M2 chip can reach 2.3 times the performance of the integrated graphics card in the PC notebook chip at the fastest, and the M2 chip can achieve the peak performance of the PC chip with only one fifth of the power consumption.
At the same time, Apple also introduced the latest customization technology into the M2 chip, which is equipped with the latest generation of security partition and neural network engine. The neural network engine in the M2 chip can process 15.8 trillion operations per second, 40% faster than the M1 chip.
And the M2 chip also makes up for some shortcomings of M1. For example, the video output of M1 only supports one 4K (60hz) display, and the Inter-series can generally support at least two 4K (60hz) displays.
So this time, Apple has equipped the M2 chip with ProRes video processor for hardware-accelerated encoding and decoding, which also allows the system with the M2 chip to play multiple 4K and 8K videos.
It can be said that after the comprehensive upgrade, M2 chip has further improved the excellent performance, function and energy efficiency of M1 chip. But is there a qualitative leap in M2 chip?
Because if compared with the previous chip, the learning ability of M1 neural network is 15 times that of the previous chip, and the performance is 3.5 times that of the previous chip; In terms of horizontal comparison, Intel’s Alder Lake and AMD Zen3+increase one CPU by more than 50% and double one GPU directly.
So by comparison, we will find that the performance improvement of Apple chips has also slowed down. However, the M2 chip has not achieved the expected performance leap. It is more about the improvement of energy consumption ratio, such as the enhanced version of M1.