Home > News content

Apple M1 performance reviews out power and speed x86 competitors

via:雷锋网     time:2020/11/18 20:51:55     readed:69

As a chip driven by arm, apple M1 has attracted a lot of attention, especially for many fruit powder. Therefore, we have made some typical simultaneous interpreting of these devices and their ecology, and carefully studied the performance of the new M1, compared with the traditional x86.

visit:

Ali Yunshuang 11 global carnival season to continue-double-core 8 G cloud server first 286.44 yuan

Apple online store (China)

M1 CPU is 5nm 8-core design, with 4 performance cores and 4 efficiency cores. Among them, the user-centered foreground tasks with low latency will run on the performance core, and the background tasks with low latency requirements will run on four efficiency cores.

With the exception of eight CPU kernels, the M1 in the Mac Mini also has eight GPU kernels, with a total of 128 execution units, despite the "AppleAccurate benchmarking VS non-Apple "is very difficult, but I am confident that it is indeed the world's leading design. There may be faster than M1 in existing CPU, but if power consumption is constant, there is no computer or server CPU. to compete with Similarly, high-end Nvidia or Radeon desktop cards may be used to beat M1 GPU,, but there are huge differences in power consumption, physical size, and heat.

The expansion of arm architecture

Desktop computers and traditional laptops are the last bastion of 64 bit x86 architecture, in which performance and the ability to run classic operating systems and software stacks are the most important evaluation criteria. In recent years, there are more and more desktop computers based on ARM architecture, although they often run slowly and most of them are in the low-end market, such as the pinebook Pro device.

However, Apple's new M1 SOC is definitely not one of the low-end products. M1 was designed from the beginning as a powerful and competitive product compared with traditional PC architecture.

How about M1?

Geekbench 5.3.0

We try to compare M1 directly with its competitors based on x86 architecture. In our equipment evaluation, we usually rely on the general comprehensive benchmark test suite, which can carry out various tests on the platform, and finally generate an intuitive score. Unfortunately, not all benchmark packages can run on MacOS, few packages can run on Apple silicon, and almost no suite can run on MacOS 11 of Apple silicon.

However, geekbench 5.3.0 is an exception, and its new version already runs locally on Apple silicon MacOS..

In geekbench's tests, M1 beat its competitors with the highest scores, whether single threaded or multi-threaded. In the OpenCL GPU test, M1 still beat i7-1185g7 and ryzen 7 4700u with a score of 1948 2.

Cinebench R23

The latest version of cinebench R23 provides support for the arm version of MacOS, which is generally more popular than geekbench. Although some people criticize that its use of Maxon graphics rendering software is too narrow, we think that it can not only distinguish the CPU gap, but also be closer to the real world expectation and passmark general benchmark than geekbench.

In the cinebench test, apple M1 easily beat the 4-core 8-thread i7-1185g7 and 8-core 8-thread ryzen 7 4700u with 7804 points, while the score of 8-core and 8-thread ryzen 9 5950x reached 11850, far higher than M1. However, it is worth noting that only four of the eight cores of M1 are high-performance versions of firestorm, and the power consumption (TDP) of 5950x is more than three times that of the whole mac mini system.

In the next single thread test, M1 scored 1520 points, almost equal to Intel's i7-1185g7, and ryzen 9 5950x surpassed M1 with 1625 points. If the world's leading single threaded x86 CPU ryzen 9 5950x and M1 are limited to four threads, and running cinebench r235950x again with the four high-performance cores of M1, M1 performance will be improved by 8.3%.

Game and browser testing

A browser benchmark is a test that can translate well between different architectures, because it can measure the performance of relatively real tasks, such as complex operations, in web browsers. Although benchmarks such as jetstream 2.0 and speedometer are still comprehensive, they can simulate the world operations that each user expects to work and perform some detailed tasks.

Through the test browser, the Mac Mini driven by M1 performs well. When Safari is used on Apple silicon, the score is higher than Acer swift 3 driven by ryzen 4700u and x86 Google Chrome with Rosetta. However, it is possible that these test results can not be directly compared with the actual browsing experience. They are all very fast running devices, which will be more fluent on the web and other places, and the mini and its M1 arm architecture are not slow.

Mac Mini's M1 processor with the latestiPadandiPhoneA12 Z and A14 bionic share the arm architecture, while Apple provides the slingshot extreme mobile game test suite of 3D mark for these devices in its app store. As can be seen from the chart, M1 performs best. If you want to play mobile games on the Mac Mini, you will get a first-class experience as long as the app adapts to the computer.

power waste

The way to test the power consumption of ryzen7 4700u and mac mini is different.

For the Mac Mini, it's easy to test the power consumption of the entire system, plug it into the Kill-A-Watt meter and watch the display. When testing Acer swift 3 driven by ryzen 7 4700u, the idle state and running status of pigz-p1, pigz-p4 and pigz-p8 were tested. The Linux utility powertop was used to run each workload repeatedly until the battery discharge was stable to no more than 0.1W fluctuation.

It should be noted that this comparison is not particularly fair. Acer swift 3 must power the display, while the Mac Mini uses separate displays. However, it is still clear that the M1 driven Mac Mini still consumes less power than swift.

Upgrade of competition between arm and x86

Although there are some limitations in trying to benchmark the new architecture on a few common platforms, it is obvious that M1 does combine high performance with high efficiency. The Mac Mini tested here can outperform some high-performance mobile CPUs when its power consumption and heat dissipation characteristics are not limited.

However, M1 is not very powerful either. If competitors bring more cores and threads into large-scale parallel competition, M1 may be defeated. However, competitors can only do so when the power and heat differences are serious, and there may be huge differences in manufacturing costs.

In addition to being confused about the large-scale parallel workload, Apple has obviously broken through the limitations of high-performance arm desktop and notebook design, that is, it can build an arm system that can compete with X86 even at high performance level.

It should be pointed out that this is only the first generation of Apple's arm laptop, and there is still a lot of room for growthMac ProThe subsequent design of may have eight performance cores instead of four.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments