Apple A9 chip reach desktop level?

Starting with the A4 on IPhone4, Apple's self-contained a-series processor performance has been dramatically improved. At the Apple press conference this year, Apple was very rare to mention the performance of the A9 processor in the process of introducing the processor. Apple mentions that the performance of the A9 processor is 70% higher than the previous generation, comparable to desktop-level CPUs, while GPU performance is up 90% to host-level GPU performance. So, actually, is the A9 chip really that powerful?

A9 Performance Parameters

In the press conference, and the introduction of A7, A8 is different, this time Apple introduced A9 did not mention the specific parameters of the processor, just a brief introduction of the next A9 is the third generation of 64-bit processor, with the new M9 coprocessor, using a new transistor architecture, while the real application of the use of optimization, The specific optimization content is not mentioned. Chip technology, it should be the use of FinFET technology, but in the end is Samsung 14nm FinFET or TSMC 16nm FinFET process is unknown.

In addition to processor technology and architecture, people are most concerned about memory capacity issues, but according to past practice, Apple still does not give specific memory values. However, Apple's previous generation A8 processor is a second-generation 64-bit processor with a dual-core architecture paired with the M8 coprocessor and 1G of memory, while the A8X processor on the ipad Air2 uses a three-core architecture and is equipped with 2G of memory for the first time. Apple said in an introduction to the ipad Pro that A9X is 1.8 times times more powerful than A8X and is equipped with 4G of ultra-large memory. Presumably, the A9 processor overall is not worse than a8x, but less than a9x, so should also be equipped with at least 2G of memory, and the core number is likely to reach the level of four cores.

A9 can a series of chips reach the desktop level?

In the current market based on the arm version of the CPU, the fastest is Samsung's Exynos 7420, relying on 14nm technology pull high frequency, in the use of ARM A57 core, Samsung put Exynos 7420 Geekbenche 3 single-threaded runs to 1400-1500 points, which is currently the highest performance of the public version of the architecture.

Apple, from A6, is no longer the same as the previous arm-based cortex-a9 modification, but uses its own development of the SWIFT CPU architecture, although the overall remains an ARM architecture processor, but it is neither A9 nor A15. Relying on a high level of design, Apple A8 processor, in the use of 20nm backward process, the Geekbenche 3 single-threaded running points to achieve more than 1600. And this time the A9 processor is A8 70%, that is the level of 2700 points, this performance is quite scary.

It is worth mentioning, however, that the Geekbench 3 test because there are some test items optimized for the Apple processor, so the running points are not completely credible. More realistic tests are SPCEint2000, specfp2000,spceint2006, SPECfp2006, which are less easily optimized. According to this test A8 about 1.67Ghz of the Core 2 level, then the A9X single-threaded level should basically reach the level of the first generation i3, and will not go straight to i7. As a result, the A9 series of chips barely counted on the desktop level, but Cook said the desktop computer beyond 80% is a bit "violating advertising law" feeling.

ARM not comparable to X86

Currently, the CPU of the ARM architecture occupies the mobile market, while the X86 architecture occupies the PC side. The two camps are constantly competing on performance, but not in the actual end market. So, simple performance beyond is a more one-sided argument.

ARM has been an advance in the past few years, but has never been in the field of PC and other X86, after all, their design starting point is not the same, so there is not much comparability. X86 is not capable of arm power, and arm is difficult to achieve X86 performance. This is why atom has been unsuccessful for a long time. Intel tries to use its shortcomings to confront other people's strengths, and the result is naturally not very good, but if Intel had the most advanced semiconductor technology on the planet, Atom would never have appeared. and arm if try to and X86 to spell performance, that result naturally also not where to go.

Back to the Apple A9 processor, although always limited by the production process, Moore's Law and instruction set differences and other reasons, the performance of the arm structure in a short time is still difficult to go beyond the traditional X86 structure of the processor. But the performance of Apple's A-series chips is evident in similar products, and Apple's 64-bit trend in 2013 led the Android camp to catch up for a year. Over the next few years, Apple's processors may be in the realm of ARM architecture processors.

Summarize

Apple's strong performance on mobile processors is commendable, although ARM processors cannot be used in PC terminals such as Macs in the short term, but perhaps one day Apple is really studying the possibilities of the ARM architecture pc.

Apple A9 chip reach desktop level?