Why can Apple beat Intel and Qualcomm? M1 vs. Core i9/Snapdragon 8cx
Apple Mac started to switch to its own chip, which is undoubtedly a blow to Intel.
Apple Mac started to switch to its own chip, which is undoubtedly a blow to Intel. When Apple mentioned its M1 chip at last year's MacBook series product launch conference, it did not forget to talk about its CPU performance increased by 2.8 times compared to previous generation MacBook products; GPU performance increased by 5 times; the comparison is the Intel processor. The battery life of the equipment has been increased to 20 hours.(Find more in Regoo iphone malaysia)
These numbers are simply incredible in terms of the stagnation of Moore's Law in the semiconductor industry. Of course, the content and objects of comparison are debatable. However, this is not the first time Apple has used such comparison numbers. When the Mac was still using the PowerPC processor (1999), Apple mentioned in its publicity that the PowerPC G3 was twice as fast as the Intel Pentium II at the time. So familiar.
This so-called "2 times faster" statement can roughly see Apple's processor philosophy in the same line, even if Apple did not have its own chip design team at the time. In fact, the so-called 2 times faster at the time was when Apple compared Macs and Windows PCs at the press conference, using PowerPC G3 processors and Intel Pentium II processors, respectively. The content of the comparison was that both parties automatically performed a set of Photoshop tasks.
The Mac in the demo can complete these tasks in half the time of a Windows PC. The key to this is that PowerPC G3 adds an AltiVec acceleration unit, which is a 128bit vector processing unit that can perform 4-channel single-precision floating-point mathematical operations in a single cycle. Photoshop can make full use of this unit, and AltiVec is more efficient than Intel's MMX extension instructions at that time.
In simple terms, it relies on dedicated units to achieve performance and efficiency rolling. Although PowerPC has long been beaten by x86, this concept has largely continued to today's Apple chips (including A series and M series SoCs). The following will describe this from a technical point of view, which is reflected in the M1.
In fact, the appearance of the M1 chip not only makes the x86 camp feel the white terror, but also poses a great threat to participants in the Arm camp such as Qualcomm. Qualcomm has cooperated with Microsoft for two generations of chips (SQ1 and SQ2) products, and has launched Surface Pro X (and its Windows 10 on Arm) for two years, but it is not worth mentioning in front of M1.
This article is "Can Intel change the CEO to make chips like Apple M1?" This is a companion article to the article "PC Processors and Nuclear Counterattack". This article will discuss from the perspective of the M1 chip, how Apple technically achieves a deterrent against Intel, Qualcomm, and other competitors; and try to explore why Apple can To do this, and whether competitors like Intel and Qualcomm can do it. Except for the first two parts of this article, the rest can be read selectively according to interest.
In the past 5 years, the single-threaded performance of Intel processor products has increased by about 28%, while Apple CPU has achieved a performance increase of 198%. In other words, the single-thread performance of the A14 processor is nearly three times that of the A9.
The M1 and A14 chips share the Firestorm large core, and it can be roughly considered that the two have similar single-thread performance (M1 frequency is higher than A14, L2 cache is also larger, etc.). The performance of Apple's chips in 2020 has just formed a scissor-fork trend compared to Intel processors. So the latest MacBook started to use its own chips, it seems logical in terms of performance.
In fact, Apple declared during the A7 chip (iPhone 5s) period that its Cyclone core is a "desktop architecture." But not many people cared. And before the emergence of M1, more people still firmly believed that compared to x86, the Arm processor can only do low power consumption, but cannot achieve the high performance like x86. Apple has proved the error of this view with actual actions.
Many comparison tests mentioned that M1 is more powerful than Intel Core i9. Is this an exaggeration? Let's roughly sort out the performance level of the M1 chip compared to x86 and Arm's mainstream PC-oriented processors.
Officials have not announced the TDP (or approximate power consumption) of the M1 chip. AnandTech test believes that the CPU TDP of M1 is 20-24W. The test of Geek Bay basically confirmed that its CPU TDP is around 25W (peak power consumption is 24W, and daily peak power consumption is around 15W). This value may not be accurate, and it may include DRAM power consumption, but it is also inseparable.
According to Geek Bay data, the peak power consumption of the GPU part at full speed is about 10W; the peak power consumption of the entire M1 chip is about 34W. AnandTech mentioned in the test that the GFXBench Aztec High test had a power consumption of 17.3W. This test should hardly reflect how much GPU alone occupies. On the whole, such power consumption and TDP level are about the same as low-voltage processors from Intel and AMD.
What Apple said is several times the improvement over the previous generation MacBook, which seems to be completely absent. The SPEC2017 test results are similar to those of SPEC2006, but M1 and AMD Zen 3 (Ryzen 9 5950X) are mutually successful. The M1's integer performance is not as good as Zen 3 as a whole, and the floating-point performance surpasses.
This comparison is also based on the fact that the highest frequency of the M1 core is only 3.2GHz, while the core frequency of Intel's 11th generation Core (i7-1185G7) is 4.8GHz, and the core frequency of Core i9-10900K can reach 5.3GHz; AMD Zen 3 (Ryzen 9 5950X) has a turbo frequency of 4.9GHz. Compared with M1, the power consumption of the comparison object will be significantly higher.
In other words, Apple M1 has actually become the IPC king of CPUs today. IPC is the ability to execute instructions per cycle, or how many things can be done per Hz. It can be seen that M1 can get rid of the most powerful x86 processor.
In terms of multi-core performance testing, the M1 itself is limited by only 4 large cores and 4 small cores, which is naturally incomparable with the big guys with 8-core, 10-core and hyper-threading support. However, in terms of performance, M1 can still crush Intel's low-voltage version of the 11th-generation Core (i7-1185G7, 4 cores and 8 threads, turbo frequency 4.8GHz, TDP 28W). If M1 expands the number of cores and processor scale, it should not be difficult to achieve multi-threaded performance.
Other test results are roughly similar, including Geekbench 5 single-threaded test, M1 easily won the championship, super AMD Zen 3 (Ryzen 9 5950X) and Intel Core i9-10900K; CineBench R23 single-threaded performance is slightly weaker than AMD Zen 3, multi-threaded performance is better than AMD Ryzen 4800U (Zen 2, 8 cores and 16 threads) is about 15.6% weaker and 60% stronger than Intel's 11th generation Core (i7-1165G7, 4 cores and 8 threads).
In order to realize the compatibility of M1 chip and x86 software, Apple made Rosetta 2 intermediate translation layer. In this way, the previous x86 software does not need to be modified and can run on the M1 chip. This scheme will make the program execution efficiency lower. Even with Rosetta 2 translation, the performance of the M1 chip running traditional x86 programs can reach the level of Intel's eighth-generation Core (for programs that rely less on AVX instructions, the performance of M1+Rosetta 2 is similar to that of the tenth-generation Core).
But the summary is: in CPUs of the same core size, the Apple M1 chip can achieve higher performance than x86 processors with much lower power consumption when running conventional performance tests. Under the same scale, it is not a problem at all to completely flip Intel. Even if it has played back and forth with AMD, it is AMD at the expense of power consumption (in the single-core performance test, AMD Zen 3 can reach the power consumption level of 49W, while Apple M1 is the overall power consumption of 7-8W).