How many cores does a smartphone have? This is one of the most common questions we ask when comparing two smartphones. The number of CPU cores is indeed an important indicator of smartphone performance, but it is not the most accurate one. At the end of last year, Qualcomm acknowledged that the octa-core chips were merely a marketing gimmick. Judging chip performance solely by the number of cores is almost meaningless. Many people find this laughable, but if we think calmly, if the number of CPU cores does not directly correlate with smartphone performance, then what determines smartphone performance?
The child who speaks the truth in “The Emperor’s New Clothes”
At the end of 2015, Qualcomm released the Snapdragon 820. At a time when the industry was madly pursuing the number of smartphone processor cores, the 820 had 2 high-performance Kryo cores and 2 power-saving Kryo cores, totaling just four cores. The Snapdragon 820 faced controversy upon its debut. When Qualcomm admitted that octa-core chips were merely a marketing gimmick, it sparked a massive upheaval in the consumer market. Since the data transmission rate has a significant impact on user experience, it has become a better way to test smartphone performance. Of course, regarding this processing solution, we hope Qualcomm can provide more data to convince consumers.
CPU is important but not everything
CPU (Central Processing Unit) is the most familiar parameter to everyone. From the name, we can see its importance to the chip. Many people even equate it with the entire chip. In reality, the more accurate term for smartphone chips is System on Chip (SoC), and the CPU is just one part of it, responsible for controlling calculations and directly determining the smooth operation of the smartphone.
When it comes to judging the performance of a CPU, ordinary consumers are often misled by various shameless marketing strategies and salespeople with half-knowledge, focusing only on how many cores there are and how high the clock speed is. However, these are only part of the final performance. The most critical determining factor is architecture!
Currently, mainstream SoCs, regardless of their origin, almost all use ARM’s Reduced Instruction Set Architecture. The difference is that capable manufacturers only use ARM’s ARMv7/v8 instruction sets and conduct their own architectural designs, such as Apple and Qualcomm being typical representatives. Since the A6, Apple has started to design its own CPU architecture. A year later, the A7, based on the Cyclone architecture, became the first 64-bit mobile processor, leading the industry by a year; Qualcomm established its position as a leader in the 32-bit era with the Krait architecture based on the ARMv7 instruction set, and the latest Snapdragon 820 is based on ARMv8 to create the new Kryo architecture.
Another approach is to directly purchase and use ARM’s designed public version architecture. The commonly seen Cortex-A7/A8/A9/A53/A57/A72 are the architecture names designed by ARM. Among them, there are 32-bit ARMv7 instruction sets and 64-bit ARMv8 instruction sets. MediaTek and HiSilicon adopt the public version scheme.
To give a vivid example, the first scheme is like ARM laying a foundation, and manufacturers can freely choose how to build the house. The advantage is strong flexibility, and performance and power consumption are often better controlled than public version architectures, but it requires high demands in terms of money, time, and technology. The second scheme is equivalent to ARM not only laying the foundation but also drawing the blueprints. Manufacturers only need to follow the blueprints for construction after purchasing, greatly reducing the development time and cost of the entire SoC chip.
Different architectures directly determine the performance basis. For example, the processors commonly found in thousand-yuan machines on the market, although they have eight cores and a clock speed exceeding 2GHz, often use the low-power A53 architecture, and their performance is even inferior to that of dual-core high-performance A72 architecture, especially in single-threaded computing capabilities. This is why Apple has consistently chosen dual-core processors but has always maintained a significant performance lead. In fact, this is similar to computer processors. Intel’s famous Tick-Tock strategy focuses on alternating improvements in architecture and process technology in different years, ultimately achieving annual performance increases. Therefore, besides architecture, process technology is also crucial.
Process technology refers to the distance between circuits within a chip. The current mainstream processes are 28nm and 20nm, with the most advanced being 16/14nm. Advanced processes can reduce processor power consumption and heat, shrink chip size, and enhance performance. Understanding this, we can see why the Snapdragon 810 with a big.LITTLE architecture of 4 A57 + 4 A53 cores, built on a 20nm process, suffers from heat and power consumption issues, while Samsung’s Exynos 7420 built on a 14nm process performs excellently.
If we go into more detail, semiconductor process technology can also be divided into 2D structure MOSFET (Metal-Oxide-Semiconductor Field-Effect Transistor) and 3D structure FinFET (Fin Field-Effect Transistor). The FinFET architecture mainly modifies the gate that controls the current, significantly improving circuit control and reducing leakage current.
Even with the same process technology, iterations may arise due to technological improvements. For example, last year’s controversy over Apple’s A9 foundry: theoretically, the more advanced 14nm process did not perform as well as TSMC’s 16nm process. The reason is that Samsung jumped directly from 28nm to 14nm LPE technology, while TSMC transitioned step by step from 28-20-16nm, making the technology relatively mature, with better yield and leakage control. However, Samsung has clearly learned from its lessons, recently launching the second generation of 14nm LPP technology, which improves performance by 15% and reduces power consumption by 15%. TSMC also launched an improved version of 16nm FinFET+ this year. Therefore, seemingly identical processes can also differ due to specific versions.
After the above discussion, I believe everyone has a general understanding of how to judge the quality of a CPU. To summarize, architecture and technology are more important than the number of cores and clock speed, and high-performance architectures greatly outperform low-power architectures, while more advanced processes lead to stronger processors.
What Determines Gaming Performance is the GPU
A good CPU is just the first step in creating a high-end chip. While it ensures smooth and stable operation, in today’s world where mobile entertainment is becoming increasingly important—especially as young people prefer to use smartphones instead of handheld consoles for gaming—the GPU is undoubtedly a crucial component that cannot be ignored.
The GPU, short for Graphics Processing Unit, relates to the strength of graphic rendering capabilities, directly determining whether games can run smoothly. Currently, the main mobile GPU manufacturers are ARM, Qualcomm, Imagination Technologies (hereinafter referred to as Imagination), and NVIDIA. Qualcomm’s Adreno and NVIDIA’s Maxwell GPUs are only used in their Snapdragon and Tegra chips, so only ARM Mali and Imagination PowerVR series are sold externally.
Although there are often rumors about Samsung and Apple developing their own GPUs, so far, all chip manufacturers can only purchase ARM or Imagination’s solutions to integrate into their chips. For example, Huawei HiSilicon and Samsung Exynos use the ARM Mali series, while Imagination PowerVR is more commonly used in Apple’s A series and MediaTek chips.
Due to the different architectures adopted by different manufacturers, such as unified rendering and separate rendering, it is not possible to judge the performance of mobile GPUs solely based on architecture or cores. The industry usually looks at their triangle output rate and pixel fill rate. Of course, since various evaluation software has emerged, comparisons have become more intuitive. If everyone can’t remember various rankings, it doesn’t matter; understanding the naming rules of each manufacturer will give you a general idea.
Generally speaking, the naming of mobile GPUs consists of letters and numbers, such as Adreno 530, Mali-T880, PowerVR GT7600, etc. The first digit usually indicates the generation, with higher numbers indicating newer generations; the second digit represents positioning, with higher numbers indicating stronger performance. The three GPUs mentioned above are already the latest high-end models in commercial use. However, ARM’s Mali GPU is a bit special; in addition to looking at the model, we also need to pay attention to the number of cores (the last part of the name MPx, where x indicates the number of cores). For example, although both Kirin 950 and Exynos 8890 use the latest Mali-T880, Kirin 950 only has 4 cores while Exynos 8890 has 12 cores, so their performance is naturally incomparable.
Making Calls Relies on Baseband Performance
The most core function of a smartphone is still communication. In this regard, no matter how powerful the CPU and GPU are, they are powerless; ultimately, it depends on the baseband.
The baseband directly determines what kind of network standards the smartphone supports. When we make calls, browse the internet, or send messages, all these actions are executed through the upper processing system issuing commands to the baseband part, which processes and executes them. It can be said that the baseband is the most technologically advanced part of the entire chip. Many manufacturers, like Texas Instruments, STMicroelectronics, and NVIDIA, have been eliminated from the market because their SoCs do not integrate baseband or their technology is outdated.
The most intuitive way to judge the quality of a baseband is to look at how many network standards and frequency bands it supports. Generally, smartphones claiming to support all networks use high-end basebands. From 4G LTE to 3G WCDMA/CMDA/TD-SCDMA, and down to 2G GSM/EDGE, all standards are supported, allowing users to simply purchase a smartphone and insert a card to use without worrying about compatibility or card replacement issues.
In addition to standard support, another important feature of basebands in the 4G era is the UE access capability level of LTE, which indicates the transmission rates the UE can support, usually represented by Cat x, where the larger x indicates higher upload and download speeds. Unfortunately, the Kirin 950 uses an older baseband Balong 720 that only supports Cat 6, rather than the latest Balong 750 that supports Cat 12/13.
Multimedia Performance Depends on SoC’s Overall Performance
Multimedia performance itself is divided into many parts. First is the video encoder, which determines how many formats the smartphone can support for encoding and decoding. For example, ARM’s Mali-V550 and Snapdragon 820 both support 4K resolution H.265 HEVC video encoding and decoding. Some SoCs also integrate audio decoders, such as MediaTek.
Next is the display controller, which affects the final resolution and frame rate that the smartphone can support. Just a few days ago, ARM released the Mali-DP650, which can easily support 4K resolution at 60FPS frame rate screens, and can also be used on 4K TVs without any issues.
Lastly, there is the image signal processor (ISP) responsible for processing the data returned by the image sensor, affecting various basic and advanced imaging functions such as white balance, focus, exposure, noise reduction, face recognition, and motion compensation. The most famous solution in this field is Fujitsu’s Milbeaut, favored by Samsung and Smartisan. However, adding an ISP separately can increase power consumption, heat, and design complexity, so manufacturers prefer to integrate ISP into the SoC, such as Qualcomm, HiSilicon, and Apple’s chips, which are all integrated self-developed technologies.
Interesting Reading
Why Do Smartphone Processors Not Have Odd Cores?
In fact, not only smartphone processors but also computer processors generally have even cores. Occasionally, some processors with odd cores are produced, but they often involve masking one core at the hardware level, forming a 3-core processor. This isolated core is used for standby to reduce power consumption, and this is essentially still a 4-core processor.
From a technical perspective, chips are generally constructed in a rectangular manner. During manufacturing, they need to be cut in both horizontal and vertical directions. Rectangles are naturally occurring geometric shapes that are orthogonal lines. When the circuit scales up, this natural shape helps people find regularities and control the circuit among vast electrical signals. When designing processors, a core is usually drawn out and mirrored several times to create multiple cores. This includes the control circuits and clock circuits being completely symmetrical, which leads to the timing of multi-core processors being identical.
Moreover, a processor is square, and the cores are generally also square. For example, cutting a square into equally sized halves or quarters is easy to coordinate. However, cutting it into an odd number of equal parts is challenging. This is one possible reason why cores tend to be even. Of course, some native odd-core processors do exist, and this is also due to the consideration of balancing power consumption and performance. Additionally, processor manufacturers often encapsulate two cores together to form a module. Chips are composed of these modules, which leads to another possibility of chips being even-numbered.
The second possibility is due to factors such as cost, market demand, and performance.