Tech enthusiasts love to talk processing power and chips, be it from PCs and gaming consoles to the latest smartphones. We do a fair bit of it here at Android Authority, with in-depth coverage of the latest processors from Arm, Huawei, Qualcomm, Samsung, MediaTek, and others. For the uninitiated, these topics are dense with jargon and abstract-sounding ideas that can feel like a brick wall to understanding even basics questions like “what is an SoC?”
It can take years of study to properly wrap your head around the finer details of chip design, which is no good if you’re simply trying to research a potential purchase. Today, we’re going to do something a bit more beginner-friendly and explain the ins-and-outs of modern smartphone chips with as little technical witchcraft as possible.
Just what is an SoC?
SoC stands for System-on-a-Chip. As the name suggests, an SoC is a complete processing system contained in a single package. It isn’t a single processor chip package, which you might be familiar with if you’ve ever built a PC. Instead, an SoC contains multiple processing parts, memory, modems, and other essential bits and pieces manufactured together in a single chip that’s soldered onto the circuit board.
The System-on-a-Chip is the brain of your smartphone.
Combining multiple components into a single chip saves on space, cost, and power consumption. SoCs connect to other components too, such as cameras, a display, RAM, flash storage, and much more. Essentially, an SoC is the brain of your smartphone that handles everything from the Android operating system to detecting when you press the power off button.
The list below contains the most common components that you will find inside a smartphone System-on-a-Chip. We’re going to cover a few of the most important ones later on in this article.
- Central Processing Unit (CPU) — The “brains” of the SoC. Runs most of the code for the Android OS and most of your apps.
- Graphics Processing Unit (GPU) — Handles graphics-related tasks, such as visualizing an app’s user interface and 2D/3D gaming.
- Image Processing Unit (ISP) — Converts data from the phone’s camera into image and video files.
- Digital Signal Processor (DSP) — Handles more mathematically intensive functions than a CPU. Includes decompressing music files and analyzing gyroscope sensor data.
- Neural Processing Unit (NPU) — Used in high-end smartphones to accelerate machine learning (AI) tasks. These include voice recognition and camera processing.
- Video encoder/decoder — Handles the power-efficient conversion of video files and formats.
- Modems — Converts wireless signals into data your phone understands. Components include 4G LTE, 5G, WiFi, and Bluetooth modems.
Finally, talk about SoCs will often mention something called the manufacturing process. It’s listed as a number in nanometers (nm). Generally speaking, the smaller the nm size, the smaller the internal wiring of the SoC. This is better for power efficiency and silicon area size. Although there are different methods of manufacturing that can make direct comparisons tricky. 7nm is currently the smallest available manufacturing process used by smartphone SoCs.
Examples of an SoC
Now that we have a brief overview of what an SoC is, how about a few examples. In the smartphone space, Qualcomm, Samsung Semiconductor, Huawei’s HiSilicon, and MediaTek are the four biggest names in the business. Chances are that your smartphone has a chip from one of these companies in it.
Qualcomm is largest provider of smartphone SoCs, shipping chips for the majority of flagship, mid-tier, and even low-end smartphone releases each ear. Qualcomm’s SoCs fall under the Snapdragon branding. Premium chips boasting the company’s best technology come under the Snapdragon 800 series banner, such as the latest Snapdragon 865. Mid and super-mid tier products are branded with Snapdragon 600 and 700 series names respectively. Such as the Snapdragon 765 which sports 5G connectivity. Entry level products are named under the 400 series.
Samsung’s Exynos SoCs operate on a similar premium, mid, and entry tier scale. These were previously listed as the Exynos 9900, 9800, and 9600 series, with Exynos 7000 series products propping up the budget end of the portfolio. However, Samsung’s latest high-end chip is the Exynos 990, while the Exynos 980 is a 5G mid-tier chipset.
Samsung’s Exynos naming scheme now very closely resembles Huawei’s. The Kirin 990 is Huawei’s latest flagship chip, which comes in 4G and 5G variants. The Kirin 600 series is much like the Snapdragon 600 range, offering mid-tier specifications for more affordable smartphones. Finally, MediaTek’s Helio range spans affordable P series products up to the gaming focused G series, and latest flagship Dimensity 1000 with 5G.
It all starts with the CPU
You might be familiar with the term processor as this is often used interchangeably with CPU in this circle of conversation. A CPU is the most commonly used type of processor. It’s designed to be highly flexible and suitable for a wide range of tasks. As such, the CPU runs the Android operating system and your apps. It’s also partly responsible for synchronizing data between other processors inside the SoC.
As a quick overview, CPUs operate using prediction units, registers, and execution units. This is known as the CPU architecture. Registers hold bits of data or pointers to memory, often in 64-bit data formats. Execution units do something with one or more registers, such as reading and writing to memory or performing math. Multiple execution units can be used at once with the CPU, each taking a clock cycle or two to complete their function.
CPUs are flexible enough to suit a wide variety of tasks. Performance can be scaled up and down by changing the clock speed (in GHz), the number of cores, or by changing the underlying architecture to do more with each clock cycle. This latter point is often what’s referred to as building a “wider” or “bigger” CPU, which is how Apple’s phone chips are so powerful. However, there are power and efficiency trade-offs to these wider designs too.
CPUs inside smartphone SoCs come in a variety of flavors, all of which are based on the Arm CPU architecture. The latest CPU cores from Arm are the big Cortex-A77 and little Cortex-A55. Smartphone CPUs often appear in eight-core configurations, with big powerful cores for more demanding applications and smaller power-efficient cores to ensure long battery life.
Alongside the CPU, the graphics processing unit (GPU) is another piece of traditional number-crunching hardware packed into a phone SoC. GPUs are far less general-purpose than CPUs and are designed very differently as a result. They are built to repeatedly cycle through mathematical functions in parallel, which they can do much faster than a normal CPU. Remember, there are millions of pixels to fill on your 1080p display, each of which has to be calculated when you’re running an app or your favorite game.
Most graphics operations are repeated over and over again to fill all the pixels on your screen. As such, GPUs are designed to run lots of math all at once on big batches of data.
Unlike CPUs that execute one or two operations each cycle, GPUs execute tens, hundreds, and even thousands of parallel operations each cycle. This depends on the size and performance of the GPU design.
The two major GPUs in the mobile SoC space are Arm’s Mali and Qualcomm’s Adreno. Both offer bigger and smaller versions of the GPU technology, with flagship chips packing in their most powerful hardware for 3D gaming. Qualcomm doesn’t talk much about the inner workings of Adreno, but we know all about Mali. You can read about the latest Mali architecture here.
Great cameras need good processors
Smartphones are increasingly judged on their photography capabilities. While high-end sensor and lens hardware is essential, powerful image processing capabilities are an equally important part of the story. Just look at the impressive results from the Google Pixel 4’s simple camera hardware setup.
While image editing and tweaking are often done on the CPU and GPU, there’s a ton of processing performed on camera sensor data before an image is even saved to your phone. An ISP is a specialized DSP that handles common imaging tasks like Bayer transformations, focusing, demosaicing, sharpening, and noise reduction. In other words, it turns digital information from a camera sensor into a nice looking picture.
Those last two are particularly important in smartphones, where cheaper handsets tend to oversharpen and produce mushy looking details. Meanwhile, at the high-end, Huawei’s Kirin 990 is the first SoC with DSLR-grade block-matching and 3D filtering (BM3D) noise reduction. This produces great results in the Huawei Mate 30 Pro.
The bottom line is that great looking pictures require a powerful image processor.
Next-generation AI processing
Terms like neural processing units, AI processors, or machine learning cores are often used interchangeably but they all tend to mean the same thing inside modern smartphone SoCs: a processor that’s specifically optimized for the mathematics and algorithms commonly used by neural networks.
Just like how GPUs are processors optimized for graphics math and ISPs are optimized for image tasks, NPUs are processors specifically designed to run neural networks and machine learning tasks more quickly and efficiently than CPUs. NPUs feature their own local memory caches too, to speed up execution without having to use slower RAM.
Neural networks often require operations that take multiple pieces of input data to generate just a single output. The multiple-accumulate operation is particularly popular, often operating on a variety of data sizes from 16 bits down to 8 and even 4 bits of data. This is very different from the math and data types used by CPUs, although some operations can be accelerated on flexible GPUs.
NPUs are the latest specialized processor to find its way into phone SoCs. While mostly reserved for flagship-tier chips, this technology will trickle down to more affordable handsets in the near future.
4G and 5G modems for faster data
The final piece of a modern smartphone SoC is the data modem, which allows you to access data networks from your carrier. Different modems also determine the speed and quality of your data connection. The most powerful modems hit speeds above 1Gbps. There are also modems for Wi-Fi and Bluetooth data, but we’re focusing on 4G and 5G modems today.
All current smartphone SoCs boast integrated 4G modems. This means that the 4G modem is located inside the SoC. The first 5G modems for smartphones are external, meaning that they have to be connected up to the main SoC. This is less energy efficient but makes it easier to implemented high-end features and provides manufacturer flexibility while 5G networks roll-out to more consumers.
The first SoCs with integrated 5G modems and capabilities are already here. Huawei has a 5G version of its Kirin 990 and Qualcomm’s Snapdragon 765 offers integrated 5G. However, the most powerful 5G modems are still external. Qualcomm’s Snapdragon X55 and Samsung’s Exynos 5123 modems provide faster speeds than their integrated counterparts as well as mmWave support. 2020’s flagship 5G phones will be sticking with external modems for a little while longer.
More about smartphone SoCs
Phone enthusiasts love to compare CPU and GPU specs, but this is becoming less relevant as performance matures and new capabilities are required. Smartphone SoCs are increasingly less about any one single capability and more about a heterogeneous compute approach to solving processing problems.
Today’s handsets handle a wider range of workloads than ever before. As a result, the number of dedicated processors inside each chip continues to increase. From basic CPU and GPU components a few years ago, to DSPs, advanced ISPs, and NPUs today. These less-talked-about parts are only becoming more important with advances in security, machine learning, and 5G.
For more information about specific smartphone chipsets, check out our selection of guides below: