The modern smartphone is an incredibly complex device, and sadly it has a very short lifespan due to it being designed for planned obsolescence. I saw this problem first hand when my girlfriend bought a used Huawei G play mini (CHC-U23) smartphone. Her previous smartphone had been stolen and she needed something cheap to replace it. After a year and a half of usage, the battery swelled up and was unable to hold a charge for more than 15 minutes.
Unfortunately, it was only possible to buy replacement batteries from China, and it would take over 4 weeks in shipping, so my girlfriend bought a new phone. After several years of sitting in a drawer, I decided to tear the phone apart and see what was inside.
Here are the phone’s specs:
Huawei G play mini (CHC-U23)
Released: August 2015
OS: Android 4.4.2 KitKat, upgradeable to Android 6.0
SoC: HiSilicon Kirin 620, 8x 1.2GHz Cortex-A53, ARM Mali450-MP4 GPU
RAM: 2GB LPDDR3 SDRAM
Storage: 8GB eMMC NAND Flash
Display: 5” 720×1280 IPS, 294 ppi, 24-bit color, 67.11% screen-to-body ratio
Cellular modem: Dual SIM, supports GSM and UMTS bands
Back camera: 13MP (4160×3120), Sony IMX214 Exmor RS, 1080p@30 video, f/2.0
Front camera: 5MP (2560×1920), OmniVision OV5648, 720p video
Wireless: WiFi 802.11b/g/n, hotspot, Bluetooth 4.0
Port: micro-USB port, USB 2.0, OTG
Other: FM Radio, microSD
Battery: 2550 mAh Li-Ion
Body: 162 g, 71.9 x 143.3 x 8.8 mm, plastic body and case
The G Play mini was released at a time when most of the higher end phones had already moved to non-removable batteries and sealed cases. Sadly the last flagship phone with a removable battery was the LG V20, released in September 2016, but most mid-range smartphones still had removable batteries in 2015 when the G Play mini was released.
Since the back cover pops easily off the G play mini, I expected that there would be no problem replacing the battery. However, I quickly discovered that the battery in the G play mini was not designed to be removable. The only way to get to the battery is to remove a light grey inner plastic panel covering the internals, which requires unscrewing 17 screws. One of the screws has a sticker over it, with the label “Torn Invalid”:
In other words, the battery can only be replaced by invalidating Huawei’s warranty. The battery has a metal cover over it which is glued to the back of the battery. I found the glue to be so strong, that it took me several minutes of prying with a flathead screwdriver to free the battery from its metal cover, which got mangled in the process.
Today phone makers claim that they can’t give us removable batteries because they need to glue shut the case to keep out water and dust and they can make the phone thinner and lighter if the battery is not removable. However, none of those arguments apply in the case of the G play mini, since it has an easily removable back cover, and the metal cover over the battery, which weighs 8.9 grams, probably takes up just as much space and extra weight as a battery with a hard plastic cover, and probably costs more to manufacture. Huawei went out of its way to design a phone where the user can’t replace the battery, for no reason except to promote planned obsolescence. The cobalt lithium oxide batteries in smartphones typically start to degrade after 500 recharge cycles, so the G play mini was deliberately designed to have a limited lifespan.
In 2015 when the G play mini was released, 40.1% of that year’s new phone models had non-removable batteries according to the gsmarena.com phone database, so Huawei was hardly the only phone maker that was promoting planned obsolescence at the time. By 2017, 79.4% of new phone models had non-removable batteries, and by 2020, that percentage was 95.3%. Most of the phones with removable batteries on the market today are low-end models with low-resolution cameras, limited RAM, and cheap processors. The only one with competitive specs is the Samsung Galaxy XCover Pro, which is a specialty rugged model geared for corporate sales.
The other problem with the design of the G play mini is that the display is fused to the plastic frame, and that frame includes many components, including a PCB with some unlabeled component (possibly an accelerometer or magnetometer), an EMR vibration motor, a Synaptics touch IC , an ambient light and proximity sensor, 2 side button circuits and 3 front-facing buttons. These extra components make it expensive to replace the display, which is the most common type of repair done on smartphones.
The modern slate design of smartphones, which Apple popularized with the iPhone, has glass extending over the entire front face of the phone, which means that its glass face is liable to crack when dropped since it doesn’t have much of a bezel to protect it. According to a 2020 report by the Indian electronics service firm OnsiteGo, 71% of its mobile phone repairs are caused by screen damage, 8% are the phone not starting, 6% are device damage, 3% are water logging, 2% are software issues and 2% are charging. Similarly, a 2016 survey by Big Giraffe in Australia found that 62% of damaged smartphones have a cracked screen and 10.7% have liquid damage.
While it is relatively easy to take apart the G Play mini and replace the display, the fact that the LCD is fused to the frame makes it expensive to repair, which makes it more likely that the phone will be junked rather repaired when it gets damaged. The plastic bezel (which is painted gold to appear like metal) is only 1 mm think around the glass front face of the phone, so it offers minimal cushioning of the glass when dropped, but it is raised slightly (about 0.2 mm) over the glass, so the glass has a bit more protection that most phones today which have eliminated the bezel lip altogether.
Despite these criticisms, I do find some good aspects about the G play mini’s design:
- It uses plastic pressure tabs to attach the back cover, rather than glue and single-use adhesive pads and gaskets, which seal shut the case.
- It uses a polycarbonate back cover. Plastic cases are superior because they don’t block radio frequencies, so the antennas can be placed inside the phone which better protects them, instead of being built into the case which is necessary with metal cases. Plastic yields in a drop, which better protects the internal components, whereas glass cases are prone to cracking and metal transmits the force of a drop, making it more likely that the front glass will be damaged.
- The components are fastened down with 17 identical shiny screws, so there is not a problem with mixing up the screws when reassembling the phone. The only two screws which are a different size are black, so they won’t be confused when reassembling. All the screws have standard Philips heads, so it isn’t necessary to buy a specialty screw driver set just to open up the phone like Apple does with its custom pentalobe screws.
- All the components that aren’t soldered on boards have connectors or pressure pins (such as the battery, speaker, LCD, sensors, side buttons, etc.), so they can be easily replaced without using a soldering iron. Only in the case of the battery glued to its metal cover is the part difficult to replace.
- Unlike today’s phones that require the use of tiny pins to eject the microSD and SIM cards, which are on tiny trays on the outside of the phone, the 2 microSIM and microSD cards in the G play mini are placed under the back cover and they are easy to be remove using fingers and don’t require taking out the battery and turning off the phone. Each card can be removed separately which isn’t possible when multiple cards are placed on the same tray, as is common in modern phone design.
Unfortunately, Huawei is not interested in supporting its old phones like the G play mini. Huawei did support the phone for the standard 2 years after its release by providing software upgrades from Android 4.4 to Android 6.0, but the phone’s website no longer offers any manuals or software downloads for the phone and Huawei sells no spare parts for the phone. To get the G play mini’s documentation and firmware, it is necessary to download them from 3rd party sites such as manualslib.com and huawei-updates.com, which may not be reliable or safe. People seeking to repair the G play mini can buy its display+frame and battery from AliExpress (be prepared for long shipping times from Hong Kong), but the only way to get the rest of the parts is to salvage them from used units.
Sadly, Huawei’s attitude toward its old devices is hardly unique in the mobile industry, which is based on planned obsolescence. Although Apple has been a leader in promoting bad hardware design based on planned obsolescence, Apple has been the best phone manufacturer in terms of providing software updates for its phones. Most iPhone models get 5 years of software updates, whereas most Android models only get 2-3 years of software updates. Google only requires that licensees of Android provide security updates every quarter in the first year after a phone is released and once in the second year, and many of the cheaper smartphone models get little more than that.
Even when the maker of an Android phone wants to keep providing software updates, the mobile industry often prevents it. Fairphone had the goal of providing at least 5 years of software updates for the Fairphone 2, which contains a Snapdragon 801 processor and was first released in December 2015 with Android 5.1 Lollipop. None of the Android phones with the Snapdragon 800/801 that were released in 2013-4 got upgraded to Android 7 (Nougat) in 2016-7, because Qualcomm decided that it wouldn’t release updated graphics drivers for the Snapdragon 800/801 since it was no longer selling the chip. Others say that the reason the Snapdragon 800/801 couldn’t be officially upgraded to Nougat is because it lacked hardware AES encryption and full disk encryption was mandated by Nougat’s Android Compatibility Definition Document (CDD) and it couldn’t pass the encryption speed requirements of the Android Compatibility Test Suite (CTS). Because Fairphone needs to provide its users with access to the Google Play Store and Google Web Services (such as Google Maps), it can only provide upgrades to Android that meet Google’s standards.
In order to obey Google’s onerous CDD rules and pass its CTS, Fairphone had to spend €500,000 to switch from Qualcomm’s unsupported Snapdragon 801 drivers to community-developed free/open source drivers. In November 2018, the Fairphone 2 became the only Snapdragon 800/801 phone to officially receive a Nougat upgrade. Unfortunately, Fairphone has not been able to provide 5 years of software updates as promised, because Google released its last update to Nougat in October 2019 since it has a policy of only supporting its Android releases for 3 years. The Fairphone 2’s last official software update was in December 2019 and the phone is still using Android 7.1, which was first released in October 2016, and the phone is still using Linux kernel 3.4.0, which was released in May 2012.
A small phone manufacturer like Fairphone can’t easily extend the lifespan of its phones because it depends on the manufacturer of the phone’s phone’s System on a Chip (SoC) to keep providing firmware and driver updates for newer versions of Android. Huawei, however, easily could have provided more years of software updates for the G play mini if it had wanted to, since it is the manufacture of the G play mini’s SoC.
The G play mini is based on the Kirin 620 application processor, which is made by Huawei’s subsidiary HiSilicon. Huawei’s original announcement presented the Kirin 620 as a integrated mobile SoC (System on a Chip), similar to a Qualcomm Snapdragon or MediaTek Helio, but its Hi6220 application processor requires the addition of several ancillary chips to match the full functionality of a modern mobile SoC. Huawei is very tight lipped about the design of its chips, and the HiSilicon web site has very little info about the Kirin 620. However, Huawei originally planned to sell Kirin to other device makers similar to the way that Qualcomm markets the Snapdragon, so more companies got their hands on Kirin 620 documentation. Luckily, one of the companies that decided to use the Kirin 620 was 96Boards, which is the hardware arm of Linaro, that produces a lot of the free/open source drivers for ARM processors. 96Boards designed its HiKey single board computer (SBC) around the Kirin 620, and posted the Hi6220’s Function Description manual on Github, so it is possible to find some documentation for the Kirin 620, unlike subsequent Kirin processors.
The Hi6220 includes a central processing unit, a graphics processing unit, a video processing unit, an image processing unit and a digital processing unit (for audio), and the basic functionality for a cellular modem, but it lacks many of the functions found in mobile SoC’s, including WiFi, Bluetooth, a global navigation satellite system receiver (i.e., GPS), and a charge controller (like Snapdragon’s Quick Charge), plus today’s mobile SoC’s also include a neural processing unit for AI. A phone designed around the Hi6220 has to include these functions as extra chips, which makes the guts of the G play mini different from most smartphones on the market.
The two Linux phones currently on the market,1 the Purism Librem 5 and PINE64 PinePhone, also have separate chips to provide WiFi, Bluetooth, cellular modem, GNSS and charge controlling, and they are also based on SoC’s that use Cortex-A53 cores built on older planar nodes, like the Kirin 620. Aside from the fact that the Hi6220 includes a cellular modem, its functionality looks similar to the NXP i.MX 8M Quad used in the Librem 5 and the Allwinner A64 used in the PinePhone, so I was curious to see how the design of the G play mini compares to the two Linux phones. I have created component lists for the Librem 5 and PinePhone, which is possible because Purism and PINE64 are the only phone makers on the market that publish the schematics for their phones.
Because Huawei doesn’t provide schematics, I took out the G Play mini’s printed circuit board (PCB) to examine its components. Like most of today’s smartphones, the G Play mini’s PCB is designed in an L shape around the battery, with the cameras and headphone jack in the bottom of the L and the USB port at the top of the L.
Both the Librem 5 and PinePhone also have a USB port at the bottom of the phone and a 3.5mm audio jack at the top, just like the G Play Mini, but the Linux phones use two separate PCBs at the top and bottom of the phones connected by a ribbon cable, rather than a single L-shaped PCB that reaches both ends of the phone. Having two PCBs and a connecting ribbon cable costs more, but it provides more modularity. A broken USB port is easier to fix on the two Linux phones, because it only requires replacing a small USB board, rather than a unified board that contains all principal components. In both the Linux phones, they have batteries in the middle between the two PCB’s, and those batteries are designed to be easily removable without tools.
Unfortunately everything interesting on the G play mini’s PCB is covered by shielding. However, four of the shield covers were removable, allowing me to glimpse the chips under the shields. I pried off the shield covers with my fingernails, although in a few areas I had to use a flathead screwdriver to pry them up.
The Hi6220 processor has thermal paste connecting it to its shield cover to disperse heat, although a thin stainless steel sheet isn’t a great conductor. I also see a black sheet on the back of the LCD which I assume is graphite to distribute the heat from the chips on the front of of the PCB. I’m guessing that 8 Cortex-A53 cores running at 1.2GHz on a 28nm planar node don’t need as much cooling as today’s 5nm Kirin 9000 processor that can reach speeds up to 3.13GHz.
The PinePhone with 4 Cortex-A53 cores running at 1.152Ghz on a 40nm planar node gets by without any thermal paste for cooling, although there is probably some thermal dissipation through its surrounding shielding cover.
Like the G play mini, the Librem 5 also uses thermal paste to dissipate the heat of its SoC, but its thermal paste joins the SoC to an aluminum sheet that connects to the entire metal frame of the phone, so its potential heat dissipation is greater. This is possible because the Librem 5 has a metal frame, whereas both the G play mini and PinePhone have plastic frames that can’t conduct heat like metal does. The i.MX 8M Quad runs at 1.5GHz on a 28nm node and has a bigger GPU so it probably needs more heat dissipation than the Hi6620 processor. Purism discovered with Aspen, which was the first version of the Librem 5, that the SoC got too hot, so Purism had to redesign the phone to place the i.MX 8M Quad on the other side of the PCB so it could dissipate heat through the back of the LCD and aluminum frame.
The frames of the shields on the G play mini’s PCB still covered a lot of details, so I took a pliers and ripped off the shields.
Then, I labeled the major components on the G play mini’s PCB to try to identify them:
The labels on 4 of the integrated circuits are so light that I can’t read them with my magnifying glass. Even when I was able to read the chip labels, Google searches only turned up info on 4 of the 16 chips on the G Play mini’s PCB.
|My label||Label on component||Size (mm)||Description|
|141 x 64||Printed circuit board assembled, 11.36 g + 2.80 g in shielding covers|
|A||7 x 8.5||3.5mm headphone jack|
|B||6.5 x 6.5||Pad for front (selfie) camera|
|C||7.5 x 3||24 pin front camera connector|
|D||9 x 3||38 pin back camera connector|
|1||(can’t read)||3 x 3||16 pins|
|2||(can’t read)||5.5 x 5.2|
|3||057||3.5 x 5||4 pins|
|3 x 3.2|
|5||Hi6220 GFCV10001 HP2251509|
509 – TAIWAN
|12.5 x 12.5||Hi6220 V100 Multi-Mode Application Processor 8x Cortex-A53, Hi-Fi2 audio processor, Dolby/DTS 5.1/7.1, TSMC 28nm LP (low-power process), BGA 653 pins, 0.48 x 0.48 in 13MP photo (13MP@15fps, 8MP@30fps), 1080p@30fps video encoding, 1080p@30fps video decoding, ARM Mali450-MP4 GPU, 4x 500MHz, OpenGL ES 1.1/2.0|
|11 x 12||Elpida 2GB LPDDR3- DRAM, 800MHz, 1.2V, VFBGA-178|
|13 x 11.5||Samsung 8GB NAND Flash eMMC KLM8GIGEND-B031,|
FBGA-153, 13 x 11.5 x 0.48 mm
|8||(can’t read)||4.5 x 4.5|
|6.8 x 6||HiSilicon Hi6361GFC, maybe a radio frequency IC or an audio IC, BGA-153|
|3.1 x 4.3||14 pins|
|2 x 2||18 pins|
|12||AA4||2 x 2||6 pins|
|13||(no label)||1.8 x 1.8||6 pins|
|E||L1515||8 x 5||microUSB 2.0 port, 5 pins|
|Back of PCB|
|F||2.3 x 2.3||Is it an unused antenna connector?|
|6 x 6|
|G||10 x 3||38 pin connector to LCD|
|H||6 x 3||6 pin battery connector|
|13 x 14||First mini-SIM (2FF) slot, 8 pins|
|J||SNK||11.3 x 11.3||microSD card slot, 8 pins|
|13 x 14||Second mini-SIM (2FF) slot, 8 pins|
|15||(can’t read)||2 x 4|
1508 – TAIWAN
|7.2 x 7.2||HiSilicon Hi6553 power management integrated circuit (PMIC), integrated 4-channel BUCK, 24 LDO, 1 LVS (Low Voltage Switch), a 12-channel HKADC, and 5-channel LED current driver ports, built-in 32K crystal and 19.2M crystal oscillation circuits. It supports under-voltage, over-voltage, over-heat, over-current protection. Output capacity: BUCK1=3.5A, BUCK2=3A, BUCK3=1A, BUCK4=1A|
|A||L1515||7 x 8.5||Back of 3.5mm headphone jack|
|L||5 x 3||10 pin connector to ambient light and proximity sensor|
|M||5 x 3||10 pin connector to S3207B touch IC|
|6.2 x 6.2||Touch IC for the LCD, connected to a 10 pin connector|
|8.2 x 8.2 x 5||13MP (4160×3120) back camera, Sony IMX214 Exmor RS, 1080p@30 video, f/2.0 (similar part), with 38 pin connector, 0.67 g|
|6.3×6.3×4.3||5MP (2560×1920) front camera, OmniVision OV5648, 720p video, with 24 pin connector, 0.19 g|
|Q||51J||10 x 5||EMR vibration motor,|
|R||GA3.1 F41C2 HF||15 x 6 x 3||Top speaker, 1.59 g|
|Ambient light and proximity sensor|
|15 x 11 x 3||Fingerprint reader, 1.59 g|
|40 x 96 x 5|
42.5 x 99×6
|Huawei 3.8V 2550mAh 9.7Wh Li-Polymer battery, limited charge voltage: 4.35V, 38.26 g (battery) + 8.93 g (metal cover) = 47.18 g|
|U||AD11A||141 x 70||Light gray internal plastic cover, 7.86 g, 4 tape antennas in each corner: S U 2330PA915141 (top left by back camera), S2329PA5141U (top right), S2341 PA U (bottom left), S2328PA5141U (bottom right by fingerprint reader)|
|144 x 72 x 6.5||White polycarbonate back cover with power and volume rocker buttons, 13.00 g.|
The fact that I was unable to find any info about many of the chips on the PCB indicates that these chips are probably custom orders for Huawei. For example, I can only find reference to the Synaptics S3207B touch IC being used in Huawei devices.
Three of the chips in the G Play mini were made by Huawei’s subsidiary HiSilicon, including the Hi6220 application processor, the Hi6553 and the Hi6163FC. According to the documentation for the 96Board’s HiKey, the Hi6553 is a power management IC. Unfortunately, there is no information publicly available about the Hi6163GFC. PhoneParts.com says that the Hi6163GFC is an audio IC, which is possible since the Hi6220’s logic block diagram (see below) shows that that the chip needs a external “audio CODEC & audio amplifier”. However, the Hi6163GFC has 153 balls, whereas normally an audio IC would not need to be so complex. For example, the Cirrus Logic WM8962 audio IC has 49 balls on it. The Kirin 620 also contains a LTE Cat4 modem (150Mb/s download, 50Mb/s upload), but the documentation available for the Hi6220 doesn’t mention it, except to show a modem in the chip’s logic block diagram and links to an external “RFIC” chip. A teardown in Chinese of the Honor 4C, which is nearly identical to G play mini, says that the Hi6361GFC is a radio frequency chip, so I think that it is possible that it is the unnamed “RFIC” chip in Hi6220’s logic block diagram:
There are many other questions about the G play mini’s board. I can’t identify which chips provide the WiFi/Bluetooth, GNSS, accelerometer and magnetometer. If the Hi6163GFC is an audio chip, then I’m guessing that the RF8240A (which I labeled as “14”) is the radio frequency IC, but that is just speculation based on its name that starts with “RF”.
There is also an antenna connector on the board close to the RF8240A chip which appears to be unused (which I labeled it as “F”). The internal plastic panel (which I labeled as “U”) has four tape antennas in each corner, so I’m guessing that that Huawei used that antenna connector when developing the board, but didn’t need it in the final product. I tried peeling back one of the tape antennas from the plastic panel, but they are very fragile and break easily. Both the Librem 5 and PinePhone use normal wire antennas which are easy to move around and adjust.
Despite my questions about the G play mini’s PCB, it is clear to me that considerable skill was involved in designing and manufacturing the board. The components are very densely packed in some sections of the board, and Huawei used such small components in some sections, that I can barely count them, even with a magnifying glass. The G play mini’s board makes extensive use of 0201 sized (0.6 x 0.3 x 0.25 mm) components. In a space of 13 x 2 mm behind the microSD card slot, Huawei crammed in 29 different components (see the magnified photo below). The black components (resistors?) in the photo have such low height that I think that they are less than the standard 0.25 mm height, but I don’t have the tools to measure exactly.
Huawei only assembles 10% of its phones in house, so I’m not sure whether Huawei made the G play mini’s board or it was one of Huawei’s assemblers (Flex, Foxconn/FIH Mobile, BYD, Huaqin, Wingtech, etc.), but it looks like precision work to my admittedly amateur eyes The Librem 5’s board also has some 0201 sized components, but fewer of them, and they aren’t as closely placed on the board. I see clear sealant around the principal chips on the G play mini’s board (probably to keep out moisture but it also might be heat dispersing glue), whereas I don’t see sealant around the chips on the Librem 5 and PinePhone boards.
I can’t tell what is the purpose of each component (resistor, capacitor, transistor, crystal oscillator, etc), but I counted the number of electronic components that I can see on the G play mini to compare it with the two Linux phones. In a PCB that is roughly 3500 mm2 in size, I count 513 different electronic components, plus another 20 electronic components in the frame and back of the LCD. Since I’m just using a cheap magnifying glass, I’m not sure of my count in some places, but it gives a rough idea how many components are packed into Huawei’s smartphone.
Number of electronic components in the Huawei G play mini (CHC-U23)
|Back side of PCB|
|@ antenna connector||12|
|@ RF8240A (chip 9)||19|
|@ SIM-uSD-SIM slots||30|
|Section near SIM 2||23|
|Section near chip 10||36|
|@ Hi6553 PMIC (chip 10)||80|
|Corner under cameras||18|
|@ headphone jack||16|
|Last section w/ connectors||20|
|Total back side||282|
|Front side of PCB|
|Around headphone jack||3|
|@ front camera connector||17|
|@ back camera connector||24|
|@ 057 (chip 8)||8|
|@ B512 (chip 7)||14|
|@ Hi6220 SoC (chip 1)||26|
|@ 2GB RAM (chip 2)||17|
|@ 8GB eMMC (chip 3)||10|
|@ Hi6361JFC (chip 4)||50|
|@ chips 5 and 6||14|
|@ microUSB 2.0 port||37|
|Total front side||231|
|Total on PCB||513|
|In frame/back of LCD|
|Light sensor PCB & cable||4|
|Touch IC PCB & cable||7|
|Vibrator ribbon cable||3|
I have never counted the number of components in a typical mid-range smartphone with an integrated mobile SoC like a Snapdragon or Exynos, but I would guesstimate from photos that it is around 400. The fact that the Hi6220 doesn’t have as many functions built into the chip does add more complexity to the G play mini’s PCB.
Nonetheless, the two Linux phones on the market have significantly higher component counts on their boards than the G play mini. The PinePhone has 672 components on its two PCBs, which is 31% more than the G play mini’s PCB. The Librem 5 is an extremely complex device with a 1174 components in its two PCBs. In addition, the RS9116 WiFi/Bluetooth M.2 card contains 87 components and the BM818 modem M.2 card contains 6 components, so the 4 boards in the Librem 5 contain a total of 1267 components, which is 150% more than the G play mini’s PCB.
Number of components on PCBs of Linux phones
|Type of component||Librem 5 main||Librem 5 USB||PinePhone main||PinePhone USB|
|Antenna connectors (ANTxxx)||3||2||6||4|
|Test points (T/TC/TP/TS/TVxxx)||126||4||27||3|
|Integrated circuits (U/DUxxx)||65||2||24†||2|
|Crystal oscillators (Y/Xxxx)||10||0||5||0|
|Total without test points||1133||41||627||45|
|Total with test points||1259||45||654||48|
Both the Linux phones have more circuit board area and their components don’t appear as tightly packed as the G play mini’s PCB. Some of the chips on the Librem 5 and PinePhone boards are flat packages with side pins that can be replaced without needing specialized equipment. The PinePhone was explicitly designed for people who want to hack the hardware, and Purism says that it added an M.2 slot for its WiFi/Bluetooth card because ham radio operators might want to use different wireless cards in the Librem 5.
The Linux phones have more components than the G play mini for a couple reasons. Both the Librem 5 and PinePhone have their cellular modems on separate chips than their SoC, whereas the modem is integrated into the Hi6220 in the G play mini, so it requires fewer components on the board. The hardware kill switches in the Librem 5 and PinePhone require extra circuitry to be able to cut the power (or send power off signals) to different components on the board.
The Librem 5 is the most complex phone that I have ever seen in terms of the number of components on its boards. The main PCB in the Librem 5 has 10 layers and the packages of the 69 ICs in the Librem 5 occupy over 1600 mm2 in area. In comparison, the 17 ICs in the G play mini occupy 722 mm2.
The Librem 5’s high component count is partly explained by the fact that it has a separate GNSS and audio IC, whereas those functions are part the cellular modem and SoC in the PinePhone, plus the Librem 5’s kill switches can turn off 3 more components than in the PinePhone, so they require more circuitry. The Librem 5 is the first mobile phone to put the WiFi/Bluetooth and cellular modem on M.2 cards, which allows them to be upgraded, but that requires adding extra components. Another factor is that the Librem 5 will be the first phone with a smartcard reader for using cryptographic operations with an OpenPGP card. Implementing that requires adding a separate STmicroelectronics STM32L432KC 80MHz Cortex-M4 microprocessor to the board to control the smartcard reader.
What impresses me about the G play mini is that it managed to achieve good battery life, despite the fact that the Kirin 620 was Huawei’s first attempt at making a mobile processor using 64 bit cores and its previous Kirin 910 processor had been an energy hog. Doubly impressive is that Huawei managed to achieve good energy efficiency with a 28nm planar node and using many separate chips. Having all the functionality packed into a single integrated SoC at a smaller node size is generally more energy efficient, yet the G play mini managed to match the battery life of phones with Snapdragons at 20 or 14 nm, which was an impressive feat of engineering.
In contrast, one of the biggest problems of the Librem 5 and PinePhone are their short battery lives, despite having larger batteries than the G play mini. The i.MX 8M Quad’s mainline Linux driver still doesn’t have the ability to suspend to RAM, but even once that problem is solved, the Librem 5 will still be an energy hog due to its large number of chips and most of them are built on older node sizes that are less energy efficient. The PinePhone is planning on offering an external keyboard mod with a 6000 mAh battery to extend its short battery life.
Unlike the G play mini, both the Librem 5 and PinePhone are designed to have long lifespans. Purism promises to provide lifetime software updates for the Librem 5 and PINE64 promises to sell the PinePhone for an unprecedented 5 years, meaning that both companies had to select components which are supported for a long time by their manufacturers. For example, NXP promises to manufacture the i.MX 8M Quad processor in the Librem 5 for 15 years till January 2033. Both Purism and PINE64 have worked to get the hardware in their devices supported in mainline Linux, so both devices should be able to use future versions of the standard Linux kernel without any modification. It should be easy to install the latest Linux kernels in these devices, unlike in Android devices which use antiquated Linux kernels that often can’t be upgraded because Qualcomm, MediaTek, Samsung and Huawei generally only support their mobile SoC’s for 3 years and stop issuing firmware and driver updates, which makes it difficult to upgrade them to more recent versions of Android.
It is clear looking at the design of the G play mini and its HiSilicon chips is that Huawei is following a very different path than many of its Chinese rivals, including BBK Electronics (which owns OPPO, vivo, realme and OnePlus), Xiaomi, Lenovo (which owns Motorola), Transsion Holdings (which produces the brands itel, Infinix and Tecno), TCL (which licenses the brands Alcatel, Palm and Blackberry) and ZTE. These companies use processors from companies such as Qualcomm, MediaTek and UNISOC, and their reference designs. They use radio frequency chips from companies like Avago Technologies, Skyworks Solutions and Qorvo and power management chips from companies like Analog devices, Maxim, STMicroelectronics, TI, NXP, RENESAS and Microchip.
In contrast, Huawei is following the path of the market leaders Samsung and Apple in designing its own chips and trying to control its own ecosystem as much as possible. The Hi6220 is not a particularly powerful processor, and it is built on an older 28nm planar process, so Huawei did not receive any performance benefits by creating its own mobile SoC, compared to using a Qualcomm Snapdragon or MediaTek Helio.
Creating competitive mobile SoC’s is an expensive endeavor, and so far only the three biggest phone sellers have managed to do it successfully, but they each ship between 200 and 300 million smartphones per year, so the high development costs are spread over many units.
Huawei minimized its costs and the financial risk to the company by taking an iterative approach, that gradually added more functionality and more expensive process nodes with each iteration. Huawei started by building the K3V2, which was a simple mobile SoC based on an inexpensive 40nm planar node that Huawei released in early 2012. The K3V2 contained four Cortex 9 cores and a Vivante GC4000 GPU and Huawei used it in 6 of its lower-end phones and one of its tablets. Having gained experience with K3V2, Huawei then designed the Kirin 910, based on 4 Cortex-A9 cores running at 1.6GHz and a low end Mali-450 4MP GPU, but it added a cellular modem and used a slightly more expensive 28nm HPM node. The Kirin 910 used too much energy so it was mostly used in tablets which have bigger batteries. With the Kirin 620, Huawei aimed for more energy efficiency, so it dialed back the clock to 1.2 GHz, but moved from the 32bit Cortex-A9 to the 64 bit Cortex-A53 and added 4 additional cores. With the Kirin 65X series, Huawei moved to a more expensive 16nm FINFET+ node, and added built-in WiFi, Bluetooth and GNSS. With each successive generation (92X, 93X, 95X and 960), Huawei got a little closer to the performance of the top-tier Snapdragons and Exynos chips. Huawei focused on energy efficiency and its P10 smartphone with the Kirin 960 was able to achieve several days of battery life on a single charge. With the Kirin 970 released in late 2017, Huawei finally caught up with the market leaders in performance by offering a 5.5 billion transistor chip at 10nm FinFET node.
Seeing the success of Samsung, Apple and Huawei, Xiaomi also tried to create its own mobile SoC, but it was impatient and tried to get there all at once, and its first attempt largely failed. Xiaomi’s Surge S1 processor didn’t fail for technical reasons; it got good performance reviews and was competitive with other mid-range mobile SoC’s when it was released in 2017.
Instead, the Surge S1 failed because it didn’t have enough volume to be able to be produced economically. Xiaomi was the sixth largest smartphone maker on the planet, shipping 92.7 million units in 2017 and 122.6 million in 2018, according to IDC. The Chinese company should have been able to put its Surge S1 processor in millions of devices, that would have distributed the costs. However, Xiaomi has a business model of growth at any cost, which meant that it has to shave its margins to undercut its competitors, so it doesn’t make sense to put its own processors in its devices, since using its own processor raises the costs more than using Snapdragons from Qualcomm.
In order to cut costs Xiaomi outsources most of its phone design to Huaqin, Longcheer and Wingtech, which are Chinese smartphone ODMs (original design manufacturers). In 2019, the design of 77% of Xiaomi’s phones was outsourced, and Xiaomi does almost none of its own manufacturing, because it relies on the ODMs, plus the electronic assemblers Foxconn, Flex, DBG and BYD. Among the major phone makers, only Motorola outsourced more than Xiaomi in 2019. In contrast, Huawei only outsourced the design of 17% of its phones in 2019. In Huawei’s recent Mate 30, roughly 20% of the phone’s bill of materials comprised chips produced by HiSilicon.
The advantage of designing in house is that Huawei can easily add its Kirin processor to many different phone models, so the cost of developing the mobile SoC are spread over millions of different phones. In contrast, Xiaomi only used its Surge S1 chip in one phone model, the Mi 5C which was only released in China, so its unit costs were much higher. Xiaomi could not rely on ODMs to design phones for its experimental SoC.
Another issue is that Xiaomi loaded everything into the Surge S1, which increased the costs and risks. It included a Realtek RT5659 for its audio codec, an ambient light sensor from Liteon, a Synaptics DSX touchscreen, an audio amplifier from NXP, and various sensors from Texas Instruments. In contrast, all of these were external chips with the Kirin 620, which made it simpler to design, so Huawei was able to reduce the costs and risks during early iterations of its Kirin processors.
Huawei’s methodical and incremental approach took a long time, but Huawei’s management was committed and didn’t give up. In contrast, Xiaomi’s management regarded the Surge as a backup plan, just in case something goes wrong with its current reliance on Qualcomm’s Snapdragon for the vast majority of its phones. According to Wired:
That’s why the Surge S1 launches, at least, mostly as a hedge. Xiaomi says it remains committed to working with Qualcomm and its other partners, and vice versa.
Xiaomi didn’t use the Surge S1 in any other devices after the Mi C5, and it looked like Xiaomi had decided to abandon making its own chips as a experiment that was too costly to continue. This wasn’t surprising, considering that the established chipmakers like Texas Instruments, nVidia, Intel and Freescale found that they couldn’t compete with Qualcomm’s mobile processors and they all bowed out of the mobile market. In early 2019, Xiaomi announced that it was willing to try its hand again at making a mobile SoC and pledged to invest $1.5 billion in artificial intelligence and smart devices over the next five years, but the company recently announced that it will spend $10 billion to enter the electrical vehicle market, so it is looking unlikely that it will have the resources to dedicate to designing its own mobile processor line.
Likewise, Oppo is reportedly working on its own mobile SoC. Oppo and Its parent company, BBK Electronics strike me as steadier companies than Xiaomi. They are more likely to commit to putting up the necessary resources and stick with it, but BBK Electronics operates on much of the same business model as Xiaomi, by selling at just above costs in order to gain greater market share, so it wouldn’t surprise me if Oppo also finds that it can’t justify the extra costs it will incur.
The reason why many of the major smartphone makers want to make their own processors is because it allows them to break out of the commodity trap that is gripping their industry. By the years 2011-12, when Google’s Android took over the smartphone industry outside Apple and Qualcomm’s Snapdragon acquired an unassailable dominance in integrated mobile systems on a chip, the smartphone makers lost their ability to distinguish their devices from their competitors, since they were all using the same operating system from Google and the same application processors and cellular basebands from Qualcomm. Google put the phone makers in a straight jacket though the Open Handset Alliance and the Android Compatibility Test Suite, which prevents them from doing much customization of their software compared to their competitors.
The only way to compete was to slash their prices or offer more expensive hardware in their devices, which cut their profits to the bone. After 2011 when the phone makers lost the ability to distinguish their devices from their competitors, most of the major phone makers lost money (Nokia->Microsoft, Blackberry, Motorola->Google->Lenovo, Sony and HTC) or barely broke even (LG and TCL). Only Apple, which makes its own operating system and designs its own application processors, and Samsung which makes a lot of its own hardware (application processors, image sensors, RAM, Flash memory, displays, etc) have consistently made profits over the years in the smartphone industry. It is clear that controlling one’s own ecosystem and being able to distinguish one’s devices from competitors is the secret to success in the mobile industry.
After Huawei starting making more of the chips in its own devices, it garnered more profits than its Chinese competitors. However, it will be difficult for other companies to get to enough volume where making their own chips makes financial sense.
The reason why I believe that Linux phones will eventually be a market success is because they break out of the Android+Snapdragon commodity trap that afflicts the smaller phone makers who can’t control their own ecosystems. Companies that make Linux phones can offer unique features that none of the Android phone makers can match. In my database of 1300 innovations in the mobile phone industry since 1973, I count 6 innovations in the Librem 5 and 4 innovations in the PinePhone, which makes them extremely innovative devices, in comparison to the sea of nearly identical Android phones churned out every year by the industry.
Innovations of the Librem 5:
- First easily-accessible hardware kill switches (3 switches for cellular model, Wi-Fi/Bluetooth and camera/microphone)
- First replaceable cellular modem and Wi-Fi/ Bluetooth (on two M.2 cards)
- First smart card reader (for 3FF OpenPGP card)
- First running 100% free/open source software on main CPU cores (PureOS with Linux/Wayland/GTK/phosh)
- First convergence based on downsizing existing desktop software with adaptive classes (libhandy) to use in both mobile and desktop PC (as opposed to convergence by using separate mobile and desktop software or upsizing mobile software)
- First to promise lifetime software updates, because designed to avoid planned obsolescence
Innovations of the PinePhone:
- First where all software is outsourced to multiple community OS projects
- Ported to more OSes (18) and more interfaces (8) than any other phone in history
- First physical switch to convert the headphone jack into a UART serial port
- First to promise 5 years of production when released
Linux phones have the potential to charge higher prices than Android phones and achieve good profit margins because they aren’t caught in the Android+Snapdragon commodity trap that has plagued the mobile industry for the last decade. They can follow their own independent course, just like Apple has done, and develop loyal customer bases who value them for their unique features that distinguish them from the rest of the industry.
Meanwhile, Huawei is continuing to develop its own hardware ecosystem that sets it apart from many of its competitors. Huawei has advanced in the design of its own phones. In recent years, it has led the industry in image processing, which allowed it to introduce phones with good night photo mode before Samsung, Samsung, OPPO, vivo and Xiaomi. It impressive just how much Huawei now controls of its own hardware ecosystem, as is shown by the following TechInsights teardown of the Mate 20 X (5G), which was released in July 2019. With the G play mini, 3 of its 17 ICs were designed by HiSilicon, but 19 of the 40 ICs in the Mate 20 X were made by HiSilicon, which is an indication of how much Huawei has advanced in 4 years of gaining control over its own hardware.
Looking at how Huawei has progressively replaced more and more of the ICs in its phones with its own chips, I have to question the US government’s decision to choke off Huawei’s access to American technology. Denying access to most high-end chips and cutting-edge fabs may work to control a company like ZTE, which quickly knuckled under to the US demands, but this strategy may backfire in the long run with a company like Huawei. Many of the smartphone makers have little capacity to plan and sacrifice in the short term to gain in the long term, because they don’t have the leadership nor the profit margins and capital to execute over the long term. Huawei, however, demonstrated with its Kirin processors, Balong modems, and the other chips in its phones and cellular base stations that it is willing to patiently and incrementally work to achieve its goals, even when it involves years of slow building that require many years of investment and high costs for the company.
Examination of the insides of Huawei’s phones shows that Huawei is relentlessly focused on controlling its own destiny and building up its own ecosystem. Samsung and Apple are famous for their control of their own ecosystems, but teardowns of their flagship phones show that they actually use a lower proportion of their own chips than Huawei. 50% (20 out 40) of the ICs in the Mate 20 X (5G) are designed by HiSilicon, whereas 40% (12 out of 30) of the ICs in the Galaxy S10+ are designed and fabbed by Samsung and 14% (5 out of 36) of the ICs in the iPhone 11 Pro Max are designed by Apple. Huawei is at a distinct disadvantage compared to Samsung, since it doesn’t make its own RAM, Flash memory and image sensors like the South Korean company, but in half a decade Huawei learned to make many of the components in its phones, whereas Samsung has been working on its components for close to 3 decades.
When I look at the teardowns of Samsung’s flagship phones, I see a company that is content to use parts from other companies which do that part better, so that Samsung can concentrate its efforts in the areas where it has a sizeable market share to defend. Samsung is increasingly acting more like American tech companies, that outsource as a cost-cutting measure and only focus on the profitable parts of the company. In order to compete with the low prices of Chinese brands at the low and mid-range of the market, Samsung decided in late 2019 to outsource the design and manufacture of 60 million of its phones per year to Chinese ODMs. It also decided to stop trying to customize the CPU cores in Exynos (like Apple and Huawei do with their processors) and now uses ARM’s standard cores. Maybe Samsung in the 1990s was like Huawei today, but today Samsung appears more interested in focusing on profits than control of its ecosystem.
In all likelihood, Huawei will take its losses in the smartphone market and retreat from countries which its telecom equipment is no longer welcome, but I doubt that the US government’s restrictions on Huawei will harm the company in the long term. Instead, the trade war is giving the Chinese state the impetus to build up its own independent semiconductor industry, which is free from technology controlled by American firms. Huawei doesn’t strike me as the type of firm that will give up. I don’t know whether the US government will eventually realize that it is doing more harm than good to its own semiconductor industry, but if the trade war continues, I suspect that it will eventually lead to a new technological cold war with every country being forced to choose sides between the dueling factions.
It is clear to me that the advisors to the Trump and Biden administrations who favored starting a trade war with China didn’t take the time to look at how Huawei designs their products. A company like Huawei that is relentlessly focused on replacing chip after chip with its own technology is not going to just roll over and give up because a foreign government makes demands and threatens it with loss of profits and market share. Huawei is currently taking steps to establish its own tech ecosystem, which is free of Western control, and the Chinese state is helping it by investing massively in creating a independent semiconductor industry that can create its own chips without using any American technology. As an advocate for software freedom and open hardware, I look askance at the kind of ecosystem that Huawei is creating. In fact, I ranked Huawei’s HiSilicon as the worst of the mobile chip manufacturers in terms of the respect for free/open source software and openness in general. It is pretty hard to be worse than Apple in that regard, but Huawei/HiSilicon manages to do it:
Like Apple, Huawei locks the bootloaders of all its devices and there is no way to gain root access for most of its devices except to pay for expensive proprietary cracks. Nobody but Huawei gets documentation for its HiSilicon Kirin processors. Unlike Apple which maintains and contributes to some FOSS projects, Huawei appears to have an active aversion to FOSS, although it is hard to know what is inside its Harmony OS. Huawei is not state-owned like UNISOC, but it has a history of ties with the Chinese military, and it is a facilitator of surveillance by the Chinese government.
Despite the fact that I don’t like the kind of separate tech ecosystem that Huawei is creating, it isn’t hard to see that the actions of the US government are pushing Huawei and many other Chinese tech companies toward greater independence and the creation of a separate Chinese tech sphere, that will be much harder for Western governments and tech companies to control in the long run.
In their own way, the new Linux phones, the Librem 5 and PinePhone, are also establishing their own type of independence, free from the existing tech giants that currently control the mobile industry. By slowly building up mobile Linux and an app ecosystem that is outside the reach of Google and Apple, Purism and PINE64 are charting a path of independence, that gives users a way to escape the control of the two reigning tech giants and their Surveillance Capitalism and “walled garden”. It will be many years before mobile Linux will have a software ecosystem that can provide close to the same functionality as the 3.1 million apps in the Google Play Store and 1.9 million apps in the Apple Store, but desktop Linux applications are rapidly becoming adaptive to mobile screens with the addition of libhandy/libadwaita and Kirigami classes.
By not using an integrated mobile SoC with an incorporated cellular modem, GNSS, WiFi, Bluetooth, charge controller, digital signal processor, image signal processor and neural processor, Purism and PINE64 are freed from many of the constraints that shackle the other phone makers. By using only free/open source drivers which the community can maintain and selecting chips with long production lives and long support cycles, Purism and PINE64 are freed to manufacture their phones for as long as they like and support them as long as they like. This longer production and support gives Purism the freedom to credibly promise lifetime software updates and freedom from planned obsolescence. It gives PINE64 the freedom to create a panoply of different devices (laptops, tablets, phones and SBCs) all based on the same processor, which can all share the same mods, so that PINE64 can create a vibrant ecosystem for hardware tinkers and mod makers. Fortunately, the new ecosystem being created by mobile Linux devices is far more open and respectful of the rights of users and doesn’t promote planned obsolescence, so it is a far more positive development for the world than the tech being produced by Huawei.
1. Currently it is possible to buy the Xperia X, XA2 and 10 with Sailfish OS preinstalled from JollaDevices, but they use Android drivers through libhybris. The Volla Phone and F(x)tec/XDA Pro1 X which are still in the crowdfunding stage will offer Ubuntu Touch preinstalled as one of the options, but they also use Android drivers through libhybris. The PinePhone and Librem 5 are the only phones currently on the market that use Linux drivers.