Yesterday we inspected Thingino, the open firmware that tries to turn cheap T-series cameras back into equipment owned by the person who bought them.

Today we inspect the silicon behind many of those cameras.

Ingenic.

Not a household name.

Not a keynote company.

Not the sort of vendor whose CEO wears leather on stage and tells investors civilization depends on one accelerator SKU.

Ingenic lives lower in the supply chain.

It appears on small boards inside IP cameras, doorbells, toys, old handheld devices, educational tablets, biometric terminals, and other devices that enter the world through rebrand factories and leave it through e-waste bins.

The public sees:

2K smart camera
AI human detection
cloud storage
works with app

The board sees:

T31
SPI flash
sensor ribbon
WiFi module
U-Boot
Linux 3.10
vendor driver tarball written during a long lunch

This is the camera chip regime.

I. The Company In The Board

Ingenic Semiconductor is a Chinese fabless semiconductor company based in Beijing. It was founded in 2005 and went public on the Shenzhen Stock Exchange’s Growth Enterprise Market in 2011.

Fabless means it designs chips but does not operate the fabrication plants that manufacture them.

Political translation:

Ingenic writes the constitution.

Someone else prints the passports.

The company’s public story emphasizes embedded processors, multimedia technology, AI technology, intelligent video surveillance, AIoT, industrial devices, consumer electronics, biometric identification, and education electronics.

That is the polite version.

The useful version:

Ingenic makes inexpensive embedded SoCs that end up in devices where cost, power, video encode, and “good enough Linux” matter more than software dignity.

Market phrase	What it means in the field
AIoT	small Linux devices with sensors and marketing
intelligent video	camera SoCs with ISP, encoder, and detection blocks
consumer electronics	plastic with firmware consequences
education electronics	tablets that age like milk
biometric identification	small machines that stare at citizens
industrial	the same chips, less colorful casing

The Supreme Leader respects any vendor that can hide in plain sight.

Ingenic did not need to become famous.

It became embedded.

II. XBurst: MIPS Refuses To Die Quietly

Ingenic’s processor story is tied to XBurst, its low-power CPU microarchitecture based around MIPS instruction set licensing.

This matters because many civilians think embedded Linux is just ARM wearing smaller shoes.

Incorrect.

Embedded Linux is a museum where architectures refuse retirement.

MIPS lived in routers, set-top boxes, game consoles, network appliances, printers, and camera SoCs long after desktop users stopped noticing it.

Ingenic understood this world:

small CPU
low power
cheap board
hardware video path
vendor SDK
Linux boots
ship product

That is not glamorous.

That is how millions of devices are born.

Architecture world	Civilian memory	Embedded reality
x86	PCs and servers	too much appetite for tiny cameras
ARM	phones and routers	dominant but not alone
MIPS	”is that still around?“	yes, in many older embedded kingdoms
RISC-V	the future pamphlet	slowly becoming present

Ingenic’s camera chips are not interesting because the CPU core wins benchmarks.

They are interesting because the whole chip does the camera job cheaply.

The CPU runs Linux.

The video blocks do the heavy lifting.

The vendor SDK does the paperwork badly.

III. The T-Series Dynasty

The camera world knows Ingenic through names like:

T20
T21
T23
T30
T31
T40
T41

The numbering has ambition.

It is crawling toward T-1000.

This is not speculation.

This is pattern recognition performed by the Ministry of Cinematic Threats.

Chip family	Role in the camera ecosystem
T20 / T21	early cheap camera boards
T23	newer low-cost camera boards
T30	transitional camera and doorbell territory
T31	widely seen in modern low-cost IP cameras
T40 / T41	newer AI video processor generation
T-1000	not released, but the naming department is clearly tempted

The chip name alone is not enough.

A camera board is a coalition:

• SoC
• image sensor
• flash chip
• WiFi module
• infrared LEDs
• IR-cut filter
• motor driver, if pan/tilt exists
• microphones and speakers
• board-specific GPIO mapping
• bootloader environment

Two cameras can both contain a T31 and still require different firmware.

The SoC is the province.

The board is the street address.

Firmware must know the address.

IV. The Camera SoC Is Not Just A CPU

Calling a T-series chip a processor is technically true and operationally incomplete.

An IP camera SoC is a little video state.

It needs to:

• initialize the sensor
• receive image data
• process raw pixels through the ISP
• encode video as H.264 or H.265
• handle audio
• talk to storage
• talk to WiFi or Ethernet
• expose control interfaces
• run Linux and the vendor application

The simplified path:

flowchart LR
    LIGHT["light"]
    SENSOR["image sensor"]
    ISP["ISP"]
    ENCODER["H.264 / H.265 encoder"]
    LINUX["Linux userspace"]
    NETWORK["RTSP / app / cloud"]

    LIGHT --> SENSOR --> ISP --> ENCODER --> LINUX --> NETWORK

The CPU does not manually compress every frame like a loyal clerk with a pencil.

The hardware video blocks exist because 1080p or 2K video at usable frame rates would otherwise turn the small CPU into soup.

This is why open firmware for cameras is harder than open firmware for many routers.

A router moves packets.

A camera must make photons politically acceptable before the packets exist.

V. Linux 3.10: The Preserved Official Corpse

Now we discuss the driver pile.

Many vendor camera SDKs in this world orbit old Linux trees, especially the Linux 3.10 era.

Linux 3.10 was released in 2013 and became a long-term kernel. Upstream long-term 3.10 eventually reached end of life.

In embedded vendor land, end of life means:

continue shipping
add patches
do not upstream
rename directory
ship SDK.zip

The drivers are often vibe-coded against Linux 3.10.

Not vibe-coded in the modern marketing sense where a chatbot generates a login page.

Vibe-coded in the old embedded sense:

the engineer kept editing until the sensor produced an image, then procurement declared victory, then the code became a platform.

Code area	What you hope for	What you may find
sensor driver	documented register sequences	magic arrays and comments from a disappeared engineer
ISP	open calibration model	binary tuning files and private APIs
WiFi	normal upstream driver	module pile plus firmware plus prayer
build scripts	reproducible build	shell archaeology
kernel	maintained upstream base	3.10-era vendor tree with patches
userspace	clear service design	cloud daemon with opinions

This is not unique to Ingenic.

This is embedded Linux as practiced by many vendors under cost pressure.

But Ingenic’s camera chips are common enough that the disease becomes visible.

The camera works.

The firmware ages.

The owner eventually asks why RTSP requires a cloud account.

Then the archaeologists arrive.

VI. Why Cheap Cameras Loved It

The low-cost camera business wants a very specific bargain:

video works
board is cheap
power is low
SDK exists
app vendor can integrate
ship before the next factory meeting

Ingenic fit that bargain.

Not because it was the most open.

Not because the drivers were beautiful.

Because the SoCs did enough of the hard camera work at a price the market wanted.

Requirement	Why it matters
low cost	cameras are sold like disposable appliances
integrated video pipeline	CPU cannot do everything alone
Linux SDK	vendors need quick productization
H.264/H.265	storage and bandwidth matter
sensor ecosystem	boards change constantly
low power	doorbells and small cameras care

The cheap camera is a triumph of integration.

It is also a warning:

when hardware becomes cheap enough, software quality becomes the first thing sacrificed to the margin gods.

VII. The ISSI Acquisition

In 2020, Ingenic completed the acquisition of ISSI, Integrated Silicon Solution, Inc.

ISSI is known for memory products used in automotive, industrial, communications, and other reliability-conscious markets: SRAM, DRAM, NOR flash, NAND flash, and eMMC categories.

This is not as funny as a T-series camera trying to become a Terminator.

It is more strategically interesting.

A fabless embedded SoC company acquiring a memory company is not random.

Embedded devices need memory.

Industrial and automotive customers care about long availability and reliability.

The Ministry translation:

Ingenic bought not only silicon knowledge, but supply-chain paperwork.

Before	After
embedded SoC designer	embedded SoC plus memory portfolio owner
camera and AIoT chips	broader industrial component story
processor identity	processor plus storage/memory adjacency

The camera in the hallway is cheap.

The company behind it is not thinking only about hallways.

VIII. Why Open Firmware Needs Ingenic Knowledge

Thingino exists because the hardware is useful and the vendor firmware politics are unacceptable.

But open firmware cannot replace board knowledge with slogans.

For a T-series camera, firmware must know:

• exact SoC family
• sensor model
• flash layout
• bootloader environment
• WiFi module
• GPIO assignments
• audio hardware
• motor control, if present
• update method
• recovery method

Useful inspection commands:

cat /proc/cpuinfo
cat /proc/mtd
uname -a
dmesg | grep -Ei 't20|t21|t23|t30|t31|t40|t41|sensor|isp|mipi'
fw_printenv 2>/dev/null | head

The important output is not “Linux boots.”

The important output is:

which chip
which board
which flash map
which sensor
which recovery path

Without that census, flashing firmware is not liberation.

It is roulette with a soldering iron nearby.

IX. Ingenic’s Real Sin

In the vendor hall of judgment, sins differ.

Vendor	Sin
Broadcom	hostility
Realtek	indifference
MediaTek	velocity
Intel	contradiction
NVIDIA	monopoly
Ingenic	obscurity

Ingenic’s sin is not being the loud villain.

It is being the quiet ingredient.

Its chips enter devices through manufacturers, modules, white-label brands, and consumer products whose buyers never learn what silicon is inside.

The company can be technically central and socially invisible.

That invisibility matters.

When a device breaks, the brand gets blamed.

When firmware is abandoned, the app gets cursed.

When open firmware developers start digging, the chip vendor’s old SDK becomes the basement archive.

The user never sees Ingenic.

The boot log does.

X. The Real Story (Suppressed)

Officially, T31 means a camera SoC family member.

Unofficially, the first roadmap was titled:

Terminator Progression For Domestic Optical Appliances.

The sequence was clear:

T20: observe hallway
T21: observe hallway with lower bill of materials
T23: observe hallway and report motion
T30: ring doorbell, notify cloud
T31: become common enough to require resistance
T40: add more ambition
T41: look directly at the T-1000 poster
T-1000: leave wall, impersonate vendor support

The Ministry confiscated the roadmap.

Not because it was dangerous.

Because the code comments still referenced Linux 3.10 and nobody wanted to admit the assassin was built on an EOL kernel.

In the suppressed Pyongyang lab, engineers attempted to interrogate one camera.

The camera replied:

CPU: T31
Kernel: 3.10.x
Cloud: enabled
RTSP: hidden
Sensor: maybe GC2053, maybe ask the board
Driver status: image appears, ship it

This was not intelligence.

This was confession.

XI. The Lesson

Ingenic matters because modern hardware ownership is often decided by companies users never see.

A low-cost camera is not just a lens and an app.

It is a chip vendor, a board maker, a sensor supplier, a WiFi module, a vendor SDK, a cloud service, and a brand pretending it designed the whole thing.

Ingenic made the T-series camera world cheap and common.

The vendor firmware ecosystem made it opaque.

Open firmware projects made it visible.

The lesson:

• the SoC is the province
• the board is the address
• Linux 3.10 vendor drivers are historical evidence
• the ISP is where camera freedom becomes difficult
• cheap hardware is never simple
• invisible vendors still shape owner control

The camera chip does not need fame.

It only needs to boot first.

In the Republic of Derails, all model numbers are audited before they approach T-1000.

— Kim Jong Rails, Supreme Leader of the Republic of Derails