UNISOC T606 vs Google Tensor. Specifications, performance, tests

Background information about the processors being compared, series, generation and market segment.

• CPU group
  UNISOC 4G 12nm Google Tensor
• Family
  UNISOC 4G Google Tensor
• Generation
  0 1
• Segment
  Mobile Mobile

Core counts, threads, cache and clock modes describe the CPU configuration. Performance depends on the workload and comparable measured results.

• A-Core
  2x Cortex-A75 2x Cortex-X1
• A-Core Frequency
  1.6 GHz -
• B-Core
  6x Cortex-A55 2x Cortex-A76
• B-Core Frequency
  1.6 GHz -
• Core architecture
  hybrid (big.LITTLE) hybrid (Prime / big.LITTLE)
• CPU Cores
  8 8
• CPU Threads
  8 8
• Hyperthreading / SMT
  No No
• Overclocking
  No No
• Base Frequency
  - 2.8 GHz
• C-Core
  - 4x Cortex-A55
• Turbo Frequency (1 Core)
  - No turbo
• Turbo Frequency (8 Cores)
  - No turbo

Internal Graphics does not affect the performance of the CPU, performs the work of the graphics card in its absence or on mobile devices.

• DirectX Version
  12 12
• Execution units
  2 20
• Generation
  Vallhall 1 Vallhall 2
• GPU Frequency
  650 MHz 760 MHz
• GPU name
  ARM Mali-G57 MP1 ARM Mali-G78 MP20
• GPU Turbo
  No turbo No turbo
• Max. displays
  2 1
• Max. GPU Memory
  4 GB -
• Release date
  Q2/2020 Q4/2021
• Shader
  32 320
• Technology
  7 nm 5 nm

Built-in codecs used to encode and decode content. Significantly speeds up the required operations.

• AV1
  Decode Decode
• H.264 / AVC
  Decode / Encode Decode / Encode
• h265 / HEVC (8 bit)
  Decode / Encode Decode / Encode
• h265 / HEVC (10 bit)
  Decode / Encode Decode / Encode
• JPEG
  Decode / Encode Decode / Encode
• VC-1
  Decode / Encode Decode / Encode
• VP8
  Decode / Encode Decode / Encode
• VP9
  Decode / Encode Decode / Encode

Types, channel quantity of RAM supported by Google Tensor and UNISOC T606. Depending on the motherboards, higher or lower memory frequencies may be supported.

• ECC
  No No
• Max. Memory
  - 12 GB
• Memory channels
  0 2
• Memory type
  LPDDR4X-1600 LPDDR5-5500

• AES-NI
  No No

Compare the TDP requirements of TDP UNISOC T606 and Google Tensor Power specifications describe different manufacturer limits. TDP, base power and maximum turbo power are separate values; they do not establish measured system consumption.

• TDP (PL1, legacy label)
  - 10 W

Architecture, interfaces, additional instructions supported by UNISOC T606 and Google Tensor, virtual machine technologies and process technology.

• Documents
  Technical data sheet -
• Instruction set (ISA)
  ARMv8-A64 (64 bit) ARMv8-A64 (64 bit)
• L2-Cache
  - 8 MB
• Operating systems
  Android -
• Release date
  Q3/2021 Q4/2021
• Technology
  12 nm 5 nm
• Virtualization
  None None