From WikiChip
Kirin 970 - HiSilicon
< hisilicon‎ | kirin(Redirected from hisilicon kirin 970)

Edit Values
Kirin 970
kirin 970.png
General Info
DesignerHiSilicon,
ARM Holdings
ManufacturerTSMC
Model Number970
Part NumberHi3670
MarketMobile
IntroductionSeptember 1, 2017 (announced)
September 1, 2017 (launched)
General Specs
FamilyKirin
Series900
Frequency1,800 MHz, 2,360 MHz
Microarchitecture
ISAARMv8 (ARM)
MicroarchitectureCortex-A53, Cortex-A73
Core NameCortex-A53, Cortex-A73
Process10 nm
Transistors5,500,000,000
TechnologyCMOS
Die96.72 mm²
9.75 mm × 9.92 mm
Word Size64 bit
Cores8
Threads8
Max Memory8 GiB
Multiprocessing
Max SMP1-Way (Uniprocessor)

Kirin 970 is a 64-bit octa-core high-performance mobile ARM LTE SoC introduced by HiSilicon in mid-2017 at the 2017 IFA. This chip, which is fabricated on a 10 nm process, features four Cortex-A73 big cores operating at up to 2.36 GHz along with four Cortex-A53 little cores operating at up to 1.8 GHz. The 970 incorporates ARM's Mali G72 (12 core) IGP operating at 850 MHz and supports up to 8 GiB of quad-channel LPDDR4X-3732 memory.

Overview[edit]

Introduced at the 2017 IFA, the overall core organization is identical to the Kirin 960 which was introduced the previous year, but features 20% power efficiency and 40% smaller die area due to the process shrink. The 970 ballooned to over 37.5% more transistors from 4 billion in the 960 to 5.5 billion. The 970 adds many enhancements, including a more powerful Mali G72 GPU and incorporates a new Neural Network Processing Unit (NPU) designed for AI acceleration. The 970 has two improved ISPs and a more powerful LTE modem supporting up to User Equipment (UE) category 18 capable of reaching a maximum downlink of 1.2 Gbps (4x4 MIMO, 256 QAM, 3CC CA).

Cache[edit]

Main articles: Cortex-A53 § Cache and Cortex-A73 § Cache


For the Cortex-A73:

[Edit/Modify Cache Info]

hierarchy icon.svg
Cache Organization
Cache is a hardware component containing a relatively small and extremely fast memory designed to speed up the performance of a CPU by preparing ahead of time the data it needs to read from a relatively slower medium such as main memory.

The organization and amount of cache can have a large impact on the performance, power consumption, die size, and consequently cost of the IC.

Cache is specified by its size, number of sets, associativity, block size, sub-block size, and fetch and write-back policies.

Note: All units are in kibibytes and mebibytes.
L1$512 KiB
524,288 B
0.5 MiB
L1I$256 KiB
262,144 B
0.25 MiB
4x64 KiB  
L1D$256 KiB
262,144 B
0.25 MiB
4x64 KiB  

L2$2 MiB
2,048 KiB
2,097,152 B
0.00195 GiB
  1x2 MiB  

For the Cortex-A53:

[Edit/Modify Cache Info]

hierarchy icon.svg
Cache Organization
Cache is a hardware component containing a relatively small and extremely fast memory designed to speed up the performance of a CPU by preparing ahead of time the data it needs to read from a relatively slower medium such as main memory.

The organization and amount of cache can have a large impact on the performance, power consumption, die size, and consequently cost of the IC.

Cache is specified by its size, number of sets, associativity, block size, sub-block size, and fetch and write-back policies.

Note: All units are in kibibytes and mebibytes.
L1$256 KiB
262,144 B
0.25 MiB
L1I$128 KiB
131,072 B
0.125 MiB
4x32 KiB  
L1D$128 KiB
131,072 B
0.125 MiB
4x32 KiB  

L2$1 MiB
1,024 KiB
1,048,576 B
9.765625e-4 GiB
  1x1 MiB  

Memory controller[edit]

The Kirin 970 supports 4-channel LPDDR4X up to 1866 MHz. Each channel supports at most two ranks.

[Edit/Modify Memory Info]

ram icons.svg
Integrated Memory Controller
Max TypeLPDDR4X-3732
Supports ECCNo
Max Mem8 GiB
Controllers1
Channels4
Width16 bit
Max Bandwidth27.82 GiB/s
28,487.68 MiB/s
29.871 GB/s
29,871.498 MB/s
0.0272 TiB/s
0.0299 TB/s
Bandwidth
Double 13.91 GiB/s
Quad 27.82 GiB/s

Graphics[edit]

[Edit/Modify IGP Info]

screen icon.svg
Integrated Graphics Information
GPUMali-G72
DesignerARM Holdings
Execution Units12Max Displays2
Frequency746 MHz
0.746 GHz
746,000 KHz
OutputDSI

Standards
DirectX12
OpenCL1.2
OpenGL ES3.2
OpenVG1.1
Vulkan1.0
Hardware Accelerated Video Capabilities
Codec Encode Decode
Profiles Levels Max Resolution Profiles Levels Max Resolution
MPEG-2 (H.262) Main High 1080p (1920 x 1080)
80 Mbit/s or 60 fps
MPEG-4 AVC (H.264) Baseline, High 5.1 3840×2160
720p@240 fps
High 5.0 4K x 2K (3840 x 2160)
135 Mbit/s or 4K x 2K@30 fps
HEVC (H.265) Main Main 3840×2160
720p@240 fps
Main 5.1 4K x 2K (3840 x 2160)
160 Mbit/s or 4K x 2K@60 fps
VC-1 Simple, Main, Advanced M, H, 3 1080p (1920 x 1080)
45 Mbit/s or 60 fps
VP6 1080p (1920 x 1080)
50 Mbit/s or 60 fps
VP8 1080p (1920 x 1080)
50 Mbit/s or 60 fps
VP9 2 4K x 2K (3840 x 2160)
100 Mbit/s or 4K x 2K@60 fps

Wireless[edit]

  • LTE Modem
    • Up to User Equipment (UE) category 18
      • Downlink of up to 1.2 Gbps (4x4 MIMO, 256 QAM, 3CC CA)
      • Uplink of up to 150 Mbps (2x20MHz CA, 64-QAM)
  • Wi-Fi 802.11 ac Dual Band
  • Bluetooth 4.2
  • NFC
  • GPS / A-GPS / GLONASS / BDS

Expansions[edit]

Neural Network Processing Unit (NPU)[edit]

kirin 970 npu.png

The Kirin 970 incorporates a new Neural Network Processing Unit (NPU) designed specifically to be used as an AI accelerator. According to CEO Richard Yu, who also introduced the processor at 2017 IFA, the NPU uses up the die area of roughly half of the CPU while consuming 50% less power and performing around 25 times faster than a traditional CPU for tasks such as photo recognition. The NPU is said to deliver 1.92 TFLOPs (HP 16-bit) through 256 MAC/cycleS. While the exact architectural details of the NPU have been withheld, the NPU appear to be a licensed IP design from Cambricon Technologies.

Utilizing devices[edit]

  • Huawei Mate 10
  • Huawei Mate 10 Pro
  • Huawei Mate 10 Porsche Design
  • Huawei Mate RS Porsche Design
  • Huawei P20
  • Huawei P20 Pro
  • Huawei Nova 3
  • Huawei Nova 4
  • Honor V10 (Honor View 10)
  • Honor 10
  • Honor Play 2
  • Honor Note 10
  • HiKey 970
  • Honor 8 Pro
  • Honor Play

This list is incomplete; you can help by expanding it.

Documents[edit]

Bibliography[edit]

  • Huawei Kirin 970 Keynote, 2017 IFA
Facts about "Kirin 970 - HiSilicon"
base frequency1,800 MHz (1.8 GHz, 1,800,000 kHz) + and 2,360 MHz (2.36 GHz, 2,360,000 kHz) +
core count8 +
core nameCortex-A53 + and Cortex-A73 +
designerHiSilicon + and ARM Holdings +
die area96.72 mm² (0.15 in², 0.967 cm², 96,720,000 µm²) +
die length9.75 mm (0.975 cm, 0.384 in, 9,750 µm) +
die width9.92 mm (0.992 cm, 0.391 in, 9,920 µm) +
familyKirin +
first announcedSeptember 1, 2017 +
first launchedSeptember 1, 2017 +
full page namehisilicon/kirin/970 +
has ecc memory supportfalse +
instance ofmicroprocessor +
integrated gpuMali-G72 +
integrated gpu base frequency746 MHz (0.746 GHz, 746,000 KHz) +
integrated gpu designerARM Holdings +
integrated gpu execution units12 +
isaARMv8 +
isa familyARM +
l1$ size512 KiB (524,288 B, 0.5 MiB) + and 256 KiB (262,144 B, 0.25 MiB) +
l1d$ size256 KiB (262,144 B, 0.25 MiB) + and 128 KiB (131,072 B, 0.125 MiB) +
l1i$ size256 KiB (262,144 B, 0.25 MiB) + and 128 KiB (131,072 B, 0.125 MiB) +
l2$ size2 MiB (2,048 KiB, 2,097,152 B, 0.00195 GiB) + and 1 MiB (1,024 KiB, 1,048,576 B, 9.765625e-4 GiB) +
ldateSeptember 1, 2017 +
main imageFile:kirin 970.png +
manufacturerTSMC +
market segmentMobile +
max cpu count1 +
max memory8,192 MiB (8,388,608 KiB, 8,589,934,592 B, 8 GiB, 0.00781 TiB) +
max memory bandwidth27.82 GiB/s (28,487.68 MiB/s, 29.871 GB/s, 29,871.498 MB/s, 0.0272 TiB/s, 0.0299 TB/s) +
max memory channels4 +
microarchitectureCortex-A53 + and Cortex-A73 +
model number970 +
nameKirin 970 +
part numberHi3670 +
process10 nm (0.01 μm, 1.0e-5 mm) +
series900 +
smp max ways1 +
supported memory typeLPDDR4X-3732 +
technologyCMOS +
thread count8 +
transistor count5,500,000,000 +
used byHuawei Mate 10 +, Huawei Mate 10 Pro +, Huawei Mate 10 Porsche Design +, Huawei Mate RS Porsche Design +, Huawei P20 +, Huawei P20 Pro +, Huawei Nova 3 +, Huawei Nova 4 +, Honor V10 (Honor View 10) +, Honor 10 +, Honor Play 2 +, Honor Note 10 +, HiKey 970 +, Honor 8 Pro + and Honor Play +
word size64 bit (8 octets, 16 nibbles) +