2021DSPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28nm 1P8M CMOS Process3.6mm x 3.6mm0.72V ~ 1.1V3D Object Recognition ProcessorMaximum: 250MHz440mWJul. 2021
2021Neuro-CIM
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28nm 1P8M CMOS Process3.2mm x 0.9mm1.1VNeuromorphic Network Acceleration in Mobile DevicesMaximum: 200MHz105.44mW ~ 241.44mWJun. 2021
2021SNPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28nm 1P8M CMOS Process3.2mm x 1.9mm0.63V ~ 1.1VSpike-Neural-Network Acceleration in Mobile DevicesMaximum: 200MHz0.31mW ~ 186.84mWJun. 2021
2020PNNPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased Date65nm 1P8M CMOS Process4mm x 4mm0.78V ~ 1.1VMobile Intelligent 3D Vision NPU50MHz ~ 200MHz331mWAug. 2020
2020HNPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28nm 1P8M CMOS Process 3.6 mm x 3.6 mm 0.58~1.04VMobile DNN Inference & Training + Privacy-preserving Cloud-edge Co-trainingMaximum: 250 MHz1.95 mW ~ 1162 mWJul. 2020
2020OmniDRL
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased Date28nm CMOS Process3.6 mm x 3.6 mm0.68V ~ 1.1 VDeep Reinforcement Learning~250 MHz 3.1 mW ~ 282.8 mW (Peak)Aug. 2020
2020MAC
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28nm FDSOI2.7mm x 2.7mm0.3 ~ 1V (Analog), 1V (Digital)ECG Arrhythmia detection1kHz ~ 30kHz (Analog), 100MHz (Digital)1uWJan. 2020
2019EIT
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased Date65nm 1P8M CMOS Process4.0 mm x 4.0 mm3.3 V (Analog), 1.8V (Digital)Electrical Impedance Tomography10 kHz ~ 10 MHz9.6 mW/ChJuly. 2019
2019GANPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased Date65 nm 1P8M CMOS Process8.1 mm x 4.0 mm0.70 V ~ 1.1 VGenerative Adversarial Network Inference & TrainingMaximum 200 MHzMaximum 647 mWJul. 2019
2018LNPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased Date65nm 1P8M CMOS Process4mm x 4mm0.78V ~ 1.1VDNN Learning50MHz ~ 200MHz43.1mW ~ 367mW(Peak)Jul. 2018