2024IRIS
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm Logic CMOS4.5 mm × 4.5 mm0.7 ~ 0.9 VReal-time 3D Rendering AcceleratorMaximum : 200 MHz584 mWAug. 2024
2024BROCA
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm LPP4.5 mm × 4.5 mm0.7 ~1.1 VMobile social agent SoCMaximum: 200 MHz52.4 mW (@ 50MHz, 0.7V), 559.2 mW (@ 200MHz, 1.1V)Aug. 2024
2024EdgeDiff
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS4.5 mm × 4.5 mm0.68 ~ 1.0 VEfficient few-step diffusion acceleration50 ~ 250 MHzGeneration Energy 418.4 mJ/InferenceAug. 2024
2024Slim-Llama
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS4.5 mm × 4.5 mm0.7 ~ 0.9 VUltra-low-power LLM acceleration25 ~ 200 MHz4.69 mW (@ 25MHz, 0.58V), 82.07 mW (@ 200MHz, 1.0V)Aug. 2024
2023Space-Mate
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS Process4.5mm × 4.5mm0.7V ~ 0.9VNeural rendering-based SLAM processorMaximum: 200MHz114.1 ~ 623.4 mWAug. 2023
2023C-Transformer
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS Process4.5mm × 4.5mm0.7V ~ 1.1VComplementary DNN-Transformer / Spiking-Transformer ProcessorMaximum: 200MHz47.5mW ~ 469.2mWAug. 2023
2023NeuGPU
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS Process4.5mm × 4.5mm0.68V ~ 0.9VNeural Graphics Processing UnitMaximum: 200MHzModeling: 559.5 mW / Rendering: 728.4 mWAug. 2023
2022C-DNN
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS Process4.5mm × 4.5mm0.7V ~ 1.1VComplementary DNN ProcessorMaximum: 200MHz~40mWAug. 2022
2022MetaVRain
TechnologyChip SizePower SupplyFunctionOperating FrequencyPower ConsumptionReleased DateSamsung 28 nm 1P8M CMOS Process4.5mm × 4.5mm0.6V ~ 0.95VNeRF-based 3D Neural Rendering Processor50MHz ~ 250MHz133mW ~ 899mWAug. 2022
2022DynaPlasia
TechnologyChip SizeFunctionPIM CellsPeak ThroughputReleased DateSamsung 28nm 1P8M CMOS Process4.5mm × 4.5mmeDRAM IMC-based Mobile DNN Inference9600Kb19.5 TOPS (4b-5b)Aug. 2022