Neural Network Inference Latency on ARM Embedded Processors
0.071Latency (ms)Deterministic NN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 0.071 | — | |
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 0.186 | — | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 0.23 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 0.341 | 901.8 | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 0.347 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 0.742 | 990.2 | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 0.754 | — | |
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 0.933 | — | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 1.134 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 1.2 | 255.7 | |
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 1.64 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 1.89 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 3.131 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 3.16 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 3.75 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 3.897 | 205.6 | |
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 4.59 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 4.726 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 4.989 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 5.182 | 149.6 | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 6.565 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 6.89 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 9.524 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 9.63 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 10.022 | 119.4 | |
| Deterministic NNArchitecture=MLP, batch-size=10, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 14.02 | — | |
| PFPArchitecture=MLP, batch-size=10, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 15.26 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 15.358 | — | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 16.3 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 18.6 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 21.14 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 21.23 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 31.51 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 35.159 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 36.33 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 41.697 | — | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A76, Tuning Status=tuned2025.11 | 45.039 | 55.3 | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 45.81 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 70.08 | — | |
| PFPArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 76.09 | — | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A72, Tuning Status=tuned2025.11 | 116.33 | 23.3 | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A76, Tuning Status=not tuned2025.11 | 119.76 | — | |
| Deterministic NNArchitecture=MLP, batch-size=100, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 137.8 | — | |
| PFPArchitecture=MLP, batch-size=100, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 147.61 | — | |
| Deterministic NNArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 209.28 | — | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A72, Tuning Status=not tuned2025.11 | 240.94 | — | |
| SVIArchitecture=MLP, batch-size=100, Processor=Cortex-A762025.11 | 306.89 | — | |
| SVIArchitecture=MLP, batch-size=10, Processor=Cortex-A762025.11 | 307.52 | — | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A53, Tuning Status=tuned2025.11 | 383.68 | — | |
| SVIArchitecture=MLP, batch-size=10, Processor=Cortex-A722025.11 | 734.74 | — | |
| SVIArchitecture=MLP, batch-size=100, Processor=Cortex-A722025.11 | 775.32 | — | |
| PFPArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A53, Tuning Status=not tuned2025.11 | 801.33 | — | |
| SVIArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A762025.11 | 801.4 | — | |
| SVIArchitecture=LeNet-5, batch-size=10, Processor=Cortex-A722025.11 | 1,196.42 | — | |
| SVIArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A762025.11 | 2,488.73 | — | |
| SVIArchitecture=LeNet-5, batch-size=100, Processor=Cortex-A722025.11 | 2,708.16 | — |