Skip to main content

Table 2 Latency benchmarks of evaluated soft-output MIMO detection algorithms for a \(2 \times 2\), 64-QAM scenario. B represents the batch size during parallelized inference. All values are expressed in milliseconds and represent average execution times for the entire batch

From: Joint high-dimensional soft bit estimation and quantization using deep learning

  EQ-Net EQ-Net NN-Det NN-Det OAMP- MMSE- ZF- SD
  (L) (P) (L) (P) Net2 SIC SIC  
Latency CPU [ms], \(B = 1\) 1.511 1.903 2.269 2.689 2.125 0.603 0.559 0.6935
Latency CPU [ms], \(B = 16\) 1.767 2.292 2.565 3.073 2.376 0.831 0.787 4.547
Latency GPU [ms], \(B = 8192\) 6.523 8.767 12.61 15.91 25.94 7.907 5.311