Fig. 2From: How Amdahl’s Law limits the performance of large artificial neural networksThe dependence of the efficiency E on the number of the cores and the value of \((1-\alpha _{\mathrm{eff}}^{X})\) , as described by Equ. (4). For comparison both the corresponding measured \((1-\alpha _{\mathrm{eff}}^{X})\) values of the first five computers (as of 2018 November) are also displayed, as measured by the benchmarks HPL and HPCGBack to article page