C # performance

For fun, I want to show the community a graph that demonstrates the performance of C # for matrix operations, compared to C ++, C ++ with OpenMP, and also Parallel Extensions. X axis - the size of the matrix. On the y-axis, the execution time of the multiplication operation in seconds (log scale).



What does a respected community think about this topic? Does it make sense to develop performance-critical applications on .NET? Or should we invest more energy in the study of Intel-evskie bodies ? (Intel TBB, MKL, IPP)? Or just use OpenMP ?
')
PS: Picture taken from my blog. Full essay (in English) here .