This page displays the same data as the previous one, but it also plots the HPCG performance of the supercomputers.
The LINPACK benchmark implements Gaussian Elimination, one of the principal method of Linear Algebra. However, the method can not be applied to large sparse matrices that arise in practically all computer simulations, because the system matrix becomes dense triangular during elimination, thus requiring too much memory.
Consequently, many authors doubt that LINPACK can give a good assessment of the behavior of a computer running simulations in, e.g., CFD (Computational Fluid Dynamics), NWP (Numerical Weather Prediction), and similar fields.
This is why J. Dongarra (one of the LINPACK authors),M. Heroux, and P. Luszczek developed a new benchmark, HPCG, which is an implementation of Conjugate Gradient Method, the basis for most algorithms for solving systems of linear equations with a sparse matrix.
The HPCG behavior of the top 10 systems is shown by red columns, and is very bad.