Fork me on GitHub
We are now running a linear algebra performance benchmark; the results are plotted below. As we move right within each test, the matrix size increases.

Performance (MFLOPS; higher is better).
Geometric mean of scores: