According to John Fruehe of AMD in his latest blog.
- AMD Opteron processors: 11 (44%)
- Intel Xeon processors: 8 (32%)
- IBM PowerPC processors: 4 (16%)
- Fujitsu SPARC processors: 1 (4%)
- AMD Opteron + IBM Cell: 1 (4)
In the CPU-only realm only Sun SPARC has scored higher on the industry standard LinPack benchmark.
From the site:
LINPACK is a collection of Fortran subroutines that analyze and solve linear equations and linear least-squares problems. The package solves linear systems whose matrices are general, banded, symmetric indefinite, symmetric positive definite, triangular, and tridiagonal square. In addition, the package computes the QR and singular value decompositions of rectangular matrices and applies them to least-squares problems. LINPACK uses column-oriented algorithms to increase efficiency by preserving locality of reference.
LINPACK was designed for supercomputers in use in the 1970s and early 1980s. LINPACK has been largely superceded by LAPACK, which has been designed to run efficiently on shared-memory, vector supercomputers.
AMD has been a staple of the Top500 since Opteron was released in 2003. Cray has created the most scalable HPC architecture including the Gemini interconnect that is said to scale out to HALF A MILLION CPU cores.
Upcoming versions of AMD Cray SCs are the XK6, their first truly hybrid SC made by combining AMDs Opteron 6200 (Codenamed Bulldozer) processors and NVidia’s Tesla X2090 (Codename Fermi) GPUs. Cray estimates this to provide up to 50PF (PetaFlops) of LinPack performance.
Users of the Cray systems include the NOAA (National Oceanic and Atmospheric Administration), the Lawrence Livermore Labs; the Oak Ridge National Lab; the Finnish Meteorological Institute and many others as these installations lease out “rack space” too many universities around the world.
The 6200 Opterons look to continue this trend as the NOAA reported the latest upgrades to their XE6 system will take it from 260TFlops to 386 TFlops using Interlagos 16 core CPUs. A new XE6 will be added natively based on Opteron 6200 processors and will achieve a theoretical 720TFlops. Cray’s current champ, the Jaguar achieves 1.7PetaFlops of performance, slightly behind the 2.4PF of the Tianhe GPU Hybrid but well behind the new leader, the SPARC based K Computer in Japan’s Riken Institute, which scores 8.1PF, though it does so with 2X the cores and power.