This report compares the performance of various computer systems in solving dense systems of linear equations. It includes data from approximately 100 computers, ranging from supercomputers like the Earth Simulator to personal computers. The results reflect only one specific problem area: solving dense systems of equations. The data is collected over time and may change as new machines are added and hardware/software improves. The programs used to generate the data are available online. The report provides performance information on a wide range of computers, from home PCs to supercomputers. The first table reports three numbers for each machine: the performance of the LINPACK benchmark for a 100x100 matrix, the performance of solving a 1000x1000 system, and the theoretical peak performance. LINPACK programs are characterized by high floating-point operations, and the routines used in this study use column-oriented algorithms. The BLAS routines, which are used in the timing study, are Level 1 BLAS that reference one-dimensional arrays. The first case involves a small problem size (100x100) with no changes to the software. The second case involves a larger problem size (1000x1000) with possible algorithm modifications. The theoretical peak performance is calculated based on the cycle time of the hardware. The report also discusses parallel processing and highly parallel computing, and provides information on how to obtain the software and run the benchmarks. The report is updated periodically and can be obtained via email. The report includes tables with performance data for various computers, and the data is based on the performance of solving linear equations. The report also includes information on the accuracy of the solutions, with a residual check to ensure the solution is accurate. The report acknowledges the help of many people in putting together this collection.This report compares the performance of various computer systems in solving dense systems of linear equations. It includes data from approximately 100 computers, ranging from supercomputers like the Earth Simulator to personal computers. The results reflect only one specific problem area: solving dense systems of equations. The data is collected over time and may change as new machines are added and hardware/software improves. The programs used to generate the data are available online. The report provides performance information on a wide range of computers, from home PCs to supercomputers. The first table reports three numbers for each machine: the performance of the LINPACK benchmark for a 100x100 matrix, the performance of solving a 1000x1000 system, and the theoretical peak performance. LINPACK programs are characterized by high floating-point operations, and the routines used in this study use column-oriented algorithms. The BLAS routines, which are used in the timing study, are Level 1 BLAS that reference one-dimensional arrays. The first case involves a small problem size (100x100) with no changes to the software. The second case involves a larger problem size (1000x1000) with possible algorithm modifications. The theoretical peak performance is calculated based on the cycle time of the hardware. The report also discusses parallel processing and highly parallel computing, and provides information on how to obtain the software and run the benchmarks. The report is updated periodically and can be obtained via email. The report includes tables with performance data for various computers, and the data is based on the performance of solving linear equations. The report also includes information on the accuracy of the solutions, with a residual check to ensure the solution is accurate. The report acknowledges the help of many people in putting together this collection.