Back to main homepage Background of current staff Download electronic publications Download sample codes Search the website

CPU Benchmark Data

This page presents data for the STARS CFD benchmark. The benchmark is intended to provide information about the relative speed of different processor, operating system, and compiler combinations for a single-threaded, floating point, computationally intensive CFD code.

The newer euler3d multi-threaded benchmark is presented on a separate page.

Description of the Benchmark Testcase

The benchmark testcase is the AGARD 445.6 aeroelastic test wing. The wing uses a NACA 65A004 airfoil section and has a panel aspect ratio of 1.65, taper ratio of 0.66, and a quarter-chord sweep angle of 45�. This AGARD wing was tested at the NASA Langley Research Center in the 16-foot Transonic Dynamics Tunnel and is a standard aeroelastic test case used for validation of unsteady, compressible CFD codes.

The CFD grid consists of 67,435 nodes and 366,407 tetrahedral elements. Figure 1 shows a close-up of the CFD surface grid.

STARS CFD Benchmark: Tetrahedral Surface Grid
Figure 1: Surface grid for AGARD 445.6 aeroelastic test wing.

The benchmark executable advances the AGARD solution 200 time-steps. Canonical benchmark results are a cycle frequency obtained by dividing 200 by the total runtime in seconds. A pdf file (benchmark.pdf) and a MS Excel file (benchmark.xls) are available to view the raw numbers. Also, please send us an E-mail ( to report your benchmark result. We appreciate all submissions!

Benchmark Results:

STARS CFD Benchmark: Results

Figure 3: STARS cycle frequency versus CPU frequency.

Discussion of Results

Benchmark results tend to lie on three curves. The lower curve fits 32 bit Intel Pentium 2 through Pentium 4 and 32 bit AMD Athlon systems with a slope of about 0.25 Hz per GHz. The Pentium 4 Netburst architecture scales linearly with slight positive curvature with frequency. AMD chips tend to show more identifiable chip families.

The upper curve fits 64 bit AMD and Intel Processors with a slope approximately 0.5 Hz per GHz. The Intel Pentium M is the first Intel chip to lie on the upper curve (thus matching AMD's performance per clock cycle). The Core 2 Duo family (one data point) lies on third slope of 0.74 Hz per GHz. Interestingly, the 64 bit DEC Alpha chip has a (one data point) slope of just under 1 Hz per GHz.

Memory bandwidth is important. The Pentium 4's RAM (RDRAM at 2.53 GHz, and 667 MHz DDR2 at 3.73 GHz) and the Athlon's DDR RAM (at 1400 MHz) improved the benchmark's performance. L2 cache size is positively correlated with benchmark results. A low performance memory system as seen in the Celeron chip gives poor benchmark results. Increased front side bus (FSB) frequencies likely contribute to the super-linear Pentium 4 behavior.

The RISC 64 bit DEC Alpha chip still shows strong performance considering its introduction in 1992 precedes the current x86 heritage 64 bit chips by more than a decade. On a CPU clock frequency scale, the DEC Alpha still has almost twice the performance of modern 64 bit chips. The RISC 32 bit IBM RS 6000 has similar characteristics. The King is dead, long live the King.

The current best is a Intel Core 2 Duo 6700 at 2.66 GHz with 800MHz DDR2 memory. We strongly recommend the Core 2 Duo family.


If you have an x86 compatible PC and want to see how your machine compares to those on this list, download the files provided below to run the benchmark testcase. The benchmark1.exe executable is a hard-wired version of STARS steady that will solve one and only one problem. You'll also need the agard2.geo file, which is a large binary data file defining the agard2 geometry.

A benchmark spreadsheet, in MS Excel format, is available to view the raw numbers. Also, please send us an E-mail to report your benchmark result. We appreciate all submissions!

Benchmark Links

If you want to find more benchmark data regarding the performance of various processors, we recommend that you check-out the SPEC benchmark suite and related performance data at

Oklahoma State University