Computer Hardware: Components, Operating Systems, and Everything Nice
One of the biggest concerns among FLOW-3D users is achieving accurate simulations. Another concern of equal importance is the time required to complete simulations. While much can be done to make simulations more efficient by choosing simulation parameters wisely, we suggest users run their simulations on the fastest computer hardware available. What should FLOW-3D users look for—and what can they expect to gain—by upgrading to the latest hardware? This note is intended to give some general guidance to users considering a change. For more specific advice on a given selection, please feel free to contact us.
The graph below shows the relative performance of four different hardware platforms: 3.2 GHz Pentium 4, 1.66GHz Core Duo, 1.8 GHz Opteron, and 2.4GHz Core 2 Duo. While the actual performance of any hardware can vary slightly from simulation to simulation (based on the type of numerical methods used), the relative performance will follow the trend shown below.
The most important hardware factor in improving the performance of FLOW-3D is the processor since it handles both number crunching and graphics. The fastest processors available are IBM's Power 5 series which are roughly 10 to 20% faster than the fastest Intel x86-64 processors. However, on a cost/performance basis, the x86-64 processors offer the best option for FLOW-3D users. Multi-core processors are available with up to 4 cores per processor. Due to memory issues, a dual-core processor provides the best cost/performance benefit at this time. Most computers available today are 64-bit processors which can run both 32 and 64 bit operating systems.
Most graphics card with OpenGL 2.0 support will work well with FLOW-3D. Look for graphics cards that specifically list OpenGL support for the operating system you're intending to use. Gaming graphics cards such as ATI's Radeon X-series and Nvidia's Geforce Series 6, 7, and 8 will perform well even for very large simulations. Workstation graphic cards such as ATI's FireGL and Nividia's Quadro cards will also work but usually cost more than gaming cards.
Both AGP and PCI-Express interfaces perform well. Onboard graphics such as Intel's 945GM Express Graphics can be expected to display graphics reasonably well for simulations up to a few million cells. For larger simulations, a dedicated graphics card with 512 MB of memory is recommended.
A minimum of 1GB of memory is recommended for running FLOW-3D on most modern operating systems. This will allow simulations with roughly 2 million cells to be run. The speed of the memory is usually an important factor in the simulation time but it is difficult to decouple the effect from the processor speed. Generally, it is recommended to buy the fastest memory available.
Both Windows and Linux are available in 32 and 64 bit versions. If the available memory on your hardware is 2GB or less, there is no benefit to running a 64-bit operating system. If your hardware has 3GB of memory (or more), running a 64-bit operating system will allow large (>6 million cells) simulations to be run. On 32-bit operating systems (Linux and Windows), simulations will be limited to roughly 3 to 4 million cells regardless of how much memory is available. On 64-bit operating
systems, the size of the simulation is only limited by the available memory on the computer.
There is no noticeable difference in simulation time on Windows vs. Linux. Users should choose whichever operating system they are comfortable with. Users should keep in mind that after Version 9.2, Flow Science will no longer support Unix operating systems.