So says Algor Product Manager Bob Williams. "Upgrading computer hardware can help maximize the return on CAEsoftware investments because simulations run on more-powerful hardware. And that lets users examine more design possibilities." The next time computer hardware is in the budget, Williams suggests looking for these capabilities:

Intel Xeon or AMD Opteron dual-processing or dual-core processors running at 3 GHz or more ($2,000 to $3,000 for a complete system). "Of course, a faster CPU means shorter simulation times. Workstations generally provide more power than garden-variety PCs. But if a PC makes financial sense, then select one with an Intel Pentium 4 or AMD Athlon processor running at 1.3 GHz or higher," says Williams.

Plan on 2-Gbytes RAM or more. ($500 buys about 2 Gbytes.) "More random-access memory contributes to shorter compute times for data-intensive applications such as CAE simulations. Given the low cost, get the most RAM supported by your computer, a minimum of 512 Mbytes," he says.

Plan on 30 Gbytes of free disk space or more (costs: $200 to $500). Engineering analyses create large files which must be stored on a computer's hard disk. Similar to RAM, maximize a system's data storage capability, at least 5 Gbytes of free disk space.

A graphics card should have 128 Mbytes or more of onboard memory with OpenGL accelerated graphics. Such cards run $100 to $1,800. "A dedicated graphics card, as opposed to a default card on the motherboard, is essential for fast dynamic viewing while building a model or examining results. However, you might get by with a midrange card with OpenGL support," he adds.

No one OS is best for engineering software. Standard Windows runs about $300 and provides easy-to-use graphical tools. Linux runs clusters well, and it's free, but $300 buys Red Hat support. And Unix, though a bit pricier at about $750 per processor, supports high-end hardware.

"Features such as distributed processing and remote submissions let engineers build finite-element models on standard PCs, send computationally intensive models to high-performance 64-bit computers or clusters of networked computers, and then examine results on standard PCs," says Williams.

True, hardware costs rise for each additional computer added to a system of distributed processors (networked computers that might be idle at night). "But considering the relatively low price for hardware, it makes sense to buy the units that get the most from a software investment," he says.

MAKE CONTACT
Algor Inc., (412) 967-2700
algor.com