HP Professional sp750 Better Solutions for Parametric Technology Corporation's - Page 29

WHITE PAPER (cont.) A NOTE ABOUT W ORKSTATION BENCHMARKS Pro/E Benchmark Rating When Compaq submitted three Windows NT systems for Bench99 testing with Pro/ENGINEER, the following results were shown (as printed in ProE: The Magazine): § The XP1000 (667Mhz) with a Compaq PowerStorm 350 graphics card was the fastest Windows NT workstation and Fastest CPU § The AP200 with an Oxygen GVX1 graphics card has the best price/performance and best value § Because of improvements in design due to constant changes in technology, workstation performance continues to dramatically improve. Here's an example of performance improvements year over year using the exact same Bench99 test (lower is better): Compaq Workstation Performance Over Time 700 600 500 400 300 200 100 0 1999 seconds 2000 2001 (projected) Running Your Own Benchmark There is no better benchmark to compare performance than one that you create and run yourself using your own assemblies/parts and the Pro/E features you use most often. 1. When running a benchmark comparison, make sure you have enough memory installed to prevent paging to disk. Once a benchmark pages to disk, it becomes disk i/o bound, and will not give a true measure of overall system performance. Paging to disk occurs when the memory required to run the benchmark is greater than the amount of physical memory available. 2. For a quick estimate of how much memory a benchmark requires, run the benchmark right after a reboot. Go into the Windows NT Task Manager select the Performance tab, and look at the Peak Commit Charge (K) number. This is the maximum amount of memory used. This number should be smaller than the physical RAM you have on the workstation. If not, you are paging to disk and not getting an accurate picture of the overall workstation performance. (If you find that peak memory used is greater than the amount of RAM available a majority of the time, then you will want to consider increasing the amount of memory on the workstation.) 29