First, we looked to two entertainment titles that commonly push our PC components to their brink of performance limits, Quake3Test v1.08 and Rage's Expendable Timedemo.

Each offers the flexibility in settings that would help us isolate the system components over the video accelerator and CPU, helping to show differences caused by the two memory types.

For our synthetic tests we chose a little old and a little new, beginning with the rapidly-progressing-towards-retirement Ziff Davis WinBench99 CPUMark99 synthetic test.

We had expected to receive WinBench2000 at Comdex, which is traditionally where Ziff-Davis makes their newest benchmarks available, but delays have put off the introduction of their newest component-level benchmark until next year.

Luckily Ziff-Davis did have a new benchmark we were particularly interested in that was completed in time for Comdex: 3D WinBench 2000.

The much maligned 3D WinBench series of performance benchmarks have never been on our favorites for video accelerator performance measurement, but for CPUs the benchmark now has a great Transformation and Lighting simulator that has been redesigned to use DirectX7.

Rounding out our test suite for comparing the VC820 to the CC820 are two benchmarks provided by Intel, which focus on the actual bandwidth needed for true AGP4X support. By throwing massive amounts of information across the system bus (two gigabytes) for a period of time the benchmark measures sustained (not peak) transfer rates for whatever platform it's run on.

While the argument over whether platform tests should favor peak data transfer rate simulations (which tend to favor DDR-SDRAM) versus sustained data transfer rate simulations (which tend to favor RDRAM) rages on, we're inclined to believe that both are important at this point towards achieving a well-rounded entertainment and business PC.

As more tests become available that accurately simulate all possible scenarios that a PC can encounter we'll continue to add and subtract from our own test suite.