The world of 3D graphics is a very complicated one and many potential graphics card customers donâ€™t want to spend their free time learning about the intricacies of the technology, they just want to know which card to buy. This is where technology journalists like myself come in â€“ we grab the latest hardware as soon as we can get our paws on it and run a series of tests in order to determine how it performs in relation to what else is out there. There are generally two established methods for testing graphics hardware â€“ you can run retail games on it and compare the performance, or you can use whatâ€™s sometimes known as a â€œsynthetic benchmarkâ€.
In the world of graphics card testing there have been few things more contentious over the past few years than synthetic benchmarks like 3DMark. Some journalists have refused to use 3DMark because â€œitâ€™s not a real gameâ€, citing that graphics manufacturers can tailor, adapt or optimise their hardware and drivers in order to produce good results. However, thereâ€™s nothing to stop a graphics manufacturer taking exactly the same approach to the most popular games â€“ itâ€™s not like it hasnâ€™t happened before!
Another advantage to a benchmark like 3DMark is that itâ€™s consistent, repeatable and comparable. That means that anyone whoâ€™s using a version of 3DMark can compare that score to any other, whereas using timed demos in a real game would require a copy of that demo, most of which are recorded by each individual.
But if I have had one criticism of 3DMark in the past, itâ€™s that it isnâ€™t always at the cutting edge of technology. Back in 2003 when DirectX 9 was new, 3DMark03 didnâ€™t really make too much use of the new features. This meant that 3DMark03 wasnâ€™t the best performance indicator for new hardware, although if you were still running older DirectX 7 or 8 games youâ€™d get a good idea of hardware capability.
By coincidence I bumped into Tero Sarkkinen (Executive Vice President of Sales and Marketing â€“ Futuremark) at IDF last March and I asked him when Iâ€™d see a benchmark that supported Shader Model 3.0 effects and was multi-threaded to take advantage of the new breed of dual-core chips. Tero told me that everything I had asked for was being worked on and as if by magic, today the fruits of that labour have broken cover in the shape of 3DMark06.
I got my hands on the final code for 3DMark06 a little while ago and Iâ€™ve been playing with it a lot, running it on different hardware setups with different settings and comparing results. My initial reactions to the benchmark were â€œThis looks amazingâ€ followed by â€œthis is really hammering the hardwareâ€. One thingâ€™s for sure, I canâ€™t criticise 3DMark06 for not using cutting edge features â€“ Tero has been as good as his word, and then some.