|
3DMark05 Preview DX9 Performance Benchmark |
|
Written by Mavke
|
|
Wednesday, 29 September 2004 |
TweakTown has a preview of the upcoming 3DMark05 benchmark. There screen shots at the site reveal a very detailed looking benchmark.
 3DMark05 Preview DX9 Performance Benchmark
Today FutureMark Corporation is releasing the latest version in the 3DMark series called 3DMark05. It is said to push the technology bar by exclusively using shader models 2 and 3 for all pixel and vertex processing which are complied using DX9c libraries, making 3DMark05 a true DX9 benchmark. This is different to 3DMark03 which had minimal DX9 tasks to perform since the majority of graphics cards back then only had DX8 or lower hardware support. Now all modern graphics cards support hardware DX9 for the latest graphics effects and 3DMark05 is here to give you an indication of how your hardware will stand up against the competition over the next year and a half until the next version is released probably sometime in 2006.
3DMark was never designed and never will be the absolute last word when it comes to 3D benchmarks for modern games. Although 3DMark05 does provide a good indication of how your system (mainly graphics card) will generally stand up in games (especially those coded with DX9 graphics programming techniques) to be released over the next year and a half at a range of different resolutions and graphics settings which are chosen inside the game as usual. If you want to know exactly how your system will perform under a specific game such as Doom 3, the best idea would be to check benchmarks from that actual game for the absolute real-world final word. The beauty of 3DMark is that it is good at estimating 3D gaming performance for you but at the end of the day, 3DMark05 is a synthetic benchmark and that should never be forgotten.
As you can see from the benchmarks we performed, 3DMark05 puts much more stress on the graphics card over 3DMark03 which mostly used older and less intensive DX8 graphics programming techniques. The most extreme result showed us a 160% decrease in the overall 3DMark score from the old version to the new. This is a good sign as 3D graphics over the next 18 months will only become more impressive and as a result, require more power from your GPU to render those graphics.
Just as we expected, CPU and memory subsystems dont play much influence on the overall 3DMark score in 3DMark05. We overclocked our test system processor by a little over 600MHz and we only saw an overall score increase of 5% at the most. This goes a long way to telling us that 3DMark05 (just like 3DMark03) is very much a graphics card dependent benchmark more than an overall system benchmark which would take into account such things as CPU, memory size, HDD, memory speeds and so on. Even though it is unable to deliver a real-world result, it does a nice job of providing a good generalized representation of 3D gaming performance through its three different game tests which cover a range of different game types.
Heck, even if you don't intend on using 3DMark05 as a benchmark, you'll have pleasure in running the splendid DX9 game tests or the demo for hours on end which includes sound (perfect for retail store owners to demo on their machines) over and over again as you marvel at the graphics which FutureMark Corporation use to help build the popularity of the 3DMark series with each new version. Now let's wait and see if 3DMark05 has any of the driver optimization cheating problems which 3DMark03 saw in the first few months it was released back in 2003.
|
NiBiTor v6.03