I'm not too big of a fan of 3dmark benches since they are synthetic and are not "real-world" performances. They can give you a good idea, but I'm not sold on the fact of those percentage increases translate to real FPS numbers.
Bald Eagle. Try looking up other benchmarks. I think the performance increase varies in different Benchmarks. I do like 3Dmark. Just about all serious gamers use it for some kind of benchmarking. It does score CPU and graphics and when you bench where individual games or utlities do not give independent scores. Some are more graphic intensive and some are more cpu intensive. I know you are much better aware of this than me.
Some people do not realize this is a important step in looking at various benchmarks to get the total performance overview of GPU's. Some like to make negative remarks toward 3dmark or not neccesarily negative, but maybe its that not a fair comparison. It is a valid comparison. It just gives different measuring stick to compare than all the rest.
Here is 3DMark and Crysis Benchmarks in Reg, SLI, Tri and Quad. This is really cool.
The test configuration included the processor intel core 2 Quad QX9650 (4,4 GHz), Motherboard EVGA 780i SLI , 2 X 2 GB memory OCZ reaper X PC6400 DDR-2 4- 4- 3-12 , HDD western digital raptor (150 GB, 10000 rp/min, SATA). Cooling system : water-block swiftech apogee GTX block (CPU) and MCP655- B pump. Finally , as a power unit we have here zalman 1000 W. All tests were executed under 64- bit Windows vista home premium using forceware drivers 174.53 and 174.40.
This overview was done by: DRDEATH! This hasn't been plagiarised from any article. I actually did the math too!(Check it in case)
What shows here to get double the performance in Crysis you need Tri- SLI but you do not get double in 3Dmark.
The 9800GT in single SLI was 42% faster on Crysis but only 28% faster in 3DMark and Unrealtournament only showed 20% faster.
So, although 3DMark may not be a "Real World" performance, it is a measuring stick that uses CPU and FPS to give benchmark score. It then gives the end user a tool to compare his/hers score to others using similar systems or even better systems. One can also look at various systems to help in upgrades and other features. The others score on FPS run on high end built systems and are "Measuring Sticks". They change Graphic Cards and then give the comparison. Also, 3DMark runs the benchmark with 2 cpu tests and 4 graphic games not just 1 game. It then gives you a TOTAL score on the benchmarks somewhat added together. When a single game gives a benchmark, it is kinda 1 dimensional? To me 3Dmark gives a more accurate account due to this factor and is more "Real World."
