Results (continued):
Moving just a touch closer to reality, I ran all three generations of Futuremark’s 3Dmark programs to get an idea of how an increase in memory speed might affect the number of Marks each program would report.
First is 3Dmark2001. This generation of 3Dmark is quite old by now, but still shows some great increases when the memory speed is increased.
Next comes 3Dmark2003. Things got a bit tougher to render when Futuremark released this version of 3Dmark. The increases seem to stay right on par with the previous version.
Finally, the big daddy of them all, 3Dmark2005. This version of 3Dmark makes even the best of systems choke. Although during the second half of the testing frame rates drop to under 3 frames per second, I still managed to work through it and show the same kind of increase as in the last versions.
Bandwidth measurements and synthetic “Marks” are all fine and good, but how does that information translate into real world improvements in gaming? How about more frames per second. Again at each step of the climb in memory speed, I ran Far Cry, Half Life 2, and finally Doom 3 frame rate benchmarks to see how much of an improvement we might see.
For frame rate testing in Far Cry I used the Research level in v1.4.1 of HardwareOC’s benchmarking program, with all settings to Ultra, and at a resolution of 1024x768. As shown below, the increase in frame rate wasn’t as dramatic as the synthetic “marks”, but there was none the less an increase at each step.
Please read on to the next page for more... Next
Page 1 | Page 2 | Page 3 | Page 4 | Home | Forum | Review Index
|
|
