Competitive Benchmarking Tips and Tricks - Graphic Processing Tweaks
(Page 4 of 5 )
Graphic Processing Tweaks
Tweaks in this section mostly apply to "graphic processing" benchmarks such as FutureMark 3DMark01SE, 2003, 2005, 2006, Aquamark, and others. These benchmarking utilities simulate a game play load on your system that stresses your graphical processing unit. Then they determine the average frame rate, measured in FPS (frames per second), and calculate a score based on those results. Basically, you need to maximize performance to get a higher FPS. These tweaks will definitely help. Experiment a lot with different combinations to find the perfect match for your system.
First tweak: This is the near famous "LOD-tweak." LOD stands for Level-of-detail. We're going to use specific software to tweak our LOD to different values. One software that I fully recommend is RivaTuner (get it here). Install it and then go to Power User -> Global -> Nvidia (or ATI depending on your setup) -> LOD Bias Range. Experiment a lot with this setting because one LOD setting can provide a significant boost for certain tests while another setting may work better for others.
Always tweaking to the perfect combination or running tests separately can yield better results in benchmarking. Here are several recommendations I've found that work quite well. The following values were tested on several systems and they're optimized to 3DMark01SE. They should also work on benchmarking applications but the gains may be unnoticeable.
Game Test #1 - "Car Chase" - 'Low' - LOD Value: 2
Game Test #1 - "Car Chase" - 'High'- LOD Value: 5
Game Test #2 - "Dragothic" - 'Low' - LOD Value: 5
Game Test #2 - "Dragothic" - 'High'- LOD Value: 5
Game Test #3 - "Lobby" - 'Low' - LOD Value: 3.5
Game Test #3 - "Lobby" - 'High'- LOD Value: 3.5
Game Test #4 - "Nature" - 'N/A' - LOD Value: 4
The values in the table above worked best for me on several setups. Keep in mind that these results are relative and fluctuate from test to test. You may want to try several LOD values in the 1-10 range and see which ones yield the best performance. I've noticed that going higher than 5 usually degrades performance.
Second tweak: In your graphic card's control panel, disable "Antialiasing" and "Anisotropic filtering" and set "Image Quality" to "Best Performance." Play with "Vertical Sync" too. Disabling it will give better overall performance but some tests actually require VSync to be enabled, so experiment.
Third tweak: Running the game tests one-by-one and not the whole benchmarking set at once increases performance quite a bit because your GFX card can cool down (wait 1-2 minutes between tests) and certain test combinations yield better results. Two famous combinations are well-known. One is from "OPPAINTER" (XtremeSystems) and the second one's source I don't remember. Check them out...
First Combination: DL - LL - CH - DH - LH - N - CL
Second Combination: LL - LH - DL - DH - CH - CL - N
DL - Dragothic Low; DH - Dragothic High
CL - Car Chase Low; CH - Car Chase High
LL - Lobby Low; LH - Lobby High; N - Nature
Use the "Change" -> "Custom" menu panels to customize and run the tests individually. As a side note, try to run the "Car Chase Low" test as many times as possible; it's the most sensitive test and basically scores (FPS) fluctuate all the time. Run it at least 5 or 8 times and save the results individually to reach your maximum.
I've focused primarily on 3DMark01 because it is still one of the most popular and commonly used benchmarking applications. It's easier to tweak and depends more on overall system performance rather than just pure GPU processing. Newer versions of 3DMark tend to focus more strongly on "pure" graphic card benchmarking (more intensive calculations are done on GFX cards than on the CPU). Anyway, feel free to try the tweaks on other benchmarking apps. Hopefully you'll end up with better results.
Next: Final Words >>
More Hardware Guides Articles
More By Barzan "Tony" Antal
