Basic Configuration

Our base configuration consists of every graphics option turned down to a minimum. If there is sufficient interest the configuration file can be posted but the main concern is the loss of key bindings. We will be working off of a resolution of 800x600. Raising the resolution will of course result in a performance hit as well as turning on Anti-Aliasing or Anisotropic Filtering. With all our options turned off we are primarily CPU bound as our framerates are within 2% of each other when running a GF4 Ti4200 compared to a GFFX 5700 Ultra. Be aware that our configuration will wipe your key bindings. Note the performance differences measured here took place on a GF4 Ti4200 64 Meg and a GFFX 5700 Ultra. The performance hit may be different on other cards. On a GF3 system, going from High Quality Textures to Maximum Quality textures was the difference between a playable game and a slide show.

Texture Settings (Character Textures, General Textures, Filtering, Bit Depth ~1% hit - 26%+)

The impact of texture quality is the most apparent when a higher setting is used on lower end cards. With our basic test case of a Geforce FX 5700 Ultra, changing the texture quality setting had very little performance difference whether it was set at lowest or extra.

Filtering type and Bit Depth were two other features that had very little impact on performance. Users of older cards that have a heavier penalty with trilinear filtering may notice more of a difference.

The picture on the left is environment textures and character textures are set to low. Environment textures include the ground, the sleeve of the arm holding the gun, and the gun itself. Notice the lack of detail on the ground and the gun in the first picture. Character textures are limited to the character models. Again there is a lot more detail present in the uniform of the soldier on the right as compared to the left.

Users with less than 128 Megs of RAM are advised to keep textures to High and not Extra. The Extra setting tends to cause a lot of thrashing over the AGP bus resulting in horrid performance. When we move to a level that is more texture heavy (mp_powcamp) the differences between Extra Detail and High Detail are immediately apparent on a 64 meg card. Check out these screenshots. 10 FPS compared to 180+

Wallmarks (~12%)

Wallmarks are the textures that get applied to walls when bullets or explosions hit them. There is a significant performance difference with wallmarks being on or off.

Ejecting Brass (~5%)

The shells you see popping out of the gun. They fly by fast, not to the side but from the gun to behind the player. Kind of a subtle effect with a pretty small performance penalty. This can probably keep this one on with little problems.

Lighting (Dynamic Lights, World Dynamic Light Quality, 0-14% hit)

When Dynamic Lights are set to none, the World Dynamic Light Quality setting does not change anything. Setting Dynamic Lights on Models Only, also has a very inperceptible difference in performance. When Dynamic Lights are set to All, we see a performance hit of 10% on Light Quality on the Fastest setting, 14% on Nicest.

Model Detail (7% - 24%)

Model detail is one of the biggest hits with the frame rate settings. The framerate hit is summerized before in the handy table. At maximum detail, there is a whopping 24% difference in framerate. Most of the differences in geometry seem to be limited to world objects like trees. I did not notice a difference in player models with regards to model detail levels.

The tree on the left has a Kate Moss kind of look without enough girth on its trunk. The amount of leaves and detail on the trees to the right are a lot more apparent.

Detail Level	Performance Hit
Minimum	0%
Low	3%
Normal	14%
High	19%
Maximum	24%

Sound (0-41%)

Our default setting was set to Miles 2D Low Quality. Raising it to High Quality resulted in no performance difference. This may change depending on the amount of memory you have. If it is on the low side, eg 192MB, there may be a bigger hit. On our system with 512 MB there was virtually no difference. Where the hit came in was changing to EAX/EAX2 with our onboard sound solution. This resulted in a drastic decline in framerate going down a whopping 37% with EAX1 and 41% on EAX2. I do not know if it is just because the Realtek Codec handles EAX that poorly or because of some other reason but I would suggest that EAX is not worth the performance penalty.