World in Conflict

Publisher: Sierra

We regard World in Conflict as one of the best real-time strategy games we've ever played. It's based on Microsoft's DirectX 10 API and, in collaboration with Nvidia's The Way It's Meant To Be Played developer support team, it incorporates some DirectX 10 specific graphics effects.

The first of these is a soft particle effect that removes the banding often found in particle effects like smoke, explosions, fire and debris - the effects simply didn't exist in the 3D world; instead, they were merely an add-on. With DirectX 10, the edges of the particle effects are much softer and banding is almost non-existent as the effects now interact with their 3D surroundings, as they're actually a part of the 3D world.

Additionally there are global cloud shadowing and volumetric lighting effects in the DirectX 10 version of the game. The latter is often referred to as 'god rays' and its implementation in World in Conflict interacts with the surroundings incredibly well. On the other hand, the former is where clouds cast shadows on the rest of the environment and, because all clouds in World in Conflict are volumetric and dynamic, the shadows cast by the clouds are rendered dynamically in DirectX 10 - they adjust in relation to the size, shape and orientation of the cloud in relation to the light source.

For our testing purposes, we used a full retail copy of the game and patched it to version 1.007, which includes a few fixes and some improved performance under DirectX 10. We used a manual run through from the Invasion level, which incorporates all of the effects we've discussed above. We chose not to use the built-in benchmark because it's largely CPU-limited. We used the "very high" preset, and controlled anti-aliasing and anisotropic filtering via the advanced settings tab.







World in Conflict is a game well known for scaling very poorly on multi-GPU set ups, and this proved to be no exception with the 9800 GTX+ in SLI and HD 4850 in CrossFire. While performance for both was markedly improved over a single card, with in some cases a frame rate improvement of nearly fifty percent, neither setup is able to convincingly surpass the powerful single GPU based cards until anti-aliasing was enabled at 2,560 x 1,600.

Here, the proximity and similarity of the results from the ATI cards at the top of the table indicate that the benchmark is now CPU and not GPU limited, although the 9800 GTX+ once again suffered the consequences of running out of memory bandwidth, posting a very disappointing result of less than half a single 9800 GTX+.