Doom3 brings a fully shader enabled environment to OpenGL, utilising the functionality of OpenGL1.3 and above. Although the title makes use of elements made available in "DirectX 9 class" hardware, Doom3's primary rendering characteristic, being its unified lighting model, requires plenty of stencil fill-rate and a lighting shader that can be achieved in a single pass per-pixel light in DirectX8.1 equivalent hardware. Here we are using our "Turkey Baster" demo which has very high RAM requirements in Doom 3's Ultra quality mode.

Doom 3 isn't a particularly shader limited title, however thanks mainly to the increased shader processing capabilities of R580, the X1900 XT does have a 15% performance advantage over the X1800 XT. The performances under the boards here, though, are reasonably highly CPU bound, hence the increases for Crossfire are limited, with the X1900 Crossfire gaining only 17% over a single X1900 XTX board at 1600x1200.

With 4x FSAA and 8x AF enabled the rendering becomes more bandwidth and texture limited, cutting down the X1900's advantages to the X1800, however in this instance its only by 1% as there is still a 14% difference between the two at 1600x1200. Again the X1900 is able to render the timedemo at nearly twice the performance of the X850. As the performances in this instance are a little more graphics limited the Crossfire gains are greater here, with a 50% increase at 1600x1200 for the X1900 Crossfire over a single X1900 XTX.

With 6x FSAA and 8x AF there is a further 1% knocked off the performances between the X1900 XTX and X1800 XT because the test is further bandwidth and texture limited. The performance lead the X1900 XTX has over the X850 increases to over double at 1600x1200 in these rendering conditions. Because the test is more bandwidth limited with 6x FSAA Crossfire makes a bigger difference, with the performance being increased by 63% at 1600x1200 for the X1900 Crossfire in relation to a single X1900 XTX.