The demo recorded on the Pier map shows a flight on a hang-glider over the island. Its speed is usually limited by the CPU performance when it comes to testing top-end graphics hardware. For less advanced solutions it is pixel shader processing speed that’s important here: there’s always water on the screen and Far Cry’s implementation of water is still among the best in the industry.
The ASUS does well in this test, making all resolutions playable with enabled 4x FSAA and anisotropic filtering, but it is slower than the Radeon X1800 XL. Note that there’s practically no performance gain in comparison with the standard GeForce 7800 GT because there’s only 20MHz of difference between the GPU clock rates of this card and of the EN7800GT Top Silent.
The demo record on the Research map is different. The action goes on indoors, in catacombs. Per-pixel lighting is used here, and shader code optimizations allow rendering the lighting of the scene in one pass with SM3.0 rather than in several passes as with SM2.0. Thus, the graphics card is required to quickly render the lighting and efficiently process version 3.0 shaders to pass this test successfully.
The ASUS EN7800GT Top Silent is a good match to the Radeon X1800 XL here, being 2-3fps ahead in high resolutions thanks to faster memory. The GeForce 7900 GT looks better than on the Pier level, but its 24 pixel processors still cannot help it overtake the Radeon X1800 XT. Like in the previous case, all the cards deliver an acceptable average frame rate in all resolutions with enabled 4x FSAA.
None of the participating graphics cards can give you the required average speed in 1600x1200, but all the solutions on Nvidia’s chips can do that in 1280x1024. The Radeon X1800 XT is very close to them, but cannot yield more than 55fps due to the poor implementation of HDR support for ATI cards in the current version of Far Cry. Note that the gap between the ASUS EN7800GT Top Silent and the GeForce 7800 GT remains constant here irrespective of the resolution.
The ASUS doesn’t have any advantage over the reference card on the Research map when HDR is enabled. In this demo it is the pixel shader processing speed that’s important and the two mentioned card don’t differ much in this respect whereas the memory subsystem performance plays no big role because the scene is small and closed. Like on the Pier map, the maximum resolution available with any card (except the Radeon X1800 XL) is 1280x1024.