well, things seldom scale linearly - I'd still like to see performance numbers for the CPUs that most of the market is likely to have. I interpreted 'cpu-bound' to mean that the CPU is doing serious work during the rendering process (or at least the game loop)
You are of course right that we can't really know until we see some CPU benchmarks (or better yet - hold a copy of the game in our hands
) - however, it should be possible to predict performance with a certain degree of accuracy from this GPU test.
First of all, in a rasterized game (like all modern games) the CPU cost of rendering a certain game frame only depends on the number of batches (number of objects drawn, to put it simply) as well as the number of render state changes needed to draw these batches. Pixel resolution, material quality, number of vertices and number of triangles only affects GPU performance.
If a game is CPU-bound it means that the CPU cannot process game logic and send the batches of graphics fast enough to keep the GPU busy. If a game is GPU-bound it means that the graphics card cannot process the render commands that are being sent to it fast enough.
You can see if a game is CPU-bound if changing resolution makes no difference whatsoever to performance (for example, just guessing from my part, try an older 3d game like Quake3 and use Fraps to see FPS in 1024x768 and 1600x1200). On the other hand, a game is GPU bound if performance changes drastically with resolution, shadows etc - even when nothing major is going on. Try NWN2 in a small location, like an inn..
A game can only run as fast as the weakest component - so since this report shows that the Radeon 5870 runs Risen at 60.6fps we can deduct that the i7 is *at least* powerful enough to render the game at 60fps. Actually, since we see a performance drop even on that graphics card when going from 1280x1024 to just 1680x1050, one could speculate that the game is indeed entirely GPU-bound.
From that I would expect that the 8800GTX is the limiting factor by far in this test, so the performance numbers listed for that card is probably fairly accurate for weaker CPUs as well.
But again, all this is just speculations - although speculations that I believe I have fairly sound arguments for (not to mention quite a lot of hands-on experience in the field)