The results are interesting, unlike with the OpenGL tests the MX can't beat the Parhelia.  We see that the MX does have a higher maximum frame rate, but without this spike the maximum frame rate would be about 79fps.  The lowest frame rate isn't very bad, and it manages to have a low about 18fps below that of its average.  The G4Ti4600-VT2D8X does very well but has some very wide ranges, from about 200fps to a low of about 63fps which is a very wide range.  Even without the high spike the maximum frame rate would still be 176fps, or about 59fps from the average frame rate.  The Parhelia has a completely different graph look, with no high spikes and only two low spikes which reach below 60fps.  The range that the Parhelia has is lower than that of the other video cards, which have a definite high frame rate spike.

For the first time the G4Ti4600-VT2D8X isn't leading the pack.  The MX is performing very poorly much like it did at 1600 in Serious Sam when AA and ansio are enabled.  The graph for the MX is almost flat, with only a small portion near the end of the demo having a little more variation to it.  The G4Ti4600-VT2D8X seems to have a fairly flat graph, with a couple of points near the end with two high spikes occurring  near the end of the demo.  The increase in memory speed does very little for the G4Ti4600-VT2D8X, apart from raising the minimum frame rate.  The Parhelia does surprisingly well in this test, more than likely due to its antialiasing technology.  There is two spikes near the end of the demo, just like that of the G4Ti4600-VT2D8X and MX, and no real low spikes to speak of.

The G4Ti4600-VT2D8X returns to the lead that it has had for most of the tests.  The MX is still in last place, about 10fps less than the nearest competitor, the Parhelia.  The minimum frame rate is a common occurrence on the graph, and in fact the graph is more like a flatter version of the graph from 1024 without any AA/ansiotropic filtering.  The G4Ti4600-VT2D8X loses a fair amount of its frame rate (about 50%) just like the MX card, but it still has an average frame rate near that of the 'magical' 60fps.  There are the two basic high spikes located near the end of the demo, which help the average frame rate improve.  The Parhelia has a graph that is flatter than before, with no real peaks in either direction, and the average frame rate is right between the lowest and the maximum frame rate measured.  Do the standings change when we add AA and ansiotropic filtering to the mix?

We can see a complete fill rate limitation on all of the video cards.  The MX has a terrible frame rate, losing 68% of its frame rate when AA/ansiotropic filtering is disabled.  Expectedly the graph is very flat, with the peaks coming at the very start and very end.  The G4Ti4600-VT2D8X is also completely fill rate limited, it is only slightly better than the MX in this case.  The loss of 77% of the performance of this card with AA and ansiotropic filtering enabled, is not very good to see on a card of this price range, as you will definetely not be able to play at 1600 with AA and ansiotropic enabled, and 1024 with AA and ansiotropic enabled is possible.  The Parhelia does better than the G4Ti4600-VT2D8X in this test, only losing about 35% of its performance compared to when AA and ansiotropic was disabled.  The frame rate graph is better than that of the other two cards, but it is still not able to provide playable frame rates at these settings.  Let us lastly look at how the MSI G4Ti4600-VT2D8X performs with the newest game in our test suite, UT2003.