This test is similar to the results seen at 1600 with Jedi Knight.  The MX does fairly well and has the least swing in the group, of 73 fps.  This shows in the graph with the MX giving the 'smoothest' graph results, with only two major spikes shown, and three or four major down spikes noticed.  The G4Ti4600-VT2D8X has the largest fluctuation of the group, with a 136-140fps difference between the highest and the lowest points.  There are two very major spikes (giving it the ~200fps maximum), which last for a second and then plummet, the lowest points are at a spot that is one of the most demanding, with multiple players on the screen at the same time in an open area.  We will see that point being the lowest, or one of the lowest, in all the Serious Sam results we see today.  The Parhelia has a lower swing than the Ti4800 but also has the lowest minimum frame rate of the group, at 35fps, which coincidentally occurs at the same spot as the G4Ti4600-VT2D8X 's minimum.  There are two to three major upper spikes in the results which give the maximum frame rate of 132, otherwise it would have been 92-93fps.  Does AA and ansiotropic filtering change anything?

The MX takes another nose dive when AA and ansio is enabled, it simply doesn't have the fill rate to take these features without losing a step.  The graph is very flat, with very few differences in frame rate and no major spikes, either upward or downward.  The G4Ti4600-VT2D8X fairs better, it only loses about 38% of its non AA/ansio frame rate, and manages to tone down the major upward spikes that we saw at 1024.  There are about four or five downward spikes but these aren't too much, at most losing about 35fps from average.  The Parhelia fairs better in this test than it has without AA enabled, thanks to FAA.  It doubles the frame rate of the MX but can't reach the G4Ti4600-VT2D8X, in fact it is about halfway between the G4Ti4600-VT2D8X and the MX.  There is still the two major upward spikes in this test, with all the other results being within 20-25fps of the average frame rate.  How do the cards fair at 1600?

There are the two 'packs' here, in one is the G4Ti4600-VT2D8X with a healthy 65+fps and in the other the Parhelia and MX have just over half of that frame rate, about 35+fps.  The MX doesn't have a graph that is as flat as that of this card with AA enabled, with two high spikes and one moderately high spike, and a couple small low spikes.  The G4Ti4600-VT2D8X has the two major upward spikes and without them would have a maximum frame rate of about 97 (overclocked) to about 87 un-overclocked.  We see a marked difference for the first time between the frame rate graphs of the two G4Ti4600-VT2D8X results, as the added memory bandwidth allows the card to have a higher spike than the standard clocked card.  The Parhelia has the same spikes as before, and without them would have a maximum of 53fps which moves it closer to the results of the MX in this case, though the Parhelia's graph has more movement.

The results here show that even a high performance card, such as the G4Ti4600-VT2D8X, can still be brought to it's knees by an older game.  The MX is simply unable to give a decent performance, as the graph shows it is very flat, with no 'spikes' to speak of.  The G4Ti4600-VT2D8X does the best of all the cards tested here, but it also has a mostly flat graph, with a difference of about 19fps from lowest to highest or about 9fps either side of the average frame rate.  The Parhelia gets closer to the G4Ti4600-VT2D8X, thanks to its bandwidth and FAA, but still can't quite catch them, in fact it still has the two upward spikes that it had previously, though less pronounced in this case.  Do the standings we see set from the OpenGL tests remain when we look at UT2003 and Max Payne.