The results here seem to speak more of driver efficiency, as the MX and the Ti4600 both perform the same, with the Parhelia trailing behind due to its drivers, which are not as mature performance wise as the GeForce 4 series drivers.  Looking a the line graphs we see each card has a slightly different time where it peaks frame rate wise.  The performance difference between the overclocked and non overclocked GF4 is simply a matter of the error range associated with this benchmark.  Now do the results change with AA and ansio added?

What can we see from these results?  The GeForce 4 MX takes a 60% nose dive just by enabling AA and ansiotropic filtering, and while the frame rate graph is fairly flat (with one major down 'spike') it still may be playable.  The G4Ti4600-VT2D8X loses about 9% of its frame rate without AA and ansio enabled, and has a much smoother graph without an excessive high frame rate spike.  The low frame rate is also met a couple of times more than it meets the high frame rate which gives a slight indication that the card is losing its dependency on the processor to limit it.  The Parhelia has a very flat graph apart from its ~20fps difference at the beginning of the test.  What about 1600*1200, do these cards hit a fill rate limited wall?

The results are somewhat similar to those taken at 1024 without AA.  The major difference is that the MX is showing signs of fill rate limitation losing about 38% of its initial frame rate.  The spikes are slightly more pronounced (at least the downward ones), with the lowest frame rate reaching below 60fps.  The G4Ti4600-VT2D8X does very well, showing no real sign of fill rate limitation here with a graph very similar to that of its results at 1024, with a slightly more pronounced low end frame rate spike.  The Parhelia has a flatter graph then either of the other cards, as it only goes below 70fps three times (out of 25), so it is within a 8fps range for 88% of the test.  Does the same thing that happened when we enabled AA at 1024 happen at 1600?

Interesting, with AA and ansio enabled the overclocking of the 4800 shows a slight improvement, which is right along with its very slight overclocking potential.  The MX loses 63% of its frame rate from enabling these features, just as it did at 1024*768.  Its graph is very flat, with a 6fps swing on either end of the average frame rate.  The G4Ti4600-VT2D8X shows a good bit of fill rate limitation reaching it, as its frame rate drops 62%, just like its slower sibling the MX.  While the graph is not as flat as the MX it is still flatter than any of the previous results.  The Parhelia does better 'only' losing 45% of its frame rate in comparison to that of the other cards, and has more spikes upward than previously.  This is only one OpenGL game, that is showing its age, how does the GeForce 4 fair when we use the much more graphically demanding Serious Sam: SE?