NVIDIA Puts Its (New) Cards on the Table

Benchmark Impressions

Benchmark tables are a great invention. They let you compare the performance of the cards in this test easily and quickly. However, interpreting these results is a great deal harder, though, since it always depends on the frame of reference that is being used. If a card is 10fps faster than its competitor, then it is obviously the winner in that specific test. However, everything is relative. If a group of cards all churn out about 150fps, for example, then a 10fps difference is hardly noticeable. If the cards can only sustain 40fps, then obviously 10fps can decide between playable and unplayable.

What really counts is the performance delta as expressed in percentage values - but adding those numbers to our benchmark tables would only add to the numbers confusion.

Of course, everyone applies their own standards when looking at benchmark results. Since that is the case, sober percentages won't help judging a card either. Then again, if we were to draw an arbitrary line at, say, 5% to decide between "as fast as" and "slower than," that would bring up the problem of borderline cases.

A card that is only 5.5 % slower in a benchmark, for example, will not be very noticeable compared to another card in game play.

And if we came up with entire brackets within which we tried to group the cards, one could still argue about the borders between these brackets....

In the following tables we have attempted to summarize the scores of the cards participating in this review. In doing so, we didn't simply add the number of victories and losses; we also took the percentage differences into account. The gauge by which we measure these results is naturally subjective and has no pretension to being the sole true measure. Image quality does not factor into the score here also. And everyone can make his or her own choice according to the benchmarks.

The GeForce FX 5950 doesn't look to good in this comparison. It does especially bad in 4xFSAA and 8x anisotropic filtering tests.

In most of the standard tests, the FX 5700 is on par with the Radeon 9600 XT. It too tends to take a performance nosedive when 4xFSAA and 8x anisotropic filtering come into play.