Comprehensive Performance And New Questions
Let us take a closer look at the performance improvements. We used percentage ratings over pure numerical values for presenting these results. The numerical results may vary on different systems, but the percent differences between the versions are relative and much more consistent. The Radeon HD 5870 with Catalyst driver version 9.12, as used in the previous articles, serves as the reference point in all measurements.
Text
As expected, there is not much of a difference between the cards and driver versions when rendering text. And while text rendering was never one of the problem areas in our first articles, we still found a small performance increase in the new drivers. It might be measureable, but most likely not noticeable for the average user, as the performance levels are so high already.
Moving on, we will now turn our focus to the test areas where we could really measure significant differences in performance.
Lines
The rendering of lines was, in fact, one of the previously-reported problem areas. With the new drivers, the performance increase we measured is significant. These new results should be enough to ensure smooth 2D performance at all times. The GeForce GTX 285 is a bit faster than the Radeon HD 5870, but not so much that you can tell outside of a benchmark.
Interestingly, the GeForce GTX 480 results reveal that the Nvidia drivers have not been optimized a lot, and that there is still lots of room for improvement.
Polygons
The GeForce GTX 480 is a positive exception, while the differences between the other results are rather marginal. The lack of polygon performance is an issue, which is clearly visible to the user in some cases. This weakness is not of much relevance outside of a few specific applications. But nevertheless, all of the values are too low considering the real potential of these cards (especially if you know that a Radeon HD 5870, in some cases, gains a 250% performance boost simply because the task manager is open). We will investigate this ‘paradox’ in more detail on the next page.
Ellipses
The ellipse problem can be checked as a solved issue, as the ATI driver developers now are up to speed.
Blitting
Blitting results in the same kind of performance as stretching or drawing rectangles. The performance differences are small and all results indicate good 2D performance.
Interim Conclusion
ATI has done its homework properly. There are (except perhaps for the polygons) no longer any areas of its 2D performance that will actually restrict your productivity, even if there is still room for further optimization.
This is, in turn, related to certain security features of Windows 7 and affects both graphics companies to the same degree. Check out the next page for more information about this as well as its relation to the ‘paradox’ we mentioned.
