New drivers appear regularly from AMD’s kitchen, and this is something that we’ve always appreciated about Radeon graphics cards. It’s one of the best ways to let your buyers know that they are being covered in terms of technical support and that they won’t be left on their own after they’ve purchased your product. Simply put, if something isn’t working properly, or you’re encountering trouble with a specific application, chances are that the next month’s driver revision will amend it all. What’s even more useful is opening the release notes of the new driver revision and reading the list of improvements, particularly performance jumps in certain games, which ranges from a few percent to a few dozen percent.
This list used to be rather short back in the day, but has increased ever since, and with both manufacturers at that. It seems that the same games keep appearing all over again from release to release, and since the percentages seem to be increasing as well, we were interested in whether the actual improvements are as big as they claim them to be. Simply enough, we’ve been unable to evade the impression that manufacturers have found another path for marketing their cards with sometimes ridiculous performance improvements announced on various websites, without a hint at the conditions under which the new results have been recorded and in what resolutions/at what detail settings. We’ve therefore decided to have a first-hand test of just how much improvement different driver revisions show for one of the most popular AMD graphics cards - Radeon HD 6870.
We’ve chosen this particular model because it’s based on the VLIW5 architecture and because AMD seems to quote it the most in the release notes of each subsequent driver revision. Besides, it’s not a new card, as it was released simultaneously with the appearance of driver version 10.10. On the other hand, AMD quotes their 11.4 as the most drastic performance update. In order to get the best overlook of the progression through time, we’ve also picked an intermediate revision, namely 11.2. This made the task more arduous, but also more thorough. As for the actual games used for testing, we’ve picked the ones quoted most often in the notes, again. Other than that, we’ve also chosen a few randomly picked tests from our usual battery, none of which have been mentioned in the notes, just to check how well the improvements scale across all applications, outside of what AMD has targeted specifically. Nothing has been changed during testing other than the drivers used, and the specific card we’ve been using is Radeon HD 6870 1 GB GDDR5 in AMD’s own referent version. This means that the clocks are referent as well, of course, with the GPU clocked at 900 MHz and the memory at an effective 4.2 GHz. In order to get results as usable as possible, we’ve tested in the two most commonly used resolutions today: full HD (1920x1080) and the most popular 16:10 resolution of 1680x1050. These two resolutions are what we believe to be a universal benchmark, as most users nowadays are running monitors with one of these being native. Furthermore, the detail levels have always been kept maxed out. Some tests have been done with anisotropic filtering and antialiasing, and some without. Therefore, we believe we have covered the maximum number of possible configurations and all common usage scenarios. Let’s see what the results have to say.
