by Alexey Stepin , Yaroslav Lyssenko
02/18/2009 | 09:36 PM
As we have already told our readers, it has become a tendency in contemporary consumer 3D graphics industry that the driver may often be of greater importance than the hardware itself, especially now that multi-processor graphics systems are getting more popular. The developers are constantly improving the software side of their solutions trying to achieve the most optimal performance in as many contemporary gaming titles as possible that is why new graphics cards drivers are being released on a regular basis. Sometimes they may also roll out an urgent fix for some critical performance or image quality issue.
In our previous article we discussed the new ATI Catalyst 9.1 driver version and found out that it really provided a noticeable performance improvement in certain games compared with the previous ATI Catalyst 8.12. This time it is Nvidia’s turn. Although the company often releases interim beta-versions without the WHQL status, we do not use them in our reviews on purpose, as we prefer to work with certified drivers only. Just a few days ago the company released another version of their certified GeForce 182.06 driver, and they claimed this version would have pretty significant improvements and innovations. Here is a brief list of the most significant changes:
Moreover, version 182.06 acquired supports for hardware PhysX acceleration for all Nvidia cards starting with GeForce 8 series that have at least 256MB of local video memory onboard. They also claimed to have fixed certain problems connected with switching power-saving modes on GeForce 8/9 solutions.
Just like with ATI Catalyst 9.1, we decided to check out if Nvidia’s claims are actually true and how beneficial driver upgrade with the new version 182.06 may turn out for a mainstream gamer. To achieve these goals, we ran a few tests showing the contemporary graphics accelerators performance with the new driver in a number of popular games, including the ones mentioned above.
We are going to investigate the performance of new Nvidia GeForce 182.06 driver using the following testbed:
Both versions of Nvidia GeForce driver were configured to provide the highest possible quality of texture filtering with minimal effect from default software optimizations. We enabled transparent texture filtering. As a result, our driver settings looked as follows:
This time we used the following gaming titles for our performance tests:
Although Nvidia promised a significant performance improvement in Half-Life 2 shooter, we didn’t include it into the list of games for this test session, because Left 4 Dead game uses the same Source engine.
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. We didn’t use any non-standard settings, except a few situations, described separately later in the article.
For our tests we chose two most illustrative representatives of Nvidia GeForce GTX 200:
We ran our tests in the following resolutions: 1280x1024, 1680x1050, 1920x1200 and 2560x1600. We added MSAA 4x antialiasing to the standard anisotropic filtering 16x in all tests. We enabled antialiasing from the game’s menu. If this was not possible, we forced it using the appropriate Nvidia GeForce driver settings.
Performance was measured with the games’ own tools using the original demos. We measured not only the average speed, but also the minimum speed of the cards if possible. If there were no built-in tools for taking performance readings, we used Fraps utility version 2.9.8 to record our performance measurements in these games. In the latter case we ran the test three times and took the average of the three readings for further performance analysis.
The part responsible for OpenGL support in the new GeForce drive version seems to have been significantly improved, as you can see from the results in the only game in our tests using this API. Depending on the resolution, the performance improvement is different. For example, the biggest gain in 1280x1024 of over 20% belongs to GeForce GTX 295, while GeForce GTX 285 improved only by 9%.
In 1680x1050 resolution the gains made about 10% and 8% respectively, in 1920x1200 – about 1.5% and 3%, and in 2560x1600 – the biggest performance boost was from GeForce GTX 285 – 7% vs. 2.5% by its elder dual-chip brother.
We can state at this point that the performance does improve and in some cases quite significantly. Now we have to find out how things are in Direct3D games.
They promised up to 8% performance improvement in this game in high resolutions with enabled FSAA. It was really so in the maximum 2560x1600 resolution: GeForce GTX 295 sped up by over 6%, while GeForce GTX 285 – by almost 9%. Both solutions provide a pretty comfortable performance level. In 1920x1200 the improvement is not so dramatic and makes only 2.5% and 4.5% respectively. In the lower resolutions the gain gets almost zero, just as the manufacturer claimed it would. Only in 1680x1050 the new driver delivers about 2.5% average speed gain to GeForce GTX 285.
The second round is up, and we seem to be getting everything we’ve been promised so far. Good job!
New driver version didn’t really promise any improvement in this game, but the reality turned out quite different: in some cases we detected about 2-3% performance difference between the old and the new GeForce driver versions.
Moreover, the minimal performance of GeForce GTX 295 more than doubled in 2560x1600. So, now the owners of this graphics accelerator finally got the opportunity to enjoy quite good gaming experience in this mode.
According to the GeForce 182.06 driver release notes, we should see up to 10% performance boost in F.E.A.R. 2. However, we observed an even greater improvement: GeForce GTX 285 sped up by 18% in 2560x1600. In other cases this number varied between 3% and 8%, which is also quite good. Note that the owners of monitors supporting resolutions over 1920x1200 will benefit most of all from the new Nvidia driver version in this game, while in 1280x1024 the performance with the new driver got even a little slower. It is quite insignificant and could be attributed to the measuring error of our testing methodology that allows using Fraps in manual mode.
GeForce 182.06 driver delivered even more than had promised in this game: instead of the 11% improvement we recorded a 20% to 40% performance gain!
Looks like they have optimized the SLI technology support; however, even a single-GPU GeForce GTX 285 benefitted from the new driver, especially in 2560x1600. It sped up by 27%! In lower resolutions it managed to catch up with and outperform a dual-GPU GeForce GTX 295 running on 181.22 driver version! As we can see, Nvidia doesn’t just stick to their promises: they sometimes deliver way more.
The last game in our today’s test session was supposed to show up to 10% performance improvement in high resolutions with enabled full-screen antialiasing. And that is the only case when Nvidia didn’t fully comply with their promises: GeForce GTX 295 gained 3% at best, while GeForce GTX 285 did in fact hit 10%-13% higher numbers.
Moreover, the minimal performance of the flagship solution would occasionally drop with the new driver.
As usual, we try to minimize the CPU influence in 3DMark Vantage by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering).
Since they didn’t declare any performance changes in 3DMark provided in the new driver, we were not surprised to see that driver versions 181.22 and 182.06 delivered similar results. When we checked out the individual benchmarks, the results coincided with up to two tenths of second precision. However, a pragmatic gamer will hardy take 3DMark results seriously into consideration, because they are rarely considered representative of a graphics accelerator/driver performance in contemporary games.
Our mini-investigation confirmed Nvidia’s claims of gaming performance improvement in certain titles provided by the latest GeForce driver version 182.06. For your convenience we grouped the results in tables, one for each of the graphics cards participating in our today’s tests.
The biggest performance boost had been promised for high resolutions, namely 1920x1200 and 2560x1600, while in 1680x1050 and 1280x1024 we could even see the new driver falling slightly behind. Nevertheless, we did find a few exceptions.
The modifications introduced by Nvidia software developers in their new GeForce driver version 182.06 showed their real best in Left 4 Dead game based on Valve Source engine: the maximum gain for GeForce GTX 295 made almost 40%, and for GeForce GTX 285 – 25%.
The second place was taken by Enemy Territory: Quake Wars online shooter: updated OpenGL driver improved the average gaming performance of GeForce GTX 295 solution by 21% in 1280x1024 resolution and by 10% in 1680x1050 resolution. GeForce GTX 285 sped up by about 8-9%, and about 7% in 2560x1600 resolution.
In all other cases, the new GeForce driver delivered on all Nvidia’s promises. Only in F.E.A.R. 2: Project Origin in 2560x1600 resolution GeForce GTX 285 got over 10% faster. I would like to specifically draw your attention to a significant increase in minimal fps level in Far Cry 2 game on GeForce GTX 295 in 2560x1600 resolution, so that the owners of these particular graphics cards can finally enjoy playing in this mode.
One way or another, it makes perfect sense to upgrade to the new GeForce 182.06 driver, especially for those who play any of the games we checked out today. However, you may get additional speed also in games not mentioned by Nvidia. Looks like the company focuses not on optimizing the drivers for selected games, but on improving the performance of their solutions in general, which we can clearly see from the results obtained with GeForce 182.06 driver in Enemy Territory: Quake Wars and Far Cry 2.
Unlike ATI Catalyst 9.1 that makes things better primarily for the dual-chip Radeon HD X2 card owners, the new Nvidia GeForce driver version will work well not only for those who have the latest generation GeForce GTX 295 and discrete SLI systems, but also for the users of common single-chip cards, such as GTX 285/280/260, and maybe even GeForce 9800/9600.