by Alexey Stepin , Yaroslav Lyssenko
06/17/2009 | 04:53 PM
Our express-test sessions upon release of new ATI and Nvidia graphics card drivers have already become a good tradition. We only make exceptions when the developer claims there are no performance improvements or any significant functionality changes. The recently released Catalyst 9.6 is not one of those exceptions for sure, because Advanced Micro Devices promises a significant performance boost in several games, such as:
We believe this is reason enough to perform another test session and check out if the above mentioned promises really take place in the new Catalyst 9.6 driver version. As usual, besides performance benefits, the new ATI driver version fixes some of the existing issues. Namely, they resolved the following:
They also claim that the following fixes are in place for the Windows Vista operating system:
Besides, they have also resolved a number of issues with Catalyst functioning in Windows 7 OS. You can find the complete list of improvements and fixes on the official Advanced Micro Devices site in the section devoted to ATI Radeon HD graphics card drivers. As for us, we are primarily interested in checking out if the claimed gaming performance improvement will prove true.
We are going to check out the performance of the new ATI Catalyst driver version 9.6 using the following testbed:
Both versions of ATI Catalyst driver were configured as follows:
This time we used the following games for our performance tests:
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 significant representatives of ATI Radeon HD 4800 family:
We also tested Radeon HD 4890 in CrossFireX configuration. This way we could estimate the improvements brought by the new Catalyst 9.5 driver not only in the common single-GPU mode but also in multi-GPU mode.
We ran our tests in the following resolutions: 1280x1024, 1680x1050, 1920x1200 and 2560x1600. We didn’t test Radeon HD 4770 in the latter one. 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 ATI Catalyst 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 appropriate tools available in the game itself, we used Fraps utility version 2.9.8 to record our performance measurements in this game. In the latter case we ran the test three times and took the average of the three readings for the performance charts.
This time besides the gaming tests, we decided to also check out the performance and efficiency during video encoding by the GPU. To accomplish that we used the official ATI Avivo converter and the new Cyberlink MediaShow Espresso converter that supports both: Nvidia CUDA and ATI Stream.
One of the most common video encoding tasks is transcoding from high-definition formats in formats supported by portable playback devices. We decided to transcode the initial file into the format supported by popular iPod players in 640x360 resolution and with 2.5Mbit bitrate.
Unfortunately, we faced a number of problems. Namely, our attempt to work with a video in H.264 1080p format with about 10Mbit bitrate (MP4 container) failed, because Avivo converter refused to work with it at all, while MediaShow Espresso didn’t demonstrate any improvement in transcoding time with ATI Stream mode enabled.
We undertook a second attempt. This time we took a WMV HD 1080p video (Windows Media container) with about 7.5Mbit bitrate. This time transcoding completed successfully with the following results:
MediaShow Espresso needed almost equal amount of time for transcoding by CPU and GPU, while the brand name ATI Avivo converter performed a little better. Nevertheless, it is still early to claim the victory of GPGPU, because a number of sources report image quality issues in videos transcoded with ATI Avivo. Things will definitely get better as time goes on and we will certainly get back to this interesting topic, but at this point ATI solutions are still not the best choice for those who work with video on a PC platform a lot.
Since they promised no performance improvements in this game with the new Catalyst version, it is only logical that there is no noticeable performance improvement. Slight performance fluctuations can be written off as a measuring error.
As we have already pointed out before, disabling “vsync” option in Company of Heroes results in game instability that is why we decided not to do it. So, maximum performance in this game depends on the screen refresh rate, which is 60fps in our case.
Despite the promised 10% performance improvement for Radeon HD 4800 graphics card family, we didn’t see anything like that even in 2560x1600 resolution. Nevertheless, the improvement from Catalyst 9.6 was present with less powerful graphics accelerators: the minimal performance of Radeon HD 4770 did increase in lower resolutions.
We didn’t run the tests for Radeon HD 4600 graphics card family that is why we can’t confirm or deny ATI’s claims of increased average performance of these graphics cards by as much as 25%. The performance level may still be not high enough even with the promised improvement, because Radeon HD 4670, not to mention Radeon HD 4650, is not such a good gaming solution after all, et least when we are talking about contemporary gaming titles.
The developers promised that multi-GPU systems will speed up by 11-13% in games based on CryEngine 2. However, we didn’t see anything like that. We didn’t reveal any slowing of the Radeon HD 4890 CrossFireX tandem either, and in resolutions starting at 1680x1050 and up its minimal performance in fact increased a little. So, we can conclude that Crysis and Crysis Warhead fans will not really benefit from upgrading to Catalyst 9.6 driver version.
We have to particularly dwell on a significant performance increase by Radeon HD 4770 in 1920x1200, which hit almost 30%. However, in absolute numbers it only means that the average performance rose from 7 to 10 fps, which is actually of no real practical value for gamers anyway.
We have always included this game into the list of our regular benchmarks for the new ATI and Nvidia drivers, because its engine is based on OpenGL API, which is a pretty rare thing these days.
ATI didn’t promise that anything would improve in Quake Wars and our tests showed that it was indeed the case: the results obtained with Catalyst version 9.5 and 9.6 are absolutely identical.
In H.A.W.X., however, the new Catalyst 9.6 driver version causes multi-GPU solutions to slow down a bit. The performance drop is not too high and makes only 3-5%, which doesn’t affect the level of gaming comfort in this game, but the fact is undeniable: the new driver does worsen the performance. Luckily, Catalyst 9.6 didn’t affect the performance of single-GPU solutions in any negative way.
The developers promised the biggest performance boost in this game – up to 30% when the system CPU becomes the bottleneck. And this time, ATI kept their promise: in 1280x1024 the average performance of Radeon HD 4890 CrossFireX tandem increased by almost 14%, and that of a single Radeon HD 4890 – by 7%. The gain wasn’t high enough to hit the promised 30%, but was nevertheless quite significant. Unfortunately, we can’t advise Radeon HD 4770 owners willing to play this game to upgrade to Catalyst 9.6, as their performance may drop by 4-10% depending on the resolution.
Trying to minimize the CPU influence in 3DMark Vantage we used the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering).
Our tests with the new Catalyst 9.6 driver in 3DMark Vantage didn’t reveal any noticeable differences from Catalyst 9.5. Therefore, for those who take the results of this benchmarking suite seriously there is no need to rush and upgrade to the new driver version: you won’t get any obvious improvement anyway.
The results of our ATI Catalyst 9.6 express-test turned out quite ambiguous, because we didn’t see the promised performance improvement everywhere where we expected to, and in some cases, performance even dropped. Just take a look at the summary diagrams and you will be able to tell if it makes sense to upgrade to the new Catalyst 9.6 version or wait until version 9.7 comes out in July:
The biggest performance gain occurs in World in Conflict: Soviet Assault in 1280x1024 resolution, where it makes around 14% for Radeon HD 4890 CrossFireX and around 7% for a single Radeon HD 4890. However, this is where we also see a significant performance drop by Radeon HD 4770 that sometimes drops by as much as almost 10%. So, the owners of these inexpensive graphics accelerators who play World in Conflict games should better refrain from upgrading to the new Catalyst 9.6 driver. Also, the owners of multi-GPU configurations on ATI Radeon chips may experience some performance worsening in Tom Clancy’s H.A.W.X. flight simulator.
In all other cases Catalyst 9.6 would be the way to go: even if it doesn’t improve the performance for you, it will definitely fix a few issues described in the release notes. As a rule, it makes sense to install the latest driver version, except for a few cases we specifically mentioned above. These cases are pretty rare, but we can’t disregard them completely, which is why of the reasons for us to continue posting the results of our ATI and Nvidia new driver express tests and provide the information updates for our readers. Please note that the results obtained in our lab shouldn’t be taken as the one and only universal truth, because the performance improvement from the use of Catalyst 9.6 driver may be different on a differently configured system.
As for the use of GPU computational capacity for video encoding tasks, AMD has a long way to go in improving performance and image quality aspects before Stream technology becomes truly popular among home video enthusiasts.