Final Words
Once you get past the simple matters of rectifying system crashes and image quality errors, AMD's driver updates give us two tangible benefits:
- New GUI interface.
- Performance changes.
| Driver | Catalyst 9.10 | Catalyst 10.3 | Catalyst 10.8 | Catalyst 11.1 |
|---|---|---|---|---|
| Date | 10/22/09 | 3/24/10 | 8/25/10 | 1/26/11 |
We see mixed results when it comes to performance. If you are looking at this from one driver update to the next, we see situations where one driver update, might give you lower performance than an even older driver if you use it in a game that just came out. On the other hand, if you look at drivers in four-month intervals, we generally see a one-half to three percent improvement. Over the course of 15 drivers, there is a 5% to 10% bump in performance.
When AMD or Nvidia release a new driver, we have been trained to think that we’re always getting beneficial performance tweaks. In reality, only some drivers actually affect performance. Other times, a driver release might address gamma control or HDMI audio output. Even when the justification for a new driver is decidedly performance-oriented, the marketing lingo "up to" ends up being an easy out for vendors when end-users don’t see large frame rate increases.
| Game | Total Improvement Claimed for Single-Card Radeon HD 5800 SeriesFrom Catalyst 9.10 to 11.1 |
|---|---|
| F1 2010 | 10.0% |
| Left 4 Dead 2 | 6-8% |
| Stalker: Call of Pripyat | 37.5% |
| Battleforge | 3.0% |
| The Chronicles of Riddick: Assult on Dark Athena | 5.0% |
| Far Cry 2 | 2-6% |
| Stormrise | 5-10% |
| DiRT 2 - DX9 | 10.0% |
| 3DMark Vantage | 13.4% |
| Call of Duty: World at War | 2.0% |
| DiRT 2 | 29.6% |
| Enemy Territory: Quake Wars | 5.0% |
| Stalker: Clear Sky | 2.0% |
| Tom Clancy's H.A.W.X. | 20.0% |
| World in Conflict | 5.0% |
| Crysis | 4.0% |
In the 15 drivers spanning close to 14 months, AMD claims to have improved performance in 16 games with the Radeon HD 5800-series operating in a single-card configuration. The biggest gains, understandably, come from older titles that were already out when the 5870 launched and have enjoyed constant updates from day one.
Our tests show basically no improvement in the relatively new F1 2010 (with the exception of 2560x1600 and 8x AA; in that one case, we see about a 40% improvement, which exceeds the claimed 10 percent). However, remember that F1 2010 came out nearly a year after the 5870. So, while AMD can be applauded for providing drivers supporting the game before it was sold, we see little gain from drivers rolled out once the game was already available.
We normally assume the largest performance bump occurs when we max out all settings, but this isn't true in every situation. In fact, we actually saw the reverse occur in Battlefield: Bad Company 2 and Aliens vs. Predators. Clearly, there are games where new drivers bring better performance at lower resolutions without anti-aliasing or anisotropic filtering enabled. With DX11 introducing better support for threading, it's entirely possible that, at some point, both games benefited from an optimization that alleviated a processor-oriented bottleneck.
On one hand, driver updates are probably going to give you less real-world performance than the best-case release notes would otherwise have you believe. On the other hand, driver updates are free. Trading up from a Radeon HD 5870 to a 6950 gives you a roughly 6% bump in Metro 2033. But getting that small increase costs you an extra $250. The moral of the story: be sure your graphics card drivers are always up to date. You might not see the gains each revision’s release notes suggest, but you will get the latest visual/stability optimizations and whatever small speed-ups might apply to your configuration.
