Nvidia's GeForce GTX 480: Before And After A Year Of Driver Updates

Final Words

Unlike AMD's Catalyst drivers, we haven't seen a major overhaul of Nvidia's user interface within the past year. Performance, on the other hand, has definitely changed. 

Over the course of almost 14 months, AMD's Catalyst drivers delivered 5% to 10% better performance. For Nvidia, the results are similar but the scale is larger. There are situations where we see little to no change in performance, but those scenarios are few in number. On average, we see a roughly 10% performance increase in as little as nine months.


Some of our results are amazing. In Battlefield: Bad Company 2 and F1 2010, the minimum improvement is 30 percent. In one rare case, the GeForce GTX 480 experienced a 75% performance jump. The only oddball is World Of Warcraft: Cataclysm, where older drivers seem to deliver better performance. Practically speaking, the performance loss may not matter much. If you are already gaming at 90+ FPS in a MMORPG, losing nine FPS is probably inconsequential. But if you are spending hard-earned dough on a graphics card, it's in bad taste to lose performance in subsequent driver updates.

Nvidia proudly advertises how much performance you can gain with each driver release. This isn't without some fine print (literally). The company specifically cites the settings it uses to test, and smartly employs "up to" verbiage to indicate that your mileage may vary. As an example, for the GeForce GTX 480, only a few games in our suite are specifically expected to show improvement.

Game (Setting)
Maximum Performance Gain from 197.41 to 266.58 (Per Nvidia)
Aliens vs. Predator (1920x1200, No AA/AF, Tessellation on)24.26%
Just Cause 2 (2560x1600, No AA/AF, Concrete Jungle)25%
Metro 2033 (1920x1200, No AA/16xAF, Tessellation on)7%
Aliens vs. Predator (1920x1200, 4xAA/16xAF, Tessellation on)11%
Metro 2033 (1920x1200, 4xAA/16xAF, Tessellation on)6%
Metro 2033 (1680x1050, No AA/16xAF, Tessellation on)7%
Battlefield: Bad Company 2 (1920x1200, 4xAA/16xAF)7%
Just Cause 2 (1920x1200, 4xAA/16xAF Dark Tower)5%

Based upon these games and their settings, we can only say that Nvidia's claims of up to 25% improvement in Just Cause 2 at 2560x1600 seem a bit high, considering we only saw a 9% gain. And we assume the company made a typo, because the game's lowest AF setting is 2x.

Beyond performance gains, the conclusions from our earlier AMD driver article are still valid. We usually assume the largest performance jump occurs when we max out all settings, but this isn't true every time. The reverse actually occurred with Battlefield: Bad Company 2 and F1 2010. There are scenarios where new drivers bring better performance at lower resolutions and lower quality settings.

When it comes to the GeForce GTX 480 specifically, drivers seem to matter much more. This makes reasonable sense, given the newness of its Fermi architecture and the expected evolution that happens over time as driver developers and ISVs get more experience from the new hardware. You can look at this one of two ways. Either Nvidia is progressively optimizing its drivers or it did a poor job of delivering good performance when its card launched.

The same can be said for AMD. It is either making fewer optimizations over time, or it is delivering more of its potential performance when a product launches. However, this only applies to the last generation of cards. It's possible that the scenario will change in the future when we look at the GTX 580 and the Radeon HD 6970.

Create a new thread in the UK Article comments forum about this subject
This thread is closed for comments
Comment from the forums
    Your comment
  • wild9
    Are the driver improvements solely aimed at games or do we also see some gains with CUDA? Many Thanks.
  • acku
    Anonymous said:
    Are the driver improvements solely aimed at games or do we also see some gains with CUDA? Many Thanks.

    That's a really good question. In a few words, graphic driver improvements are aimed solely at games (stability and performance.) There is a reason for this. When a CUDA application performs a calculation, it uses programming calls from the CUDA library. This is separate from the graphics driver, and it's something that NVIDIA updates less often. (The same goes for AMD's APP). It is the CUDA library and CUDA application that ultimately determines the efficiency of GPGPU performance.

    Say you wanted to obtain the sum of all integers from 1 to N:

    int i, sum = 0;
    for (i = 1; i <= N; i++)
    sum += i;
    printf ("sum: %d\n", sum);

    this can be rewritten as

    int sum = (N * (N+1)) / 2;
    printf ("sum: %d\n", sum);

    If N were sufficiently small and hardware performed addition and looping faster than it did multiplication and bit-shifting, this wouldn't be an improvement. But this is an example of making code within a program more efficient and perhaps more performance oriented. When NVIDIA updates its CUDA library, generally we are dealing with either a complier optimization (developer side) or a run-time optimization (user side).

    Andrew Ku
  • mi1ez
    Wow. Pity about the WoW stats. Dropped the ball there!
  • asteldian
    A little careless letting slip with WoW. Not that it matters, 480 is plenty power regardless, but given how popular that game is (for whatever reason) you would have thought making sure the update at least kept things the same with tat game would have been smart
  • Anonymous
    Humm, you should activate directx 11 on WOW. It really gives a good boost on fps and might shift results with the most recent drivers.
  • mi1ez
    Turning on DX11 gives a boost to frame rates?