Articles: Graphics

Bookmark and Share

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 ]

Instantaneous Performance and Image Quality

The results we have gathered so far indicate that Metro 2033 is a demanding application, nearly as difficult on the graphics subsystem as the notorious Crysis and Crysis Warhead. At the highest graphics quality settings you can only achieve a playable frame rate with the most expensive and advanced solutions, mostly from the ATI Radeon HD 5800 series. It is clear that Nvidia’s current products, including the GeForce GTX 295, find it hard to compete with AMD’s top-end cards. Even if their average speeds are comparable, Nvidia’s products usually have a lower bottom speed.

Now, we will try the game at different graphics quality settings and also see how much the CPU contributes to the performance of the system in when running Metro 2033. We will also check out if Nvidia’s cards can be used to play the game somehow. We took two popular cards for this test: a Radeon HD 5850 and a GeForce GTX 285. Both cards were tested together with Core i7-975 Extreme Edition and Core i7-920 processors. We ran this test at a resolution of 1680x1050 to avoid being limited by the graphic card’s performance.

For each of the five quality profiles we were recording the instantaneous performance with the Fraps 3.1.2 utility for 60 seconds. As usual, we tried to make the test scenario as variegated as possible but that was not easy due to the tunnel-like design of most of the game’s levels. Anyway, we captured a few screenshots that can help us evaluate the difference in image quality between the game’s settings profiles.

Here are the results:


First of all, we can see that FSAA does not work well in Metro 2033. It blurs the textures. Moreover, 4x FSAA leads to a 50% performance hit, which is perfectly visible in the instantaneous performance graph. The Radeon HD 5870 may even be as slow as 25fps – at a resolution of only 1680x1050 pixels! The GeForce GTX 285 is quite hopeless. Most of time its performance is below 25 fps irrespective of the CPU model, which means that the graphics card is the bottleneck. With the Radeon HD 5850, replacing the Core i7-920 with the Core i7-975 EE does improve the performance as the average frame rate grows up from 34 to 39 fps and the top speed from 48 to 56 fps. The bottom speed remains at 25-26 fps with both CPUs, though.


The Very High quality mode without FSAA delivers the best visuals although the edges of polygons are jagged as you can clearly see in the screenshot showing the primus stove with wire handle. We guess the detailed textures make up for this shortcoming, which is not conspicuous at high resolutions, especially considering the semidarkness reigning in most scenes of the game. The frame rate is overall higher here. The GeForce GTX 285 is but occasionally slower than 25 fps. The Radeon HD 5850 has no problems at all as its speed is never lower than 35 fps whatever the scene. The CPU influences the resulting performance much less than in the 4x FSAA mode. In fact, there is no influence: the results with either card are the same irrespective of whether they work together with the expensive and hot Core i7-975 EE or the more affordable Core i7-920. So, we recommend the Very High mode for all users of top-end graphics cards.


In the High mode there is some loss of detail in textures and some of the touted special effects seem to be turned off. For example, the soldier in the first screenshot looks less “glossy”. The lighting model is simplified noticeably, which can be easily seen in the bonfire screenshot, but not too much. What about speed then? Well, the diagrams suggest that there are no performance benefits in this mode. The Radeon HD 5850 was fast enough even in the Very High mode whereas the GeForce GTX 285 only speeds up from 22 to 24 fps in terms of bottom speed and from 54 to 64 fps in terms of top speed. The CPU affects the result more than in the Very High mode, but the Core i7-920 is for some reason faster than its senior cousin when working with the Radeon HD 5850. This CPU’s instantaneous performance graph looks smoother than the other CPU’s with the GeForce GTX 285, too. It must be due to the inaccuracies of the manual method of testing or, more likely, to some background services always running in the Windows 7 environment. This is the price you have to pay for enjoying multi-tasking. You can only avoid it by transitioning from the PC to game consoles.


When we choose the Normal mode, the textures lose more details and the lighting model is simplified further. This is quite clear if you compare the static screenshots scrupulously, but the difference is far from obvious in the hot action going on in the semidarkness of subway tunnels. This lowering of graphics quality settings saves the day for Nvidia’s solutions: the GeForce GTX 285 is in fact not slower than 35-40 fps in the Normal mode. There’s no point in using this mode for the Radeon HD 5850, of course. As for the CPU’s influence, the fluctuations you can see in the diagrams are largely accidental and due to the test method inaccuracies.


And finally, the Low mode is the worst-quality one in this game, yet we cannot say it is far worse than the Normal one, especially during actual gameplay. The frame rate of the game does not grow up much in the Low mode, so we don’t think it is practically useful, even for graphics cards slower than the GeForce GTX 285. The CPU’s influence is minimal and does not justify the replacement of a Core i7-920 with a Core i7-975 EE. Purchasing a Radeon HD 5850 or a Radeon HD 5870 is going to have a much stronger effect, but you wouldn’t want anything other than the Very High mode then.

Summing it up, we can note two features of Metro 2033 that you should know about to achieve highest performance and maximum image quality. First, you should not use traditional FSAA. Besides provoking a 50% performance hit, it blurs all textures. FSAA is useless in this game. And second, if your graphics card cannot deliver a comfortable speed, you should switch to the Normal mode to increase the frame rate. There is no point in choosing a lower quality mode since there is no significant increase in frame rate, especially with Nvidia’s cards. On the other hand, the game’s visuals do not degenerate much, either.

Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 ]


Comments currently: 19
Discussion started: 03/23/10 10:41:12 PM
Latest comment: 12/21/15 11:12:11 AM

View comments

Add your Comment