NFS: Carbon - Lincoln Boulevard
Sprint, 1024 x 768
2.8
GHz X1950 Pro
2.4
GHz X1950 Pro
1.8 GHz
X1950 Pro
2.8 GHz
7900GS
2.4 GHz
7900GS
1.8
GHz 7900GS
NFS:
Carbon - Lincoln Boulevard
Sprint, 1280 x 1024
2.8
GHz X1950 Pro
2.4
GHz X1950 Pro
1.8 GHz
X1950 Pro
2.8 GHz
7900GS
2.4 GHz
7900GS
1.8
GHz 7900GS
NFS:
Carbon - Lincoln Boulevard
Sprint. 1600 x 1200
2.8 GHz X1950 Pro
2.4
GHz X1950 Pro
1.8 GHz
X1950 Pro
2.8 GHz
7900GS
2.4 GHz
7900GS
1.8
GHz 7900GS
X1950 Pro 1.8 GHz to 2.8 GHz
| 1024 | 1280 | 1600 | 1024 AA AF | 1280 AA AF | 1600 AA AF | ||
| Minimum | 37.93% | 16.67% | -5.56% | 12.90% | 26.32% | 6.25% | |
| Average | 5.01% | 3.32% | 1.55% | 3.75% | 5.33% | 2.16% | |
| Maximum | 11.94% | 2.13% | 0.00% | -11.11% | 5.13% | 0.00% |
7900GS 1.8 GHz to 2.8 GHz
| 1024 | 1280 | 1600 | 1024 AA AF | 1280 AA AF | 1600 AA AF | ||
| Minimum | 5.00% | 0.00% | 12.50% | 21.43% | 0.00% | 0.00% | |
| Average | 0.81% | -0.49% | -9.78% | 4.35% | 1.08% | 0.61% | |
| Maximum | 5.88% | -4.17% | -33.33% | 14.81% | -4.76% | 0.00% |
Obviously, this game runs very well on ATI / AMD Radeon cards, which is another way saying it really sucks on NVIDIA GeForce cards. Well, except the GeForce 8 series. Despite the lack of replay, the results from repeated runs are fairly consistent and constant. The game is also very graphically intensive, making it a good graphics benchmark. We saw no difference in average frame rates between using a 1.8 GHz and 2.8 GHz processor (in this particular race) with both graphics card. However, at default settings we saw quite an increase in minimum fps - from 29 to 40 fps - definitely significant. Since it's graphically intensive, upping the resolution have a detrimental effect on frame rate.
NFS:
Carbon - Lincoln Boulevard
Sprint, 1024 x 768, 32 bit fps Progress
48 fps
32 fps
16 fps
0 fps
NFS: Carbon - Lincoln Boulevard Sprint 1024
(click here for a more detailed graph)
| 1.8 X1950 Pro | 2.8 X1950 Pro | 1.8 7900GS | 2.8 7900GS | |
| <30 fps | 0 | 0 | 89 | 97 |
| 30-45 fps | 12 | 6 | 15 | 2 |
| 45-60 fps | 78 | 72 | 0 | 0 |
| 60-90 fps | 15 | 27 | 0 | 0 |
| 90-120 fps | 0 | 0 | 0 | 0 |
| >120 fps | 0 | 0 | 0 | 0 |
| Total | 105 | 105 | 104 | 99 |
results are in seconds
Because this is not a replay, there's no guarantee the frame rate on the graph occur on the same part of the test. So don't look for specifics, but rather try to spot the general trend. It looks like our initial assesment of higher minimum fps on a 2.8 GHz E6300 might be wrong - or is it? We certainly do not see frame rates go down to the level we saw on the previous graph. However, the table above shows there IS a difference between 1.8 and 2.8 GHz with the Radeon X1950 Pro.
Conclusion:
Scalability testing adds a new dimension to graphics card reviews. In short, they allow a glimpse of what the graphics card can do when the processors no longer become the bottleneck. This is not the same as using higher resolutions or turning graphical features such as anti aliasing and anisotropic filtering. Modern graphics card utilizes performance enhancing features to minimize performance penalty from the use of higher resolutions and graphical features. One way to look at it is that graphical test (higher resolutions, AA and AF tests) shows how much performance we lost, scalability test shows us how much performance we (can) gain. However, the two are related to one another. Games and benchmarks both run on the processor and graphics cards, the two needs to be in sync for the most optimal performance - usually measured as the highest frame rate. Use a too slow processor and the graphic card will never be able to reach its full potential (unless you're pushing high resolutions with AA and AF). Using a processors that can process data faster than the graphics is not necessarily ideal either - at some point the graphics card itself may become the bottleneck. It won't scale because the graphics card simply can not render any faster.We set out to see which card scales best: the GeForce 7900GS or the Radeon X1950 Pro. Judging from what we saw, it really depends on the game / application and the settings used. The less graphical intensive settings we use, the better we can see the influence of higher performance system when paired with these two graphics cards - scalability. For now, let's only consider results from default settings.
If we were only to look at scalability rates as increase in performance, the Radeon X1950 Pro scales better than the GeForce 7900GS in most applications. The GeForce 7900GS did scale better with Quake 4 and at least the same with SW: KOTOR. However, we also have to note that in several games, the GeForce 7900GS scored the higher nominal frame rates. The table below summarizes this fact.
| 1024 x 768 | 1280 x 960 / 1024 | 1600 x 1200 | |
| Call of Duty | higher | ||
| Homeworld 2 | higher | higher | |
| SW: KOTOR | higher | higher | higher |
| Quake 4 | higher | higher | higher |
However, there are other influencing factors in play here. Three of these games are OpenGL based games (Call of Duty, Homeworld 2 and Quake 4). It is generally accepted that NVIDIA have the better or more efficient OpenGL ICD than ATI. SW: KOTOR was a game port from the original Xbox, making use of proprietary custom extensions NVIDIA used on the Xbox graphics chip that's also available on their add-on cards (but not documented) - an extension that ATI hardware lack. So, while the Radeon X1950 Pro scales better, it's not necessarily the best graphics card.
That is, until you put AA and AF performance into play.
From the Radeon 9700 onwards, ATI hardware pushed the envelope as far as AA and AF is concerned. The Radeon X1950 Pro is no different. AA and AF performance penalty have always been smaller with ATI hardware than NVIDIA's. Or you could view it from a different prespective - AA and AF performance penalty on NVIDIA hardware are bigger than with ATI hardware. This is still true for the GeForce 7900GS. Almost all the games we tested here today runs faster with AA and AF on the Radeon X1950 Pro than the GeForce 7900GS (SW: KOTOR is the exception). In all these games, the Radeon X1950 Pro scales better than the GeForce 7900GS when AA and AF are enabled, at least until we hit the graphical processing limits.of the card.
AA and AF performance on the Radeon X1950 Pro is good enough, that you're actually missing out if you're using it only at default settings. Of all the games we tested today, some games are still very much playable with the Radeon X1950 Pro at 1280 x 1024 or even 1600 x 1200 with AA and AF enabled, except for SW: KOTOR, TES IV: Oblivion, Need for Speed: Carbon and F.E.A.R where playable frame rates are pretty much restricted to 1024 x 768. The same can't be said for the GeForce 7900GS. Frame rates at 1024 x 768 are too low to be really playable in TES IV Oblivion, Need for Speed: Carbon that not only do you have to run at default settings, you have to scale down the level of detail as well. The best the GeForce 7900GS has to offer is 1280 x 960 / 1024 with AA and AF on newer games. Older games such as Call of Duty, Homeworld 2 and Richard Burns Rally are fast enough at 1600 x 1200 with AA and AF, but frame rates are lower than on the Radeon X1950 Pro. A much 'balanced' competitor for the Radeon X1950 Pro is actually the GeForce 7950GT, which cost more but still have problems with Need for Speed: Carbon and still lacks the ability to run AA with SM 3.0 HDR enabled games.
[Previous Page]
[Go to top]
[Home]