Tech-Hounds.com

Because gamers play games, not benchmarks




Deeper with the GeForce 7 Series

PCI Express x16 graphics cards
Chipset GeForce 7800GT / 7800GTX 256 MB, 256 bit

After completing our round up, we've decided to conduct more testing to see just what the GeForce 7 series have to offer. Normally, we put up a comparison to its nearest competitor (ATI's Radeon X1800 series of cards), but we're not able to get samples at this time. Hopefully, we will be able to bring you a Radeon X1800 series round up next month. For now, we will be comparing the GeForce 7800GTX and 7800GT with reference clocks, since there are no standard for the higher clocked cards (each vendor used different clocks for their cards). Higher clocked cards should offer slightly higher frame rates (but not much).

From a rendering perspective, the GeForce 7 series is not that much different from NVIDIA's own GeForce 6 series. So, image quality will be the same if not similar between the two. However, there is one major difference - the GeForce 7 series can perform antialiasing on transparent (or alpha) textures, either using multisampling or supersampling. Naturally, we're very interested in this feature and just like we promised in our last article, we will focus on this feature in this article.

Another feature that is also a concern to us (and at least some of you) is of course high dynamic range (HDR) rendering. Although the GeForce 6 series is already capable of HDR rendering, performance with HDR is still too low to offer fluid gameplay. The GeForce 7 series with higher fill rate should be able to provide higher frame rates - we will be using Serious Sam II and Splinter Cell Chaos Theory for HDR performance test. However, just like with the GeForce 6 series, HDR rendering will not work with antialiasing on the GeForce 7 series. So, we will only touch this subject briefly.

In addition to testing with our usual graphics benchmarks, we also perform some additional testing of transparent textures antialiasing with our older game benchmarks as well. We won't be publishing the results (for reasons we will go into later), since we're focusing on compatibility and not performance with these older games. So, without further ado, let's move onward.

Overview

Transparency antialiasing

If you're lucky enough to already have one (or two) cards from the GeForce 7 series, you'll be able to find the option to turn on / off this feature in the driver panel, under the Performance and Quality Settings section. Of course, you'll have to choose Advanced Settings in the Global Driver Settings to actually see and change it. NVIDIA explains transparency antialiasing as

" an advanced feature, (is) a technique used to minimize the visible aliasing on the edges of images with transparent textures. This improvement in 3D image quality is most noticeable in objects such as vegetation, chain-link fences, and so on."

If you don't have much interest in 3D graphics, you may be wondering why is this so. Well, here's a quite simplified explanation. Look at a tree, chain-link fence in real life and in games. If game developers were to model the same objects in their game in exact detail, it will take too much time, but more importantly, too slow. So, instead of modeling every single leaf or chain-link, they took (or draw) a picture of a branch with leaves or a chain link fence, then put them into a single texture. To complete the illusion of a tree, they cut off the background and mark the empty area as transparent, so you will only see the branches, leaves or chain-link when the texture is finally rendered on screen. So, game developer can get away with much less polygons and just several textures for that tree or fence.



Since this is a new feature, you have to basically 'force' antialiasing on transparent textures. No game available right now supports transparency antialiasing out of the box. Thankfully, NVIDIA have managed to simplify this feature so you won't have to force the level of antialiasing (2x, 4x, 8xS) from the driver panel. You only need to turn on the option (and the transparency anti aliasing mode - off, multisampling or supersampling) on the driver panel. This way, you can still control the level of antialiasing from a game or application. Unfortunately, there's no way of using different levels of anti aliasing - for example, you may want to use 4x AA on non transparent textures and just 2x (or 2x Quincunx) AA on transparent textures or vice versa.

Now, let's take a look at what transparency antialiasing has to offer, image quality wise. Below you can see a series of screenshots taken from Brothers in Arms with plenty of vegetation. They are in order, without transparency antialiasing, with transparency multisampling antialiasing (MSAA) and last with transparency supersampling antialiasing (SSAA)



The easiest area to focus on is the leaves near the crosshair (the ones with the sky on the background). Well, we didn't really see that much difference - looks like even the performance is the same. You'll no doubt notice that by looking at the FRAPS fps counter on the top left of all three images. Looks like this can explained by two things: either there's no performance drop with any of the transparency antialiasing modes or transparency anti aliasing is not enabled at all (on all images). Now, let's take a look at our second game, Battlefield 2. Again, the images are arranged in this order: without transparency antialiasing, with transparency MSAA and last, with transparency SSAA.



Take a good look at the leaves on the lower left part, you could definitely see (if you look closely enough) transparency SSAA doing it's work. Image quality under transparency MSAA and no transparency AA is pretty similar. So, we can see that transparency SSAA does work with this game and can provide better image quality to transparency MSAA or no transparency AA. But it seems there's quite a big hit with transparency SSAA - look at the FRAPS fps counter. With no CPU overhead (bots, physics etc), the GeForce 7800GTX we used offer up to 120 fps - well above the 100 fps cap enabled by default in the game. With transparency SSAA, 71 fps is the highest we can go - that's close to a 50 fps performance hit. Seems like even supersampling is still too heavy for the GeForce 7800GTX.



Serious Sam II, just like its predecessor, makes heavy use of transparent textures. The only game rivaling it is FarCry. Here again we can see transparency SSAA providing a better image (with less aliasing) - look at the vines hanging off the trees. You can clearly see aliasing with transparency MSAA and no transparency AA. However, the load from doing transparency SSAA proved to be too much for the GeForce 7800GTX. Understandably so considering the amount of transparent textures in use - performance dropped to 25 fps or about 1/3 or the frame rates with transparency MSAA and no transparency antialiasing. What if we reduce the number of samples? Theoretically, it will be faster but will image quality suffer? Let's take a look at what you get with 2x transparency SSAA.



Image quality is pretty close to 4x transparency SSAA (there are very subtle differences), with less of a performance hit (38 fps instead of 25 fps - still about half though). We can conclude that transparency SSAA offer the most significant image quality boost (when it works) compared to both transparency MSAA and no transparency antialiasing. Of course, for them to work, the levels (or the game) must make use of transparent textures. If no transparent textures are in use, we won't see any image quality improvement or performance difference. We can also see the amount of transparent textures in use does influence the performance levels you're getting. We really can't recommend enabling 4x transparency SSAA with Serious Sam II, no matter how good it looks. That's why we wished NVIDIA made separate options to control the levels of antialiasing between normal and transparent textures. The best compromise of image quality and performance will probably be 4x MSAA and 2x transparency SSAA (with 16x AF).

Brothers in Arms seem to be quite an anomaly, since it does use transparent textures, but we can see no (visible) differences or even any performance difference for that matter. Looks like not all games will work correctly with transparency SSAA. Another game that seems to confirm this fact is F1 Career Challenge. Below, you could see screenshots displaying artifacts in the game when transparency SSAA is enabled. In actual gameplay (when you're racing on the track), you couldn't see anything at all and performance is so slow, you'd think the PC had crashed. The Forceware drivers do have a long standing bug with this game, so this may be causing the problem.


High Dynamic Range (HDR) Rendering

Next on our list is HDR rendering. Below you can see two screenshots from Splinter Cell Chaos Theory - the left is without HDR and the right is with HDR. You can see the HDR enabled image is slightly brighter, in some parts of the wall, and there are definitely a difference on the light bulb while the wall on the right side stays perfectly dark. The HDR image has more aliasing than the non HDR image (with 4x AA enabled). Looks like HDR's performance hit is about the same with 4xAA.



By enabling HDR in Serious Sam II, you'll notice much more details and colors when playing the game. Look at the clouds and the ground on the edge of the pathway - you'll definitely see the differences. So, HDR does bring us close to 'cinematic quality' in games and real time rendering in general. Again you'll notice the non HDR image (with 4x AA) has less aliasing than the HDR image on the right. Performance levels are about the same, though it looks like HDR is slightly slower in this game than enabling 4xAA with the GeForce 7800GTX and 7800GT. We'll take a closer look at the numbers later.



Unlike transparency antialiasing (or AA and AF in general), HDR can only be enabled if the application made use of it. With quite a performance drop on the GeForce 6 series, no wonder not many games made use of this effect. Hopefully, this will change in the future, partly due to more widespread use of HDR capable hardware from both ATI and NVIDIA.

Performance

We did lots of testing for this article, since we're using new set of drivers. Unfortunately, not all of them can be shown here, it will make this article too long. So, we limit the results to the our latest graphics benchmarks (F.E.A.R, Quake 4, Serious Sam II, Battlefield 2 and Brothers in Arms). Performance with transparency AA in older games are pretty much the same, whether you chose MSAA or SSAA.

Most of the games feature application controlled anti aliasing and anisotropic filtering, sans Battlefield 2 and Brothers In Arms which only have control over one setting (AA in BF2 and AF in Brothers in Arms). We enable AA and AF in-game when possible, either through the menu or console command. For Battlefield 2, anisotropic filtering was enabled through the driver panel as did AA for Brothers in Arms. Image quality setting is set to High Quality for best image quality. Without further ado, here are the results.

Our test setup
AMD Athlon 64 3500+ socket 939
2 x 256 MB Kingston KVR 3-3-3 PC3200 DDR-SDRAM
MSI K8N NForce 4 SLI motherboard
Maxtor DiamondMaxPlus9 80 GBs Serial ATA 8 MB buffer
ASUS E-616 DVD-ROM
450 watts ATX power supply

Windows XP Professional with Service Pack 2 installed
NVIDIA Forceware 81.95 reference driver
NVIDIA NForce 4 6.66 reference driver
Creative SoundBlaster Live! 24 bit 5.12.1.512 driver.
DirectX 9.0c

The results:

F.E.A.R Performance Test, 1024 x 768
GeForce
7800GTX
GeForce
7800GT
60
109
234
28 (4x AA 16x AF)
52 (4x AA 16x AF)
114 (4x AA 16x AF)
29 (4x AA 16x AF Alpha MSAA
53 (4x AA 16x AF Alpha MSAA
111 (4x AA 16x AF Alpha MSAA
29 (4x AA 16x AF Alpha SSAA)
52 (4x AA 16x AF Alpha SSAA)
101 (4x AA 16x AF Alpha SSAA)
.
59
94
195
27 (4x AA 16x AF)
46 (4x AA 16x AF)
91 (4x AA 16x AF)
25 (4x AA 16x AF Alpha MSAA
47 (4x AA 16x AF Alpha MSAA
102 (4x AA 16x AF Alpha MSAA
23 (4x AA 16x AF Alpha SSAA)
46 (4x AA 16x AF Alpha SSAA)
88 (4x AA 16x AF Alpha SSAA)
F.E.A.R Performance Test, 1280 x 960
GeForce
7800GTX
GeForce
7800GT
45
80
155
20 (4x AA 16x AF)
36 (4x AA 16x AF)
72 (4x AA 16x AF)
19 (4x AA 16x AF Alpha MSAA
38 (4x AA 16x AF Alpha MSAA
73 (4x AA 16x AF Alpha MSAA
20 (4x AA 16x AF Alpha SSAA)
37 (4x AA 16x AF Alpha SSAA)
74 (4x AA 16x AF Alpha SSAA)
.
38
68
134
17 (4x AA 16x AF)
32 (4x AA 16x AF)
65 (4x AA 16x AF)
18 (4x AA 16x AF Alpha MSAA
33 (4x AA 16x AF Alpha MSAA
65 (4x AA 16x AF Alpha MSAA
18 (4x AA 16x AF Alpha SSAA)
32 (4x AA 16x AF Alpha SSAA)
64 (4x AA 16x AF Alpha SSAA)
F.E.A.R Performance Test, 1600 x 1200
GeForce
7800GTX
GeForce
7800GT
37
54
108
13 (4x AA 16x AF)
25 (4x AA 16x AF)
51 (4x AA 16x AF)
14 (4x AA 16x AF Alpha MSAA
25 (4x AA 16x AF Alpha MSAA
56 (4x AA 16x AF Alpha MSAA
14 (4x AA 16x AF Alpha SSAA)
25 (4x AA 16x AF Alpha SSAA)
50 (4x AA 16x AF Alpha SSAA)
.
30
47
85
11 (4x AA 16x AF)
21 (4x AA 16x AF)
43 (4x AA 16x AF)
12 (4x AA 16x AF Alpha MSAA
22 (4x AA 16x AF Alpha MSAA
43 (4x AA 16x AF Alpha MSAA
11 (4x AA 16x AF Alpha SSAA)
22 (4x AA 16x AF Alpha SSAA)
41 (4x AA 16x AF Alpha SSAA)

Of course, there's no point of using more than 1024 x 768 with 4xAA in F.E.A.R with a single GeForce 7800GTX or 7800GT. Only pay attention to the 1024 x 768 graph. We can definitely see there's no performance difference between no transparency antialiasing, 4x transparency MSAA and SSAA. Not surprising really, since F.E.A.R made little use of transparent textures, particularly in the performance test.

F.E.A.R Performance Test, 1024 x 768. 32 bit fps Progress (7800GTX)
150 fps
100 fps
50 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

F.E.A.R Performance Test, 1024 x 768. 32 bit fps Progress (7800GT)
150 fps
100 fps
50 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.


GeForce 7800GT

Default 4x AA 16x AF Alpha MSAA
Alpha SSAA
<30 fps 0 4 6 5
30-45 fps 0 27 27 27
45-60 fps 2 14 12 13
60-90 fps 29 10 10 10
90-120 fps 15 0 6 0


GeForce 7800GTX

Default 4x AA 16x AF Alpha MSAA
Alpha SSAA
<30 fps 0 2 2 2
30-45 fps 0 19 18 20
45-60 fps 1 20 20 19
60-90 fps 24 10 11 11
90-120 fps 18 4 4 3

These fps progress graph confirms our conclusion. The differences between the three transparency antialiasing settings can be attributed to normal variations between runs. With little use of transparent textures throughout the game, there's no point of enabling transparency antialiasing in this game anyway. So it looks like the best image quality is still with 4x AA.

Quake 4, Data Processing Terminal, 1024 x 768
GeForce
7800GTX
GeForce
7800GT
67
121.841
179
57 (4x AA 16x AF)
86.279 (4x AA 16x AF)
138 (4x AA 16x AF)
57 (4x AA 16x AF Alpha MSAA
86.604 (4x AA 16x AF Alpha MSAA
137 (4x AA 16x AF Alpha MSAA
56 (4x AA 16x AF Alpha SSAA)
86.032 (4x AA 16x AF Alpha SSAA)
134 (4x AA 16x AF Alpha SSAA)
.
64
119.814
172
50 (4x AA 16x AF)
77.368 (4x AA 16x AF)
126 (4x AA 16x AF)
49 (4x AA 16x AF Alpha MSAA
76.837 (4x AA 16x AF Alpha MSAA
122 (4x AA 16x AF Alpha MSAA
50 (4x AA 16x AF Alpha SSAA)
76.732 (4x AA 16x AF Alpha SSAA)
123 (4x AA 16x AF Alpha SSAA)
Quake 4, Data Processing Terminal, 1280 x 1024
GeForce
7800GTX
GeForce
7800GT
62
113.633
177
37 (4x AA 16x AF)
58.92 (4x AA 16x AF)
105 (4x AA 16x AF)
37 (4x AA 16x AF Alpha MSAA
59.189 (4x AA 16x AF Alpha MSAA
104 (4x AA 16x AF Alpha MSAA
37 (4x AA 16x AF Alpha SSAA)
58.775 (4x AA 16x AF Alpha SSAA)
103 (4x AA 16x AF Alpha SSAA)
.
62
106.3
161
33 (4x AA 16x AF)
51.462 (4x AA 16x AF)
93 (4x AA 16x AF)
33 (4x AA 16x AF Alpha MSAA
51.479 (4x AA 16x AF Alpha MSAA
91 (4x AA 16x AF Alpha MSAA
33 (4x AA 16x AF Alpha SSAA)
51.444 (4x AA 16x AF Alpha SSAA)
93 (4x AA 16x AF Alpha SSAA)
Quake 4, Data Processing Terminal, 1600 x 1200
GeForce
7800GTX
GeForce
7800GT
60
95.664
156
29 (4x AA 16x AF)
42.709 (4x AA 16x AF)
82 (4x AA 16x AF)
26 (4x AA 16x AF Alpha MSAA
43.047 (4x AA 16x AF Alpha MSAA
80 (4x AA 16x AF Alpha MSAA
25 (4x AA 16x AF Alpha SSAA)
42.979 (4x AA 16x AF Alpha SSAA)
80 (4x AA 16x AF Alpha SSAA)
.
58
85.076
135
23 (4x AA 16x AF)
37.409 (4x AA 16x AF)
72 (4x AA 16x AF)
17 (4x AA 16x AF Alpha MSAA
37.527 (4x AA 16x AF Alpha MSAA
71 (4x AA 16x AF Alpha MSAA
17 (4x AA 16x AF Alpha SSAA)
37.341 (4x AA 16x AF Alpha SSAA)
69 (4x AA 16x AF Alpha SSAA)

Its the same story with Quake 4. While there are more transparent textures throughout the game than F.E.A.R, the Data Processing Terminal level doesn't have any - at least from what we can tell. If there were any transparent textures, the frame rates at 1024 x 768 is still high enough to allow very fluid, playable frame rates, although 1280 x 1024 may prove to be too much.

Quake 4 - Data Processing Terminal, 1024 x 768. 32 bit fps Progress (7800GTX)
90 fps
60 fps
30 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

Quake 4 - Data Processing Terminal, 1024 x 768. 32 bit fps Progress (7800GT)
90 fps
60 fps
30 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.


GeForce 7800GT

Default 4x AA 16x AF Alpha MSAA
Alpha SSAA
<30 fps 0 0 0 0
30-45 fps 0 0 0 0
45-60 fps 0 8 7 7
60-90 fps 3 31 32 32
90-120 fps 14 10 10 9


GeForce 7800GTX

Default 4x AA 16x AF Default 4x AA 16x AF
<30 fps 0 0 0 0
30-45 fps 0 0 0 0
45-60 fps 0 3 4 3
60-90 fps 3 32 21 24
90-120 fps 13 19 19 16

As expected, there are only slight variations between the three transparency antialiasing settings. There is something odd near the end of the level though, both the GeForce 7800GTX and 7800GT seems to drop quite significantly in default settings to AA performance levels. Looks like that's when the elevator hits the bottom, but even then performance is still above 60 fps on both cards at 1024 x 768 at that point.

Serious Sam II - Greendale, 1024 x 768, 32 bit
GeForce
7800GTX
GeForce
7800GT
53.8
69.9
92.3
51.4 (4x AA 16x AF)
65.5 (4x AA 16x AF)
82.5 (4x AA 16x AF)
49.7 (4x AA 16x AF Alpha MSAA
63 (4x AA 16x AF Alpha MSAA
77.8 (4x AA 16x AF Alpha MSAA
19.1 (4x AA 16x AF Alpha SSAA)
22.3 (4x AA 16x AF Alpha SSAA)
26.6 (4x AA 16x AF Alpha SSAA)
.
54.4
71.5
94.8
43.4 (4x AA 16x AF)
58.2 (4x AA 16x AF)
72.7 (4x AA 16x AF)
43.9 (4x AA 16x AF Alpha MSAA
54.4 (4x AA 16x AF Alpha MSAA
66.3 (4x AA 16x AF Alpha MSAA
15.3 (4x AA 16x AF Alpha SSAA)
18.2 (4x AA 16x AF Alpha SSAA)
22.2 (4x AA 16x AF Alpha SSAA)
Serious Sam II - Greendale, 1280 x 960, 32 bit
GeForce
7800GTX
GeForce
7800GT
53.5
69.4
92.3
39.7 (4x AA 16x AF)
49.4 (4x AA 16x AF)
60.8 (4x AA 16x AF)
36.5 (4x AA 16x AF Alpha MSAA
46.1 (4x AA 16x AF Alpha MSAA
57.1 (4x AA 16x AF Alpha MSAA
13.6 (4x AA 16x AF Alpha SSAA)
16 (4x AA 16x AF Alpha SSAA)
19.7 (4x AA 16x AF Alpha SSAA)
.
52.9
69.2
90.7
33.8 (4x AA 16x AF)
42.1 (4x AA 16x AF)
52.5 (4x AA 16x AF)
32 (4x AA 16x AF Alpha MSAA
39.1 (4x AA 16x AF Alpha MSAA
47.9 (4x AA 16x AF Alpha MSAA
10.8 (4x AA 16x AF Alpha SSAA)
13 (4x AA 16x AF Alpha SSAA)
16.2 (4x AA 16x AF Alpha SSAA)
Serious Sam II - Greendale, 1600 x 1200, 32 bit
GeForce
7800GTX
GeForce
7800GT
51
64.2
81.5
28.9 (4x AA 16x AF)
34.6 (4x AA 16x AF)
42.4 (4x AA 16x AF)
27 (4x AA 16x AF Alpha MSAA
32.1 (4x AA 16x AF Alpha MSAA
39.2 (4x AA 16x AF Alpha MSAA
7.5 (4x AA 16x AF Alpha SSAA)
11.1 (4x AA 16x AF Alpha SSAA)
14.6 (4x AA 16x AF Alpha SSAA)
.
43.1
54
68.8
24.4 (4x AA 16x AF)
29.3 (4x AA 16x AF)
36 (4x AA 16x AF)
23.4 (4x AA 16x AF Alpha MSAA
27 (4x AA 16x AF Alpha MSAA
32.6 (4x AA 16x AF Alpha MSAA
5.8 (4x AA 16x AF Alpha SSAA)
8.9 (4x AA 16x AF Alpha SSAA)
12.3 (4x AA 16x AF Alpha SSAA)

At last we can see just how much penalty 4x transparency SSAA incur on both cards. Looks like performance drops to around 1/3 of normal performance levels. So, it's even worse than what the numbers on the screenshots showed. We can also see that even 4x transparency MSAA incurs a slightly higher penalty than normal 4x AA. The GeForce 7800GTX fares slightly better than the GeForce 7800GT, but in all honesty, since we're not really achieving much in anti aliasing, it's better off to choose normal 4x AA than forcing 4x transparency MSAA.

Serious Sam II - Greendale, 1024 x 768. 32 bit fps Progress (7800GTX)
66 fps
44 fps
22 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

Serious Sam II - Greendale, 1024 x 768. 32 bit fps Progress (7800GT)
66 fps
44 fps
22 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.


GeForce 7800GT

Default 4x AA 16x AF Alpha MSAA
Alpha SSAA
<30 fps 0 1 1 63
30-45 fps 1 0 0 0
45-60 fps 11 40 60 0
60-90 fps 50 22 2 0
90-120 fps 1 0 0 0


GeForce 7800GTX

Default 4x AA 16x AF Alpha MSAA
Alpha SSAA
<30 fps 0 1 1 63
30-45 fps 1 0 0 0
45-60 fps 11 15 19 0
60-90 fps 50 47 43 0
90-120 fps 1 0 0 0

Here we can see more clearly the extra pipeline and processing power of the GeForce 7800GTX really helps it to offer a higher frame rate than the GeForce 7800GT. You only lose performance on peaks with the GeForce 7800GTX. Of course, just like what the numbers show, even a single GeForce 7800GTX is unable to cope with the demands of 4x transparency SSAA.

Battlefield 2 - Gulf of Oman, 1024 x 768, 32 bit
GeForce
7800GTX
GeForce
7800GT
56.25
72.53
86.75
57.5 (4x AA 16x AF)
76.588 (4x AA 16x AF)
87 (4x AA 16x AF)
61.33 (4x AA 16x AF Alpha MSAA)
76.529 (4x AA 16x AF Alpha MSAA)
86.33 (4x AA 16x AF Alpha MSAA)
50.2 (4x AA 16x AF Alpha SSAA)
67.772 (4x AA 16x AF Alpha SSAA)
79.6 (4x AA 16x AF Alpha SSAA)
.
57.75
77.35625
88.75
48.5 (4x AA 16x AF)
65.38975 (4x AA 16x AF)
80.5 (4x AA 16x AF)
47.6 (4x AA 16x AF Alpha MSAA)
65.086 (4x AA 16x AF Alpha MSAA)
79 (4x AA 16x AF Alpha MSAA)
48 (4x AA 16x AF Alpha SSAA)
61.3244 (4x AA 16x AF Alpha SSAA)
72.8 (4x AA 16x AF Alpha SSAA)
Battlefield 2 - Gulf of Oman, , 1280 x 960, 32 bit
GeForce
7800GTX
GeForce
7800GT
59.75
79.839
93.75
47.33 (4x AA 16x AF)
62.486 (4x AA 16x AF)
76 (4x AA 16x AF)
50.67 (4x AA 16x AF Alpha MSAA)
62.771 (4x AA 16x AF Alpha MSAA)
75.67 (4x AA 16x AF Alpha MSAA)
44.75 (4x AA 16x AF Alpha SSAA)
55.22575 (4x AA 16x AF Alpha SSAA)
67.5 (4x AA 16x AF Alpha SSAA)
.
53.67
72.364
84
42.25 (4x AA 16x AF)
51.17425 (4x AA 16x AF)
62 (4x AA 16x AF)
37.5 (4x AA 16x AF Alpha MSAA)
50.26375 (4x AA 16x AF Alpha MSAA)
63 (4x AA 16x AF Alpha MSAA)
34.25 (4x AA 16x AF Alpha SSAA)
46.35275 (4x AA 16x AF Alpha SSAA)
62 (4x AA 16x AF Alpha SSAA)
Battlefield 2 - Gulf of Oman, , 1600 x 1200, 32 bit
GeForce
7800GTX
GeForce
7800GT
50.75
74.218
90.75
37.25 (4x AA 16x AF)
46.3845 (4x AA 16x AF)
56 (4x AA 16x AF)
34.33 (4x AA 16x AF Alpha MSAA)
43.14 (4x AA 16x AF Alpha MSAA)
52.67 (4x AA 16x AF Alpha MSAA)
31.67 (4x AA 16x AF Alpha SSAA)
41.334 (4x AA 16x AF Alpha SSAA)
51.67 (4x AA 16x AF Alpha SSAA)
.
58.5
75.5435
86.25
32 (4x AA 16x AF)
38.37 (4x AA 16x AF)
46.33 (4x AA 16x AF)
30 (4x AA 16x AF Alpha MSAA)
38.3815 (4x AA 16x AF Alpha MSAA)
48.5 (4x AA 16x AF Alpha MSAA)
29.5 (4x AA 16x AF Alpha SSAA)
35.5985 (4x AA 16x AF Alpha SSAA)
43 (4x AA 16x AF Alpha SSAA)

With bots and physics in play, we can see that performance levels between the three transparency antialiasing is very small on both cards. So, it looks like Battlefield 2 is a perfect match for 4x transparency SSAA, at least up to 1024 x 768. In 1280 x 960, we're already seeing frame rates going to around 45 fps on the GeForce 7800GTX and 35 fps on the GeForce 7800GT. From experience, that means around 30 fps on the GTX and 20 fps on the GT in heavy duty gameplay.

Battlefield 2 - Gulf of Oman, 1024 x 768. 32 bit fps Progress (7800GTX)
67.5 fps
45 fps
22.5 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.