Tech-Hounds.com

Because gamers play games, not benchmarks




Deeper with the Radeon X1900 Series - Part 2

PCI Express x16 graphics cards
Chipset X1900 512 MB, 256 bit

Undoubtedly, this is an interesting year for 3D graphics. From the software side, we're beginning to see games making full use of hardware capabilities. Games are not only using larger, more detailed textures but also providing us with much more detail by creatively using shaders. Even games not traditionally associated with cutting edge graphics technology is getting on the bandwagon. UFO Aftershock, which basically is a turn based strategy, is relying on normal mapping to add detail that would have been too expensive to put in polygons. DirectX 9 is now a mature technology, though many could argue that it's still as 'fractured' as DirectX 8 was. But DirectX 9 was literally quite successful in getting 3D graphics moving ahead. High dynamic range rendering, flow control and conditional operations for shaders, and floating point processing is a far cry than water effects, bloom and environment bump mapping largely used by DirectX 8 games.

On the hardware front, both vendors have released their latest generation of graphics card, supporting SM 3.0 - on the high end, ATI with the Radeon X1800 and X1900 and NVIDIA with the GeForce 7800 and 7900 series. These cards are just in time for the latest batch of games, providing us with fairly high enough frame rates for comfortable gameplay at 1024 x 768, with 4x AA and 16x AF enabled. If you can stomach it, you might be able to play at 1280 x 960 or 1280 x 1024. However, if you care enough to look under the hood, you'll no doubt will come to the conclusion that both cards are in different stages of architectural development. The Radeon X1900 is a quite new approach to 3D graphics, focusing more on shaders than textures. It also come equipped with forward looking features that should maximize its performance when working with shaders. On the other hand, the relatively unchanged GeForce 7900 is pretty much the farthest NVIDIA can extend the G70 architecture. While that may seem bad, it's not. It is proof just how forward looking the design of the NV 40 was. It's still the faster card when dealing with mostly texture dependent games and it could rely on the brute force of its many pipelines (and what's inside those pipelines) to push shaders.

Both the Radeon X1900 and GeForce 7800 / 7900 is good and fast, but they're both far from perfect. Each is powerful where the other is not. So, it was aptly said that the battle between these cards are similar to the old DirectX 8 feud between the vendors - namely with the Radeon 8500 and the GeForce 3 / 3 Titanium. The GeForce 3 Titanium was also an extension of the GeForce 3 design. Hell, even the GeForce 4 was. NVIDIA took the basic GeForce 3, tweaking it bit by bit, both in design and fabrication. The result was the blinding success of all three generations, but in the end if you look at it, all three generations didn't bring anything new to 3D graphics beyond the GeForce 3 spec. That was not the case with Radeon 8500. It slightly push forward with a more advanced specification, the 'fractured' DirectX 8.1. But most people will remember it as the first card to show just how much anisotropic filtering can improve image quality. Even with a 'broken' implementation that only worked in with bilinear filtering, the Radeon 8500 (and later the 9000 and 9200) provided a taste of 3D graphics beyond trilinear filtering. The Radeon 8500 is also known as the first card to focus on 32 bit color at the expense of 16 bit color performance. That's the good points. For users anywhere, the Radeon 8500 is a nightmare with its plethora of driver bugs, incompatibilities and promises unfulfilled (remember VIDEOSHADER on the Radeon 8500?). Not to mention some dubious rendering practices. Those of you who once had to live with these cards know what we mean.

It is about this time that ATI made some bold move and started to 'clean house'. Driver development was the primary focus those days, but they also realized that they need something revolutionary for their next generation part. The result is R300 and the rest is history. Meanwhile, NVIDIA was lulled into a false sense of security by the success they achieved, both on the PC and console markets. By the time the Radeon 9700 was launched, they knew they're in trouble. One could blame bad design decisions or fabrication problems, but the truth remains. The NV30 was no match for the R300. The following months was not a pretty sight and mark some dubious bugs and 'optimizations' from the one time leader of consumer 3D graphics. In the end, the sobering truth made both companies a little wiser, as did reviewers and gamers alike. We're now more 'picky' about optimizations and image quality in general.

Back to the present, as we said before, the best from both vendors is still to come. But let's look at what's available here and now, and not some 6 or 9 months down the road. As promised, today we're going to take a look at the 'other' features of the Radeon X1900 - adaptive antialiasing and high dynamic range (HDR) rendering.

Overview

Adaptive Anti Aliasing

Adaptive antialiasing is a new feature present on the X1K series Radeons from ATI. As the name suggest, adaptive antialiasing is a feature that actually tries to perform antialiasing on transparent textures. It works very much like the transparency antialiasing available on the GeForce 7 series of cards. Like many other image enhancing feature on 3D graphics card these days, this option can be turned on and off from the graphics card's driver control panel. You can enable (or disable) this feature without messing around with the 'standard' antialiasing options, that means you'll be able to use this feature both with application controlled and driver forced antialiasing. You can also choose which method of adaptive antialiasing you want: Quality which uses supersampling for best quality at the expense of performance or Performance which rely on multisampling to maximize performance with slightly less effective antialiasing.

Like any image enhancing feature, we need to see just how much image quality improvement this feature offer. While there's no substitute to looking at them directly on the screen right in front of you, these screenshots will have to do for now. The first screenshot is adaptive antialiasing with Performance setting and the second one is with Quality. 4x AA was enabled from the in game menu for these screenshots.

Radeon X1900



If you look close enough you'll see the second screenshot (the one with supersampling) has less aliasing than the first. That's also true for the second set of screenshots. There are still some aliasing present, even with supersampling. Now, let's look at the GeForce 7 screenshots. Like the Radeon X1900 screenshots, the first one is with multisampling while the second is with supersampling.

GeForce 7



Up close, with multisampling we can see that ATI's adaptive antialiasing is more effective than NVIDIA's. But remember, it's the supersampling screenshots we're interested in. With supersampling, it looks like the GeForce 7 is actually more effective at removing aliasing on transparent textures than the Radeon X1900. This fact is slightly mudded by the 'thickening' apparent in all GeForce 7 transparency antialiasing shots. This 'thickening' can be seen on all transparent textures, not just on the palm you can also see it on the grass, trees, etc.

One can argue for either camp, but during repeated gameplay sessions, we notice less aliasing on the GeForce 7 than on the Radeon X1900 with transparency supersampling antialiasing enabled. Aliasing can easily be seen on faraway objects with transparent textures on the Radeon X1900, where the GeForce 7 had none or just a small degree of aliasing. However, when the objects are quite near, ATI's adaptive antialiasing and NVIDIA's transparency antialiasing is at par with each other. Unfortunately, we can't show you this fact, so the static screenshots will have to do for now.

Let's see whether that's true with other games.

Radeon X1900



GeForce 7



Serious Sam II is literally filled with transparent textures, so it's perfect for evaluating adaptive and transparency antialiasing. You'll no doubt notice that there's no 'thickening' here on the GeForce 7. First of, the multisampling shots. You'll quickly notice that ATI's implementation is once again superior to NVIDIA - we notice less aliasing (on the vines) on the X1900 screenshots than we do with the GeForce 7's screenshot. As for the supersampling screenshot, it's hard to see the differences. If we were only to rely on these static screenshots, we have to say that ATI's adaptive antialiasing is as effective as NVIDIA's transparency antialiasing. But it's truly not. In motion, we notice some aliasing still present on the Radeon X1900, where the GeForce 7 has none or very little. NVIDIA's transparency antialiasing is more effective than ATI's adaptive antialiasing when supersampling.

Need more proof? Well, so far, we've only seen screenshots with 4x antialiasing. Below are some screenshots from both cards with 2x adaptive and transparency antialiasing. With less samples, we can see more aliasing so it will be easier to spot the differences. Plus, we can also see which one offer the best 2x antialiasing for transparent textures. Supersampling only, of course.

Radeon X1900



GeForce 7



With the thickening present, it's still hard to decide which one is the less aliased image. You'll have to look real close to see the differences. But in the end, NVIDIA's transparency antialiasing prove superior to ATI's adaptive antialiasing.Let's see the Serious Sam II screenshots. With no thickening issues, it should be easier to judge which one is more effective at antialiasing transparent textures.

Radeon X1900



GeForce 7



With less samples, it's easier to spot the differences between the two. The trick is to differentiate between polygon edges and transparent textures. Look at the leaves on faraway trees, you'll definitely see more aliasing on he Radeon X1900 than on the GeForce 7. An easier spot will be on the right: look at the leaves on the tree on the right. Once you know what to look for, you can see the same thing on the vines of tree in front of Sam.

So, unlike 'standard' antialiasing, NVIDIA has the upper hand in quality with transparency antialiasing. Because of this, we can't really do an apples to apples comparison here. Keep this in mind when you're looking at the benchmark results.

High Dynamic Range (HDR) Rendering

With SM 3.0 adoption, the Radeon X1K series are the first Radeon to directly support the preferred HDR SM 3.0 method used in games like Farcry and Splinter Cell Chaos Theory. However, unlike the GeForce 7 series, the Radeon X1K can still run with AA enabled when a game (or an application) make use of HDR. Let's take a look at Splinter Cell Chaos Theory first. The first screenshot is without HDR and the second one is with HDR. The third screenshot from the GeForce 7 is with HDR enabled, but tone mapping disabled.

Radeon X1900



GeForce 7



Even without HDR, you can already see there are some differences between the X1900 and GeForce 7 shots, though not by much. The far side of the room (on the right) received more light on the GeForce 7 than on the Radeon X1900. You can also see some differences on the light bulb. Everything else is similar if not the same. With HDR enabled, the differences are more noticeable. In the GeForce 7 screenshot, the room is more lit than on the Radeon X1900. You'll can also see some differences on the light bulb. It is only when we turn off tone mapping on the GeForce 7 do get pretty much the same picture from both cards. So, is ATI 'skipping' a step here? We're not sure, but we decided to perform testing on the GeForce 7 both with and without tone mapping enabled.

Radeon X1900



GeForce 7



With the latest patch, you can enable both HDR and AA on the Radeon X1K series in Serious Sam II. Unlike Splinter Cell Chaos Theory, we can see virtually no difference between HDR on the Radeon X1900 and GeForce 7 series. The detail in the dirt path and clouds are present on both, as also the lens flare on the water fall (which the non HDR image lack).

It boils down to this. ATI's adaptive antialiasing is somewhat a mixed bag. In Performance mode, the image quality it offers is better than NVIDIA under the same setting. However, this is not the preferred mode of use. In Quality mode, which corresponds to supersampling on the GeForce 7 series, the Radeon adaptive antialiasing is not as effective. It only works well on objects that are near to the camera. There's still very noticeable aliasing (more so in motion) on faraway objects with transparent textures. Gamers looking for the best image quality in games with transparent textures, should be more happy with the GeForce 7 than the Radeon X1900.

While this may seem like a big deal, it's really not. Not all games or all game levels / maps utilize transparent textures. Even those who do will not always work with both ATI's adaptive antialiasing and NVIDIA's transparency antialiasing. NVIDIA is again the better card here - its 8xS can be used in such games. We haven't looked (yet) at the X1900 performance with adaptive antialiasing, but we already seen what the GeForce 7 can do. The performance penalty can be very high, higher than just applying AA alone and that's a lot considering the performance penalty of antialiasing on the GeForce 7 series. In some cases, you'll have to drop the number of samples to 2x to maintain playable frame rates. Or maybe buy another card to run the game under SLI.

Although the GeForce 7 offer the better image quality (transparent textures wise), transparency antialiasing on both cards is not really a feature worth buying for. Look at the options. A card that looks good but ran slowly or one that's not good but is possibly faster. This matter is very subjective, but that sentence is pretty much our view on the subject.

HDR is different. Quality wise, image quality from both cards is pretty much similar and comparable to each other. Theoretically, ATI have the upper hand here with the Radeon X1K supporting HDR with AA. Keep in mind though, current games supporting HDR must first be patched to enable HDR with AA. Future games making use of HDR should come with HDR and AA support out of the box. In this regard, HDR is more important than transparency antialiasing in performance evaluations. It's also easier since we can actually perform an apples to apples comparison. For this article, we decided to include benchmark results both with just HDR and HDR plus AA and AF. Just keep in mind that you should only compare the HDR only scores between the cards, since the GeForce 7 series don't support HDR and AA.

Performance

We didn't change our test setup from the last article, it's pretty much the same. We've included results from both a standard clocked and a higher clocked GeForce 7800GTX (the ASUS EN7800GTX TOP) with comparable settings (standard 4x AA 16x AF and with transparency antialiasing enabled). Of course, we chose supersampling for best quality The X1900 XTX is still our choice for this article to represent the X1900 series. The X1900 XTX was tested in standard 4x AA 16xAF and with adaptive antialiasing enabled (under both Performance and Quality settings).

We'd like to thank both Tagan and Kingston for supplying the additional power supply and 1 GB memory modules for this article. We'd also like to thank Gigabyte for providing the Radeon X1900 XTX sample card.

Our test setup
AMD Athlon 64 3500+ socket 939
2 x 1024 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
Tagan TG530-U15 530 watts ATX/BTX power supply

Windows XP Professional with Service Pack 2 installed
ATI Catalyst 6.2 reference driver
NVIDIA Forceware 81.98 reference driver
NVIDIA NForce 4 6.66 reference driver
Creative SoundBlaster Live! 24 bit 5.12.1.512 driver.
DirectX 9.0c

The graphs are pretty self explanatory, but in case you can't see the text, they're arranged in groups of three colors: green for minimum fps, blue for average fps and red for maximum fps. The first group represent test results under 4xAA and 16x AF and the second represent results with 4xAA and 16xAF with adaptive or transparency antialiasing enabled and set to supersampling mode (or Quality on the Radeon X1900 XTX). For the X1900XTX, the third group represent results with 4xAA and 16x AF with adaptive antialiasing enabled and set to multisampling mode (or Performance). These results are in frame rate per second.

The results:

Richard Burns Rally - Harwood Forest (Transparency AA), 1024 x 768, 32 bit
7800GTX TOP
7800GTX
X1900XTX
110 (4x AA 16x AF)
157.676 (4x AA 16x AF)
227 (4x AA 16x AF)
55 (4x SSAA 16x AF)
96.174 (4x SSAA 16x AF)
214 (4x SSAA 16x AF)
.
75 (4x AA 16x AF)
152.746 (4x AA 16x AF)
227 (4x AA 16x AF)
41 (4x SSAA 16x AF)
85.665 (4x SSAA 16x AF)
200 (4x SSAA 16x AF)
.
108 (4x AA 16x AF)
159.65 (4x AA 16x AF)
234 (4x AA 16x AF)
68 (4x SSAA 16x AF)
118.175 (4x SSAA 16x AF)
224 (4x SSAA 16x AF)
103 (4x MSAA 16x AF)
146.939 (4x MSAA 16x AF)
226 (4x MSAA 16x AF)
Richard Burns Rally - Harwood Forest (Transparency AA), 1280 x 1024, 32 bit
7800GTX TOP
7800GTX
X1900XTX
94 (4x AA 16x AF)
138.237 (4x AA 16x AF)
227 (4x AA 16x AF)
41 (4x SSAA 16x AF)
69.517 (4x SSAA 16x AF)
152 (4x SSAA 16x AF)
.
48 (4x AA 16x AF)
124.459 (4x AA 16x AF)
221 (4x AA 16x AF)
37 (4x SSAA 16x AF)
61.81 (4x SSAA 16x AF)
135 (4x SSAA 16x AF)
.
109 (4x AA 16x AF)
151.812 (4x AA 16x AF)
222 (4x AA 16x AF)
47 (4x SSAA 16x AF)
87.416 (4x SSAA 16x AF)
188 (4x SSAA 16x AF)
77 (4x MSAA 16x AF)
120.297 (4x MSAA 16x AF)
221 (4x MSAA 16x AF)
Richard Burns Rally - Harwood Forest (Transparency AA), 1600 x 1200, 32 bit
7800GTX TOP
7800GTX
X1900XTX
71 (4x AA 16x AF)
104.789 (4x AA 16x AF)
177 (4x AA 16x AF)
33 (4x SSAA 16x AF)
52.607 (4x SSAA 16x AF)
106 (4x SSAA 16x AF)
.
37 (4x AA 16x AF)
92.533 (4x AA 16x AF)
155 (4x AA 16x AF)
29 (4x SSAA 16x AF)
46.648 (4x SSAA 16x AF)
94 (4x SSAA 16x AF)
.
88 (4x AA 16x AF)
129.138 (4x AA 16x AF)
219 (4x AA 16x AF)
34 (4x SSAA 16x AF)
67.51 (4x SSAA 16x AF)
143 (4x SSAA 16x AF)
58 (4x MSAA 16x AF)
93.523 (4x MSAA 16x AF)
182 (4x MSAA 16x AF)

Unsurprisingly, the Radeon X1900 XTX is faster than the GeForce 7800GTX and quite by a significant margin under the same settings. After all, its not really achieving much, so it's probably doing less work. While minimum frame rates is pretty similar on both the higher clocked GeForce 7800GTX and the Radeon X1900 XTX, the Radeon offer a slightly higher average frame rate, even at 1600 x 1200. It's a hollow victory, really because we know the quality is better on both GeForce 7800GTX.

Richard Burns Rally - Harwood Forest (Transparency AA), 1024 x 768. 32 bit fps Progress
105 fps
70 fps
35 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.


GeForce 7800GTX TOP

GeForce 7800GTX

Radeon X1900 XTX

4x AA 16x AF 4x SSAA 16x AF 4x AA 16x AF 4x SSAA 16x AF 4x AA 16x AF 4x SSAA 16x AF
<30 fps 0 0 0 5 0 0
30-45 fps 0 78 0 114 0 0
45-60 fps 0 142 0 124 0 0
60-90 fps 38 40 141 18 1 43
90-120 fps 213 1 117 0 120 120
>120 fps1003014098
Total261261261261261261
results are in seconds

This table and graph again show just how heavy transparency antialiasing is on the GeForce 7800GTX. The frame rate dropped from around 60 - 120 fps to 30 - 90 fps with supersampling transparency antialiasing. Mind you, this is at 1024 x 768 and judging by the numbers, you really don't want to use a higher resolution. Even the Radeon X1900 XTX is feeling the pinch, spending more time below 90 fps with adaptive antialiasing (43 seconds as opposed to 1 second).

Serious Sam II - Greendale (Transparency AA), 1024 x 768, 32 bit
7800GTX TOP
7800GTX
X1900XTX
44 (4x AA 16x AF)
58.496 (4x AA 16x AF)
76 (4x AA 16x AF)
16 (4x SSAA 16x AF)
19.069 (4x SSAA 16x AF)
25 (4x SSAA 16x AF)
.
44 (4x AA 16x AF)
54.14 (4x AA 16x AF)
67 (4x AA 16x AF)
15 (4x SSAA 16x AF)
17 (4x SSAA 16x AF)
22 (4x SSAA 16x AF)
.
33 (4x AA 16x AF)
49.037 (4x AA 16x AF)
65 (4x AA 16x AF)
17 (4x SSAA 16x AF)
23.841 (4x SSAA 16x AF)
32 (4x SSAA 16x AF)
29 (4x MSAA 16x AF)
37.414 (4x MSAA 16x AF)
48 (4x MSAA 16x AF)
Serious Sam II - Greendale (Transparency AA), 1280 x 960, 32 bit
7800GTX TOP
7800GTX
X1900XTX
39 (4x AA 16x AF)
46.139 (4x AA 16x AF)
58 (4x AA 16x AF)
12 (4x SSAA 16x AF)
13.914 (4x SSAA 16x AF)
18 (4x SSAA 16x AF)
.
35 (4x AA 16x AF)
40.653 (4x AA 16x AF)
50 (4x AA 16x AF)
10 (4x SSAA 16x AF)
12.453 (4x SSAA 16x AF)
16 (4x SSAA 16x AF)
.
33 (4x AA 16x AF)
43.086 (4x AA 16x AF)
55 (4x AA 16x AF)
12 (4x SSAA 16x AF)
14.784 (4x SSAA 16x AF)
20 (4x SSAA 16x AF)
20 (4x MSAA 16x AF)
24 (4x MSAA 16x AF)
32 (4x MSAA 16x AF)
Serious Sam II - Greendale (Transparency AA), 1600 x 1200, 32 bit
7800GTX TOP
7800GTX
X1900XTX
27 (4x AA 16x AF)
32.776 (4x AA 16x AF)
41 (4x AA 16x AF)
8 (4x SSAA 16x AF)
9.79 (4x SSAA 16x AF)
13 (4x SSAA 16x AF)
.
25 (4x AA 16x AF)
28.804 (4x AA 16x AF)
36 (4x AA 16x AF)
7 (4x SSAA 16x AF)
8.715 (4x SSAA 16x AF)
12 (4x SSAA 16x AF)
.
28 (4x AA 16x AF)
34.204 (4x AA 16x AF)
43 (4x AA 16x AF)
9 (4x SSAA 16x AF)
10.77 (4x SSAA 16x AF)
15 (4x SSAA 16x AF)
15 (4x MSAA 16x AF)
17.59 (4x MSAA 16x AF)
23 (4x MSAA 16x AF)

Even the slightly optimized adaptive antialiasing on the Radeon X1900 XTX won't do much in this game. It's simply too heavy for current cards to run with 4x supersampling transparency antialiasing. You'll have to drop to 2x (or 2xQ on the GeForce 7 series) if you want the best compromise between image quality and playable frame rates. There's is one interesting fact here. It appears that the Radeon X1900 XTX is doing some additional work with multisampling than just standard 4x AA and 16x AF. Interesting, but not really useful since there's still very visible aliasing. We won't display the 2x results here, because they're pretty much the same (although nominally they are higher).

Serious Sam II - Greendale (Transparency AA), 1024 x 768, 32 bit fps Progress
41 fps
27.5 fps
13 fps
0 fps
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.


GeForce 7800GTX TOP

GeForce 7800GTX

Radeon X1900 XTX

4x AA 16x AF 4x SSAA 16x AF 4x AA 16x AF 4x SSAA 16x AF 4x AA 16x AF 4x SSAA 16x AF
<30 fps 0 63 0 63 0 62
30-45 fps 0 0 1 0 29 1
45-60 fps 44 0 57 0 33 0
60-90 fps 19 0 5 0 1 0
90-120 fps 0 0 0 0 0 0
>120 fps000000
Total636363636363
results are in seconds

With supersampling, neither card can offer sustained frame rate over the minimum 30 fps mark all the way through the replay / demo. They're pretty much bunched up below 30 fps. You have to use 2x antialiasing if you want higher frame rates. Even then, you frame rate will still occasionally drop to below 30 fps at one time or another. This game proves that transparency antialiasing is only applicable to older games. Next, we move HDR test results.

Serious Sam II - Greendale (HDR), 1024 x 768, 32 bit
7800GTX TOP
7800GTX
X1900XTX
43 (16x AF)
53.93 (16x AF)
70 (16x AF)
.
41 (16x AF)
49.432 (16x AF)
63 (16x AF)
.
37 (16x AF)
48.109 (16x AF)
64 (16x AF)
34 (4x AA 16x AF)
44.525 (4x AA 16x AF)
59 (4x AA 16x AF)
Serious Sam II - Greendale (HDR), 1280 x 960, 32 bit
7800GTX TOP
7800GTX
X1900XTX
33 (16x AF)
40.385 (16x AF)
52 (16x AF)
.
30 (16x AF)
36.142 (16x AF)
47 (16x AF)
.
33 (16x AF)
43.6 (16x AF)
57 (16x AF)
28 (4x AA 16x AF)
36.21 (4x AA 16x AF)
47 (4x AA 16x AF)
Serious Sam II - Greendale (HDR), 1600 x 1200, 32 bit
7800GTX TOP
7800GTX
X1900XTX
23 (16x AF)
28.582 (16x AF)
37 (16x AF)
.
21 (16x AF)
25.524 (16x AF)
34 (16x AF)
.
29 (16x AF)
36.729 (16x AF)
46 (16x AF)
21 (4x AA 16x AF)
26.46 (4x AA 16x AF)
34 (4x AA 16x AF)

Serious Sam II is the only game in our benchmark suite that supports both HDR and AA at the same time. We can see that enabling AA with HDR come with some penalty, though not really that much at 1024 x 768. It's only at higher resolution that the Radeon X1900 XTX begin to falter. Its still somewhat playable at 1280 x 960, but you can forget about 1600 x 1200.

With just HDR, the Radeon X1900 XTX is faster once we get above 1024 x 768, while the GeForce 7800GTXs are faster at 1024 x 768. That lead allowed the Radeon X1900 XTX to push enough frame rates close to the 30 fps minimum mark at 1600 x 1200, while the GeForce 7800GTX can only offer comparable frame rates at 1280 x 960. Clearly, the Radeon X1900 XTX is faster with HDR here. With future games making use of HDR, this bode quite well for the Radeon X1900. You can still use HQ AF and at least 4x AA with HDR enabled, if your frame rates were high enough to begin with.

Splinter Cell Chaos Theory - Lighthouse (HDR), 1024 x 768, 32 bit
7800GTX TOP
7800GTX
X1900XTX
50.98 (16x AF)
83.79 (16x AF)
155.178 (16x AF)
47.74 (16x AF)
81.03 (16x AF)
155.70 (16x AF)
.
45.39 (16x AF)
74.92 (16x AF)
153.44 (16x AF)
42.34 (16x AF)
72.48 (16x AF)
155.36 (16x AF)
.
61 (16x AF)
99.59 (16x AF)
154.95 (16x AF)
(4x AA 16x AF)
(4x AA 16x AF)
(4x AA 16x AF)
Splinter Cell Chaos Theory - Lighthouse (HDR), 1280 x 960, 32 bit
7800GTX TOP
7800GTX
X1900XTX
35.5 (16x AF)
59.15 (16x AF)
136.64 (16x AF)
33.52 (16x AF)
57.54 (16x AF)
136.06 (16x AF)
.
31.07 (16x AF)
54.86 (16x AF)
139.8 (16x AF)
29.77 (16x AF)
51.11 (16x AF)
133.48 (16x AF)
.
40.62 (16x AF)
77.84 (16x AF)
147.89 (16x AF)
Splinter Cell Chaos Theory - Lighthouse (HDR), 1600 x 1200, 32 bit
7800GTX TOP
7800GTX
X1900XTX
26.91 (16x AF)
44.4 (16x AF)
114.89 (16x AF)
25.65 (16x AF)
43.39 (16x AF)
114.18 (16x AF)
.
23.87 (16x AF)
39.4 (16x AF)
102.51 (16x AF)
22.70 (16x AF)
38.49 (16x AF)
100.87 (16x AF)
.
34.47 (16x AF)
59.95 (16x AF)
134.69 (16x AF)
(4x AA 16x AF)
(4x AA 16x AF)
(4x AA 16x AF)

No doubt about it, the Radeon X1900 XTX is faster. Splinter Cell Chaos Theory's heavy shader use is also helping a lot here. Even at 1600 x 1200, we still have high enough frame rates, with minimum frame rates of around 35 fps. On the other hand, only the higher clocked GeForce 7800GTX can muster the same level of performance, but at lower resolutions.

Quite odd though, turning off tone mapping actually caused a decrease in performance on the GeForce 7800GTX. With comparable image quality from both cards, it's definitely a no brainer here. The Radeon X1900 XTX wins, hands down.

Conclusion:

ATI's latest offering, the Radeon X1900 XTX is strong, but it does have its weak points. We already covered the lack of performance with texture dependent games in our last article, and now we discovered that ATI's adaptive antialiasing is not as effective as NVIDIA's transparency anti aliasing. You don't really gain that much image quality improvement, even with Quality setting.

However, we also saw that ATI is putting the Radeon X1900 XTX as a forward looking part. HDR performance is way stronger than the GeForce 7800GTX, even in texture dependent games like Serious Sam II. In shader heavy games like Splinter Cell Chaos Theory, the Radeon X1900 XTX is simply on a higher level than the GeForce 7800GTX. Of course, the icing on the cake is the ability to use AA with HDR, something the GeForce 7800GTX lacks.

Overall, while the Radeon X1900 is not always the fastest card around in all games, it is faster with AA and AF and it is faster in new, shader heavy games. Even with texture dependent games, the Radeon X1900 is still able to supply gamers with more than enough frame rates and in most cases still retain the lead with AA and AF enabled. Not only that, it also bring to the table higher quality visuals (HQ AF) with virtually no performance loss.

Go to top
Disclaimer and Privacy policy.