ATI AND NVIDIA are locked in an epic battle for the graphics performance crown, but as sexy and exciting as technology leadership can be, sometimes it’s hard to get really excited about high-end graphics cards. ATI and NVIDIA’s latest flagships, the Radeon 9800 XT and GeForce FX 5900 Ultra, are both capable of rendering stunning environments with fluid frame rates, but $500 price tags keep the cards out of the hands of those of us who don’t have spare organs to hawk on eBay.
Fortunately, the fancy technology found in most high-end graphics cards eventually trickles down to more affordable mid-range products. Mid-range cards might not have enough horsepower to run the latest games at the highest resolutions with antialiasing and anisotropic filtering cranked all the way up, but they’re generally fast enough for all but the most demanding gamers.
Last year, NVIDIA’s GeForce4 Ti 4200 owned the mid-range graphics market, but this year has been dominated by ATI. ATI took over the mid-range graphics performance crown with the Radeon 9500 Pro, which was succeeded by the Radeon 9600 Pro. NVIDIA’s GeForce FX 5600s haven’t been able to keep up. Not content to sit idle, today ATI is beefing up its mid-range graphics line with the Radeon 9600 XT. The 9600 XT promises to set a new standard in affordable graphics performance, but it is really that much faster than the competition? Read on to find out.
The RV360 GPU
The Radeon 9600 XT is based on ATI’s new RV360 GPU, which is quite similar to the RV350 chip found in the Radeon 9600 Pro. The RV360 shares the RV350’s 4×1-pipe architecture and a host of other features that you can read about in my Radeon 9600 Pro review. Rather than rehash all the technology found in the RV360, I’d rather focus on what’s new in the chip. ATI snuck a few surprises into the RV360 that are worth exploring.
Like the recently announced R360 GPU, which powers the Radeon 9800 XT, the RV360 supports GPU core temperature monitoring. Temperature monitoring is necessary for ATI’s new OVERDRIVE automatic overclocking software, which will come to the Radeon 9600 XT in the Catalyst 3.9 driver release, slated for November. (The RV360 has all the necessary hardware support for OVERDRIVE to work, but the Cat 3.9s aren’t ready yet.) Since OVERDRIVE will initially only offer the 9600 XT overclocked speeds of 513 and 527MHz, old fashioned overclocking may be a route for experienced enthusiasts.
To help give OVERDRIVE plenty of clock speed headroom, RV360 GPUs are being fabbed on a 0.13-micron manufacturing process using a special “Black Diamond” insulator that has less capacitance than the Fluorine-doped silicate glass insulator found in the RV350. Low capacitance (low-k) insulators can help chips reach higher clock speeds, which explains why ATI is able to clock the RV360 GPU at an even 500MHz on the Radeon 9600 XT100MHz higher than the Radeon 9600 Pro. The fact ATI is rolling out OVERDRIVE support for the 9600 XT suggests the chip can handle clock speeds north of 500MHz, too.
If you look really close, you can almost see the low-k insulator
I know you can’t wait to see benchmarks, but first, let’s have a quick peek at the Radeon 9600 XT’s specs.
|Peak pixel fill rate||2000 Mpixels/s|
|Texture units/pixel pipeline||1|
|Textures per clock||4|
|Peak texel fill rate||2000 Mtexels/s|
|Memory type||BGA DDR SDRAM|
|Memory bus width||128-bit|
|Peak memory bandwidth||9.6GB/s|
|Ports||VGA, DVI, composite and S-Video outputs
Composite, S-Video inputs
|Auxiliary power connector||None|
If you ignore low-k insulators and core tweaks, the Radeon 9600 XT is basically a faster version of the Radeon 9600 Pro. We’ll see just how much faster in a moment. However, before we get into testing, let’s take in the beauty that is the Radeon 9600 XT:
The Radeon 9600 XT’s heat sink is similar to what you might find on a Radeon 9700 Pro
The card’s BGA memory chips are rated for operation at 300MHz
And no, ATI still isn’t offering dual DVI on its consumer graphics cards
Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run three times, and the results were averaged.
Our test system was configured like so:
|Processor||Athlon XP ‘Thoroughbred’ 2600+ 2.083GHz|
|Front-side bus||333MHz (166MHz DDR)|
|Motherboard||DFI LANParty NFII Ultra|
|Chipset||NVIDIA nForce2 Ultra 400|
|North bridge||nForce2 Ultra 400 SPP|
|South bridge||nForce2 MCP-T|
|Chipset drivers||NVIDIA 2.45|
|Memory size||512MB (2 DIMMs)|
|Memory type||Corsair XMS3200 PC2700 DDR SDRAM (333MHz)|
|Graphics card||GeForce FX 5600 Ultra 128MB||Radeon 9600 Pro 128MB
Radeon 9600 XT 128MB
|Graphics driver||Detonator FX 45.23||CATALYST 3.8|
|Storage||Maxtor DiamondMax Plus D740X 7200RPM ATA/100 hard drive|
|OS||Microsoft Windows XP Professional|
|OS updates||Service Pack 1, DirectX 9.0b|
We’re comparing the Radeon 9600 XT to its most direct competitors, the Radeon 9600 Pro, and NVIDIA’s GeForce FX 5600 Ultra. I’ve elected to use only publicly available and supported drivers for this review, which means you won’t find any results using the various beta versions of NVIDIA’s “Release 50” drivers. We’ll be exploring the performance and image quality of NVIDIA’s next official driver release when the driver is finalized, WHQL-certified, and available to the general public.
The test system’s Windows desktop was set at 1024×768 in 32-bit color at a 75Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- FutureMark 3DMark03 Build 330
- Codecreatures Benchmark Pro
- Comanche 4 demo benchmark
- Quake III Arena v1.31 with trdemo1.dm_67
- Wolfenstein: Enemy Territory with demo0000.dm_82
- Serious Sam SE v1.07 with Demo0003
- Unreal Tournament 2003 with trtest1.dem
- Splinter Cell v1.2 with TRKalinatekDemo.bin
- Tomb Raider: Angel of Darkness v49 patch
- Gun Metal benchmark v1.20
- ShaderMark 2.0
- rthdribl 1.2
- Halo 1.02
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Theoretical fill rate and memory bandwidth peaks don’t necessarily dictate real world performance, but they’re a good place to start.
|Core clock (MHz)||Pixel pipelines||Peak fill rate (Mpixels/s)||Texture units per pixel pipeline||Peak fill rate (Mtexels/s)||Memory clock (MHz)||Memory bus width (bits)||Peak memory bandwidth (GB/s)|
|Radeon 9600 Pro||400||4||1600||1||1600||600||128||9.6|
|Radeon 9600 XT||500||4||2000||1||2000||600||128||9.6|
|GeForce FX 5600 Ultra||400||4||1600||1||1600||800||128||12.8|
ATI hasn’t done anything to increase the Radeon 9600 XT’s memory bandwidth over the 9600 Pro, but XT’s pixel and texture fill rates lead the pack.
Despite its higher theoretical pixel fill rate peak, the Radeon 9600 XT can’t catch the GeForce FX 5600 Ultra in 3DMark03’s single-textured fill rate test. However, the tables turn when we look at multi-textured performance, where the 9600 XT has a huge lead over even the 9600 Pro.
The Radeon 9600 XT improves upon the 9600 Pro’s already impressive pixel shader performance by a small margin, but the XT really shines in vertex shader performance. A 50% improvement in vertex shader performance suggests maybe ATI spent a little time tweaking more than just the RV360’s clock speeds.
ShaderMark 2.0 is brand new and includes some anti-cheat measures to prevent drivers from applying questionable optimizations. The Radeons run the benchmark with straight pixel shader 2.0 code, but I’ve included results for the GeForce FX with partial-precision and extended pixel shaders, as well.
The Radeons are in a class all their own in ShaderMark 2.0, and the GeForce FX 5600 Ultra’s performance is nothing short of embarrassing. The FX is way behind the Radeons across the board, and refuses to cooperate with a number of the shader tests.
Focusing on the Radeons, the 9600 XT is about 20% faster than the 9600 Pro, which nicely matches the 20% core clock speed difference between the two cards.
Quake III Arena
Wolfenstein: Enemy Territory
Unreal Tournament 2003
Wolfenstein: Enemy Territory runs fastest on the GeForce FX 5600 Ultra, but Quake III Arena and Unreal Tournament 2003 perform better on the Radeon 9600 XT.
Codecreatures Benchmark Pro
Gun Metal benchmark
The Radeon 9600 XT leads the way in Comanche 4 and Codecreatures, but the card can’t quite keep up with the GeForce FX 5600 Ultra in the Gun Metal benchmark.
Serious Sam SE
The GeForce FX 5600 Ultra maintains its lead in Serious Sam SE with and without antialiasing and anisotropic filtering.
In Splinter Cell, all the cards are closely matched. Overall, the Radeon 9600 XT is the fastest of the lot, but not by much.
In 3DMark03, the Radeon 9600 XT leads the way in all but the DirectX 7-class Wings of Fury game test. The 9600 XT’s performance advantage over the 9600 Pro is particularly impressive in the complex Mother Nature test.
Tomb Raider: Angel of Darkness
Tomb Raider: Angel of Darkness gets its own special little intro here because its publisher, EIDOS Interactive, has released a statement claiming that the V49 patch, which includes a performance benchmark, was never intended for public release. Too late, the patch is already public. We’ve used these extreme quality settings from Beyond3D to give the GeForce FX 5600 Ultra a thorough workout in this DirectX 9 game, and the results speak for themselves.
The Radeons wipe the floor with the GeForce FX 5600 Ultra. To be fair, NVIDIA claims that the Tomb Raider benchmark incorrectly uses a generic DirectX 9 code path rather than NVIDIA’s GeForce FX-optimized code path for the game. If that is indeed the case, our results highlight just how much the GeForce FX line needs optimized code paths in order to be competitive.
The Radeon 9600 XT is out ahead in AquaMark3, which makes extensive use of DirectX 9 shaders.
I used the “-use20” switch with the Halo benchmark to force the game to use version 2.0 pixel shaders.
The Halo benchmark also shows the Radeons way out ahead, with the 9600 XT leading the way. Halo’s benchmark timedemo renders cut-scene footage, which is important to note since the benchmark results don’t necessarily reflect real-world gameplay performance. However, the timedemo still uses Halo’s graphics engine, so it’s fair game for comparing the relative performance of different graphics cards.
Halo also isn’t compatible with the GeForce FX’s antialiasing, but ATI’s Smoothvision antialiasing seems to work just fine.
Real-Time High-Dynamic Range Image-Based Lighting
To test the GeForce FX 5600 Ultra’s performance with high-dynamic-range lighting, we logged frame rates via FRAPS in this technology demo at its default settings. The demo uses high-precision texture formats and version 2.0 pixel shaders to produce high-dynamic-range lighting, depth of field, motion blur, and glare, among other effects.
The Radeons XT looks great running this demo, and performance isn’t too shabby, either. The XT isn’t leaps and bounds ahead of the Radeon 9600 Pro, but there’s still a noticeable performance gap between the two cards.
The Radeon 9600 XT’s 100MHz core clock speed advantage helps a little with antialiasing, but the performance boost isn’t too dramatic.
The Radeon 9600 XT is just a little bit faster with anisotropic filtering, too.
In testing, I was able to get my Radeon 9600 XT stable and artifact-free with core and memory clock speeds of 595 and 640MHz. Try as I might, I just couldn’t get the card to behave with a 600MHz core clock speed. Overall, though, I’m quite happy with the 95MHz overclock; it certainly proves that the card has plenty of headroom for OVERDRIVE to work with.
Of course, just because I was able to get my 9600 XT stable at 595/640 doesn’t mean that every card will be stable at those speeds. Some cards may be cable of higher clock speeds, and some may barely overclock at all. Either way, I have to give ATI props for the Catalyst 3.8 driver’s new VPU recovery feature. This safety feature, added in the Catalyst 3.8 driver release, attempts to reset the graphics card if it locks up, potentially saving a reboot. It works, too. VPU recover is incredibly useful for overclocking stability testing, and I can’t wait to see it become a Windows requirement.
Overclocking the Radeon 9600 XT yields impressive gains in AquaMark3 both with and without antialiasing and anisotropic filtering. It’s official: low-k dielectrics rule.
The Radeon 9600 XT’s basic rendering capabilities are unchanged from its predecessor, so we haven’t included an image quality analysis in this review. However, things do change with driver releases, so we will be looking at the comparative 3D image quality of the latest ATI and NVIDIA products with the latest drivers in an upcoming article. For now, I can say that the Radeon 9600 XT’s image quality is, subjectively, very good.
The results of our performance testing are clear. The Radeon 9600 XT improves on the 9600 Pro’s already impressive performance in a wide variety of benchmarks and real-world applications, including titles that take advantage of DirectX 9-class pixel and vertex shaders and floating point data types. NVIDIA’s GeForce FX 5600 Ultra manages to sneak in a few small victories here and there, but those victories are mostly confined to dated games whose graphics engines don’t make use of DirectX 9-class hardware.
Though the Radeon 9600 XT performs well today, OVERDRIVE support should make the card even more fearsome. I can only hope that ATI will consider bumping up the Catalyst 3.9 driver’s top OVERDRIVE clock speed from 527MHz to something a little bolder. If my sample (which reached nearly 600MHz) is any indication of what the majority of RV360 GPUs is capable of, 527MHz may only scratch the surface of the chip’s overclocking potential.
At the end of the day, I think I’m more excited about the Radeon 9600 XT than I’ve been about any other graphics card. ATI has leveraged some very cool manufacturing technology to produce a graphics chip that’s capable of incredibly high clock speeds without the need for exotic cooling or even an auxiliary power source. The Radeon 9600 XT’s performance in DirectX 9-class applications is great, which bodes well for next-generation titles like Half-Life 2 and Doom 3. What’s probably most appealing about the 9600 XT is the fact that its $199 price tag makes the card very affordable for a wide range of gamers and PC enthusiasts alike. That $199 price tag includes a copy of Half-Life 2, too. All things considered, it would almost be irresponsible for me to not give the Radeon 9600 XT our coveted Editor’s Choice award for mid-range PC graphics. Today, the Radeon 9600 XT is as good as mid-range graphics gets.