
ATI introduces their mainstream video card series based on flagship X800 technology and we take a look at what level of gaming experience the Radeon X700XT is delivering directly compared to NVIDIA's GeForce 6600GT.

Move over X600XT, it is now time for the proper video card to take its throne in the mainstream market segment from ATI. NVIDIA filled their mainstream gap when they introduced the GeForce 6600 series in mid August. A few weeks ago we had the chance to test a GeForce 6600GT in real gameplay and found it to deliver incredible performance for the price point it is at. Unfortunately the only card we could compare it to that was comparable from ATI was the X600XT. It is now time for ATI to move their current generation hardware technology down into the mainstream market. The real battle has now begun.
Let’s cut straight to the info. ATI is debuting the X700 series video cards today which are priced for the mainstream video card market, read $150 to $250. ATI is introducing three video cards, the X700, the X700Pro and the X700XT. This time the “X” in the product name does mean that it is using the same technology that makes up the X800 series.
Unlike the X800 series (X800Pro and X800XT/PE) that uses a .13 micron process with Low-K, the X700 series is using the new .11 micron process fabricated at TSMC and it is not using Low-K. The X300 series is already using the .11 micron process, so ATI has some experience to draw from in using this process for the X700. The NVIDIA GeForce 6600 series is also using the .11 micron process. This new process allows lower power consumption and thus higher clocks. The X700 series is also built to support GDDR3 memory as is the X800 series. The GeForce 6600GT uses GDDR3 spec memories as well.
The X700 series are single slot video cards that require no external secondary power; they operate completely off of the PCI-Express bus power. ATI claims this is a low noise cooling solution, but compared to the GeForce 6600GT we thought otherwise. We will explain more about this later on.
ATI wants you to know that the X700 series is built for more than just DOOM 3. They go on to say that the X700 series will offer best in class performance in DX9 shader-heavy games.
Both the X700Pro and X700XT have 8 pixel pipelines with 6 vertex engines. This compares to the GeForce 6600 series which also have 8 pixel pipelines, but only 3 vertex engines. Right off the bat this gives the X700 series a theoretically higher geometry vertex/triangle performance advantage. Just like the X800 series, the X700 series supports 3Dc, SMARTSHADER HD, SMOOTHVISION HD, VIDEOSHADER HD, and HyperZ HD. The memory interface is 128-bits wide, supporting either 128MB or 256MB framebuffer with GDDR3 memory.
In the chip diagram above, you will notice that it looks exactly like the X800 layout except that this one is using 8 pixel pipelines, a 128-bit memory bus, and has the native PCI-Express interface module.
One very interesting specification about the X700 is the fact that it has 6 vertex engines. This is the same number that the X800Pro and X800XT-PE have. In comparison, the GeForce 6600 series has 3, while the GeForce 6800 has 5 and the GeForce 6800GT and Ultra have 6. ATI is correct when they say this is the first mainstream product with that much raw vertex processing power.
As we have come to learn recently, the entire line of Radeon R3xx/R4xx (Radeon 9500 and X800/X700 series) products all support Geometry Instancing. This was thought to be an NVIDIA only feature on the GeForce 6600 and 6800 series, but apparently NVIDIA is not the only one to have it. There are, however, some differences. NVIDIA supports the official Vertex Shader 3.0 version of Geometry Instancing as defined by Microsoft in DX9.0c. ATI does not. Their method of Geometry Instancing differs slightly from the official Microsoft specification in DirectX 9.0c. In order to use ATI’s Geometry Instancing, one must utilize the Shader Model 2.0b path. Recently ATI has had a problem with getting their method to pass in the latest WHQL driver tests for DX9.0c. Therefore, in order to pass WHQL, ATI has had to disable support for this feature by default in their drivers. However, in the latest driver which we have here, there is an option to enable Geometry Instancing.
Because this uses the X800 technology, the X700 series supports long pixel shaders of up to 1,536 instructions per pass. Utilizing the F-Buffer, the potential is infinite shader lengths because of faster multi-passing. The same sparse sampling AA with gamma correction is being used as seen in the X800 series. Temporal AA is also on track and ATI’s unique adaptive Anisotropic filtering is also present. ATI has made some changes in regards to texture filtering with a little thing they call CATALYST A.I. We will go into what this is on the next page; you don’t want to miss it.
3Dc support is also included in the X700 series. As you recall, 3Dc is a method to compress normal maps better than DXTC5 can. Now, 3Dc will improve image quality if it is being used in place of DXTC5, which will allow normal maps without the artifacts that DXTC5 has. But the main purpose of 3Dc is to bring forth better performance. Through that performance increase, content developers can choose to offset it by implementing higher resolution normal maps allowing a higher image quality. Otherwise if they do not do this all you will get is a performance increase using 3Dc. We have learned that FarCry patch 1.3 will use 3Dc. 3Dc in FarCry will be used only to increase performance, not increase image quality. We also know that Half Life 2 will be using 3Dc. If used, 3Dc does have potential, especially on these mainstream cards where every little bit helps, but if it isn’t used it is a useless feature. It appears that there will be games coming out using this new technology, which is good to hear.
ATI is still on top of their game with video capabilities. All the video capabilities of VideoShader HD in the X800 series are here and present on the X700 series. For the casual gamer who wants great game performance but also wants more out of his video card such as creating a home entertainment system, the X700 series is all ready to take that role.
Summing up the specifications here is what we are presented with. The X700XT is at the top of the mainstream lineup and will be clocked at 475MHz core speed and 525MHz (1.05GHz DDR) memory speed. This puts the fillrate at exactly 3.8 Gpixels/sec. Comparing that to the GeForce 6600GT’s core speed of 500MHz and 8 pixel pipelines we have 4 Gpixels/sec. So right off the bat the GeForce 6600GT has a higher pixel fillrate. The memory bandwidth on the X700XT is exactly 16.8GB/sec, this compares to the GeForce 6600GT’s memory bandwidth of 16GB/sec. So here the Radeon X700XT has a slight advantage. The X700XT also has a vertex advantage by having 6 vertex engines versus 3 on the GeForce 6600 series.
The Radeon X700Pro has the same features and is just clocked lower at 420MHz core speed and 432MHz (864MHz DDR) memory speed and comes with only 256MB of RAM, while the X700XT has 128MB of RAM, but is clocked much higher. The Radeon X700 will be clocked at 400MHz core speed and 350MHz (700MHz DDR) memory speed and comes with 128MB.
All in all, the X700 is an X800 built on a smaller process with half the pixel pipelines of their current flagship.