Club3D X850 XT - 02 - Features

Main Features
GPU	Radeon X850 XT - R480
Memory Brand/Model	Samsung K4J55323QF-GC16 (1.6ns)
Memory Type	256 MB 256-bit GDDR3
Engine Clock Speed	520MHz
Memory Clock Speed	540MHz
Memory Bandwidth	35.8 GB/sec
Pixel Pipelines	16
Vertex Pipelines	6
Fill rate	8.6 Giga pixels /sec
DX Support	9.0
OpenGL Support	1.5
Output	VGA / DVI-I / D-Sub
Bus	AGP 4X/8X
Processing technology	0.13 micron

The R480 is not very different from the previous high end R430. It too is built using the 0.13 micron low-k fabrication process and still features 16 Pixel Pipelines and 6 parallel Vertex engines.

As you can see from the ATITool screenshot, the GPU and Memory clocks for the X850 XT are set to 520MHz and 540MHz respectively.

The card is equipped with 8x32MB=256MB of GDDR3 memory from Samsung. The parts used are the famous K4J55323QF-GC16 which are widely popular, especially for high-end graphics cards.

The 16 in the memory's name corresponds to 1.6ns memory speed. Dividing 1 by 1.6, we get the maximum supported memory clock frequency. Therefore 1/1.6=600MHz which is quite a difference from the 540MHz. Hopefully, the fansink on the X850XT will provide the card with enough cooling to sustain increasing the clock closer to 600MHz.

Here's a table containing the differences between the X800 XT PE and the new X850 XT and XT PE:

	X800 XT PE	X850 XT	X850 XT PE
Core(MHz)	520	520	540
Memory(MHz)	560(1120)	540(1080)	590(1180)

 

X800=X850 Architecture

3Dc

To understand what 3Dc does, we need to quickly run through a brief discussion on normal maps.

You can think of a normal as an arrow pointing outwards from a surface at a 90 degree angle in order to quickly access its angle difference from different points of lighting.

First, a model with an extremely high polygon count is created (i.e 15000 polygons) and then one with less (1000 polygons) and that is the one that we'll eventually keep. Running a simple program calculates the differences between the two models and stores it as a normal map texture. Before rendering the final model in the game, the normal map texture is applied on the low polygon model and pixel shader instructions are used to compute real time lighting.

A quick example of Normal Mapping (not 3Dc!)

Normal mapping is greatly used in some of the most successful game titles of our days, like Far Cry, Half Life 2 and Doom 3. However, the problem with Normal mapping is the great limitation on the texture size due to the lack of compression.

To get a peek at what would happen if none of the textures were compressed in a recent game, try Doom 3 and set the details to Ultra High. As suggested by Id Software, Doom3's publisher, the Ultra High detail level requires a graphics card with 512Mb RAM. In case you're not really into the graphics card market there is no card with 512Mb RAM, yet.

Such is the case with normal mapping. The current texture compression algorithms, DXTC and S3TC, are inapplicable on normal mapping textures as they produce block artifacts, so programmers usually avoid using them.

3Dc is the solution to this problem with an algorithm that splits up the texture into blocks. Then compressing each block, it is able to achieve up to 4:1 compression.

When using 3Dc you can get much better quality textures with the same performance from your card.

The bad thing about 3Dc is that we'll have to wait until the game developers incorporate it into their engines or release a patch that takes advantage of it. Ubisoft was supposed to be the first to include 3Dc support into their much awaited Far Cry patch 1.3. However, as we found out, there is no actual difference in the game despite the addition of the console command "EnableCompressedMaps".

 

SMARTSHADER™ HD
• Full hardware acceleration for Microsoft® DirectX®
9.0 programmable vertex and pixel shaders in
hardware
DirectX 9.0 Vertex Shaders
- Vertex programs up to 65,280 instructions
with flow control
- Single cycle trigonometric operations
(SIN & COS)
DirectX 9.0 Extended Pixel Shaders
- Up to 1,536 instructions and 16 textures per
rendering pass
- 32 temporary and constant registers
- Facing register for two-sided lighting
- 128-bit, 64-bit & 32-bit per pixel floating point
color formats
- Multiple Render Target (MRT) support
• Complete feature set supported in OpenGL®
via extensions

SMOOTHVISION™ HD
• 2x/4x/6x Anti-Aliasing modes
Sparse multi-sample algorithm with
gamma correction, programmable
sample patterns, and centroid sampling
Lossless Color Compression (up to 6:1)
at all resolutions, including widescreen
HDTV resolutions
Temporal Anti-Aliasing
2x/4x/8x/16x Anisotropic Filtering modes
- Up to 128-tap texture filtering per
AA sample
- Adaptive anisotropic filtering algorithm
with bilinear (performance) and trilinear
(quality) options

HYPER Z™ HD
• 3-level Hierarchical Z-Buffer with Early Z Test
• Lossless Z-Buffer Compression (up to 48:1)
• Fast Z-Buffer Clear
• Z cache optimized for real-time shadow rendering
• Optimized for performance at high display
resolutions, including widescreen HDTV resolutions

VIDEOSHADER™ HD
• Seamless integration of pixel shaders and video in
real time
• FULLSTREAM™ video de-blocking technology for
Real, DivX, WMV9 and WMV10 formats
• VIDEOSOAP™ noise removal filtering for captured
video
• MPEG1/2/4 decode and encode acceleration
DXVA support
Hardware Motion Compensation, iDCT, DCT and
color space conversion
• All-format DTV/HDTV decoding
• Adaptive Per-Pixel De-Interlacing and Frame Rate
Conversion (temporal filtering)