The Graphics Chip
Before going on to look at the specific boards being tested here, lets take a closer look at the chip details.
In relation to the higher end R423, by the loss of the two fragment / pixel pipeline quads the RV410 chip has only shed around 40 Million transistors, but has gained a few on the Radeon 9800's R350 chip thanks to the extra vertex engines. Given that R350 was last years high end chip and, apart from the memory bus, RV410 is at least its equal, so something must have happened to for it to be able to be offered at less than half its initial price, and this is of course down to process. R350 was based on the 150nm process whilst RV410 uses TSMC's 110nm process, allowing ATI to fit more transistors on to a smaller die than R350.
The reduction in die size, in order to get a chip of such a configuration down to the mainstream price points, is also probably the primary reason that a 256-bit memory bus is not utilised. Whilst transistor sizes scale down with new processes, the pads (chip to packaging interconnects) do not similarly scale down, hence 256-bit buses actually require a large die size in order to facilitate all the connections and the die size of RV410 may not be large enough to facilitate such a bus.
TSMC's 110nm process is an optical shrink of 130nm, which allows it further die reductions from 130nm, however it is currently targeted purely as a "value" process and hence is not offered with a low-k option. Given that ATI's high end R420/R423 both achieve high clock speeds with the use of low-k you may think that RV410 will be a little constrained as far as peak clocks are concerned, and while that is the case to some extent, the XT version is still hitting a 475MHz release speed.
The following is a list of the significant features that the RV410 chip makes available to end users:
- SMARTSHADER HD
- Support of DirectX9 Programmable Vertex and Pixel Shaders
- VS2.0 Vertex Shader functionality
- Up to 65,280 instructions including loops and subroutines.
- Single Cycle Trigonometric Operations (SIN & COS)
- DirextX9 Extended Pixel Shaders
- Up to 1,536 instructions and 16 textures per rendering pass
- 2nd Generations F-Buffer support for unlimited Shader instruction lengths
- 32 temporary and constant registers
- Facing register for two-sided lighting
- Multiple render target support
- Shadow volume rendering acceleration
- 128-bit, 64-bit & 32-bit per pixel floating point colour formats
- SMOOTHVISION HD
- 3Dc Normal Map Compression
- High quality 4:1 Normal Map Compression
- Works with any two-channel data format
- 2x/4x/6x Multi-Sampling full scene Anti-Aliasing modes, adaptive algorithm with programmable sample patterns and colour buffer compression
- Temporal Anti-Aliasing
- Lossless Color Compression (up to 6:1)at all resolutions, including widescreen HDTV resolutions
- 2x/4x/8x/16x anisotropic filtering modes
- 3Dc Normal Map Compression
- 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 Optimisations for shadow rendering
- Optimized for performance at high display resolutions, including widescreen HDTV resolutions
- VIDEOSHADER HD
- Seamless integration of pixel shaders with video FULLSTREAM video de-blocking technology
- Noise removal filtering for captured video
- MPEG-2 decoding with motion compensation, iDCT and colour space conversion
- All-format DTV/HDTV decoding
- YPrPb component output
- Adaptive de-interlacing and frame rate conversion
As our ATI Radeon X800 review details, the features of R420 are similar to the feature-set brought with R300, but with new instructions in the Vertex Shader, increased instruction lengths for the Pixel Shader, the addition of 3Dc normal map compression and architectural tweaks for better support of high resolutions. The X800 review also takes a close look at the overall X800 Pipeline from the Geometry Pipeline, to the Rendering Pipelines and finally down to the ROP's, much of which will be applicable to RV410.
