Introduction

ATI’s RADEON 9700 VPU was proud to be the first fully DirectX 9 compliant graphics core and it went on to be ‘the’ gaming card to own for more than half a year. With the heavily hyped NVIDIA GeForce FX 5800 Ultra ending up as a lackluster product, the RADEON 9700 effortlessly reigned as the best available and affordable (in contrast to the FX 5800 Ultra) high-end gaming card. While NVIDIA was nearing the period of launch of their GeForce FX 5900 Ultra, ATI revised the RADEON 9700’s R300 core and came out with the R350 core to better compete with NVIDIA’s new product.

Although numerically, it seems as if great improvements have been made over its predecessor, the R350 is mostly an update judging by the enhancements. ATI is also engaged in the model number game similar to NVIDIA and the R350 core is now named as the RADEON 9800 VPU. However, there is a problem with ATI’s model numbering scheme since the entire 9xxx series is dotted with the highest-end all the way down to the lowest end graphics cards. Due to this reason, the individual models in the entire series are not quite comparable or consistent in their graphics engine and supported features. If you were to peek in to the NVIDIA’s clearly defined series of GeForce4 MX, GeForce4 Ti and GeForce FX, the products within them are much better represented in their capabilities than if you were to look at the entire 9xxx series from ATI. From the amount of queries and uncertainties expressed by just looking at the discussions from our community, ATI has definitely done a great job in confusing users. Certainly more effort is needed to improve their product numbering/naming schemes with their next generation of products (certainly not now, as it would only add to the confusion further).


Putting that slight confusion aside, the RADEON 9800 VPU is still a 0.15-micron part and the GPU clock speed has now been raised from 325MHz to 380MHz for the ‘Pro’ version. This wasn’t much of a problem since a number of existing R300 cores could easily overclock higher than their stock speed. Faster memory chips too have been coupled to the RADEON 9800 Pro, raising its bandwidth to 21.8GB/s.

Within the core, SmartShader and SmoothVision have both been updated to version 2.1 while HyperZ III had been updated to HyperZ III+. The below is a summary of changes made to each feature:

SmartShader 2.1 – A new F-Buffer had been added to allow shader programs of unlimited length to be executed on the RADEON 9800’s VPU, essentially rivaling the CineFX Engine from NVIDIA in this aspect and a few others.

SmoothVision 2.1 – Better FSAA and AF performance due to optimized controller and Color compression of up to 6:1 ratios.

HyperZ III+ - Equipped an optimized Z-Buffer cache that is now tuned for real-time shadow rendering.

Apart from the above, the RADEON 9800 has a lot in similarity with its predecessor, RADEON 9700. Below is the comparison table between the RADEON 9700 Pro and RADEON 9800 Pro VPU, as well as the GeForce FX 5800 / 5900 Ultra:

Graphics Cores Comparison

GPU/VPU	ATI RADEON 9800 PRO	ATI RADEON 9700 PRO	NVIDIA GeForceFX 5900 Ultra	NVIDIA GeForceFX 5800 Ultra
Core Code	R350	R300	NV35	NV30
Manufacturing Process (microns)	0.15	0.15	0.13	0.13
Transistor Count	~110 Million	~110 Million	130 Million	125 Million
Core Clock	380MHz	325MHz	450MHz	500MHz
Rendering (Pixel) Pipelines	8	8	8 or (4)	8 or (4)
Peak Pixel Fill Rate (Mpixels/s)	3,040	2,600	3,600 or (1,800)	4,000 or (2,000)
Texture Units per Pipeline	1	1	1 or (2)	1 or (2)
Peak Texture Fill Rate (Mtexels/s)	3,040	2,600	3,600	4,000
Textures process-able per Texture Unit	8 to 16	8 to 16	8 to 16	8 to 16
Memory Clock	340MHz (680MHz DDR)	310MHz (620MHz DDR)	425MHz (850MHz DDR)	500MHz DDR II (1GHz DDR)
DDR Memory Bus	256-bit	256-bit	256-bit	128-bit
Memory Bandwidth	21.8GB/s	19.8GB/s	27.2GB/s	16.0GB/s
Core Features	SMARTSHADER 2.1 SMOOTHVISION 2.1 - Color Compression (up to 6:1 ratio HYPER Z III + : Z-Compression (up to 24:1 ratio) and Z-cache optimized for real-time shadow rendering TRUFORM 2.0	SMARTSHADER 2.0 SMOOTHVISION 2.0 HYPER Z III - Z-Compression (up to 24:1 ratio) TRUFORM 2.0	CineFX 2.0 Engine LMA II Intellisample HCT Technology - Color and Z-Compression (up to 4:1 ratio)	CineFX Engine LMA II (Optimized for DDR II) Intellisample Technology - Color and Z-Compression (up to 4:1 ratio)
Pixel Shader Support	2.0+	2.0	2.0+
Vertex Shader Support	2.0+	2.0	2.0+
DirectX Support	Up to DirectX 9.0		Up to DirectX 9.0
FSAA Modes	Multi Sampling (up to 6x)		Multi Sampling (up to 8x)
Anisotropic filtering Modes	Up to 16x		Up to 8x
Other Features	Dual integrated 10-bit per channel 400 MHz DACs Integrated TV Output support up to 1024x768 resolution Integrated 165 MHz TMDS transmitter (DVI 1.0 compliant) VideoShader HydraVision		Dual integrated 400 MHz DACs Integrated TV Output support up to 1024x768 resolution Video Mixing Renderer (VMR) Extended programmability with Cg nView DVC 3.0
AGP Speed	8x		8x
AGP Bus Compatibility	AGP 3.0, 2.0 and 1.0		AGP 3.0, 2.0 and 1.0