Date:
Sunday , May 11, 2003
Author:
Brent Justice
Editor:
Kyle Bennett
Google +1

GeForceFX 5900 Ultra / Radeon 9800 Pro 256MB

One article, two new video cards. NVIDIA's GeForceFX 5900 Ultra. ATI's Radeon 9800 Pro 256MB. Head to head IQ and Performance!

continued...

NV35:

Welcome to the NV35, or rather GeForceFX 5900 Ultra, the successor to the NV30/5800. With a new cycle of graphics cards, we usually have product refreshes and totally new architecture launches. The NV35 fits in neither category in our opinion, but rather falls somewhere in the middle. Let's see if NVIDIA's NV35 will really “Ignite Your Graphics”.

Article Image Article Image Article Image Article Image

The NV35 is built on the .13 micron process just like the NV30 and NV31. We do get a few new buzzwords with this release, CineFX 2.0 and Intellisample HCT, which we will go over a bit later. There are a few big improvements for the NV35 over the NV30. The hardest hitting change is the use of a 256-bit memory bus. You will recall the ATI Radeon 9700/Pro and 9800/Pro already use a 256-bit memory bus. Now the NV35 utilizes that bus width as well, allowing for much more bandwidth when compared to the NV30. Another addition is the use of 256MB total video card RAM on the high end GeForceFX 5900 Ultra board. This added memory will give you more room for framebuffer functions such as resolution and AA settings combined with texture storage.

The GeforceFX 5900 Ultra ships at an 850MHz DDR memory speed on a 256-bit memory bus, which produces a total of 27.2GB/sec of raw memory bandwidth. That is a 70% increase in raw memory bandwidth over the GeForceFX 5800 Ultra and 21% more raw bandwidth when compared to the ATI Radeon 9800 Pro 256MB card. However, this is still a 4 pixel pipeline card, as can be seen in the third slide. Two new features have been added, Ultrashadow and Intellisample HCT.

The other big upgrade for the NV35 is the doubling of its Floating Point Shader Ops. As you recall, the NV30 was much slower at pure Pixel shader speed when compared to the 9700/9800. NVIDIA has doubled the shader op performance to regain speed. This should certainly come into play with games we will see released over the next year and beyond.

CineFX 2.0

Article Image Article Image

CineFX 2.0 is the name given to the cinematic quality part of the NV35. This is where the 2x increase in shader ops comes into play as well as the Ultrashadow technology and 128-bit internal precision.

Ultrashadow is NVIDIA’s name for a new routine in the NV35 to help accelerate shadow volumes. As you know, shadow volumes will be a big part in Doom ]|[, and any card that wants to compete with the big boys will need to be able to handle these extensive stencil shadows very efficiently. The easiest way to think about Ultrashadow is simply to visualize that the card is not calculating shadows that cannot be seen by the camera. Meaning that shadows behind an object on your computer screen, or out of view of your character, are not given to the GPU to figure out. Therefore, the 5900 Ultra is not doing work that needs not be done, giving those saved GPU cycles for other priorities. This is from the official NVIDIA 5900 Handbook on Ultrashadow:

Shaders are also used to create sophisticated shadows for enhanced realism. The new NVIDIA UltraShadow™ technology accelerates the complex computations for lighting source and object interactions. For details on this patent-pending NVIDIA innovation, please refer to the NVIDIA GeForce FX: UltraShadow Technology paper at www.nvidia.com.

UltraShadow gives programmers the ability to calculate shadows much more quickly by eliminating unnecessary areas from consideration. With UltraShadow, programmers can define a bounded portion of the scene (often called depth bounds) that limits calculations of lighting source effects to objects within a specified area. (See Figure 2.) By limiting calculations to the area most affected by a light source, the overall shadow generation process can be greatly accelerated. Programmers can fine-tune shadows within critical regions, create incredible visualizations that effectively mimic reality, and still achieve awesome performance for fast-action games. The accelerated shadow generation can also free up time that can be allocated to other sophisticated but time-consuming effects.

Because stenciled shadow volumes require no texturing or color updates, the hardware “doubles up” the rendering horsepower to generate stenciled shadow volumes at speeds of up to double the standard pixel-processing rate. Other graphics solutions have to render stenciled shadow volumes in two passes. UltraShadow accomplishes the shadow volume rendering in a single pass, reducing CPU overhead and improving GPU performance. The NVIDIA approach also interoperates with NVIDIA Intellisample™ high-resolution compression technology (HCT) to make sure that shadow edges are properly antialiased. The GeForce FX 5900 GPUs maintain the stencil information on a sub-pixel basis, ensuring that shadow edges are antialiased rather than “blocky” or “jaggy.”

It is also important to note that UltraShadow will enhance application performance on any application that calculates shadows. As more passes are needed for lighting and shadows, you will see a greater performance increase when compared to not having this technology.