Gainward Golden Sample GF3

YOU WILL SEE MORE hardware reviews on the INQUIRER over the next few months, but we've also struck - think the IT phrase is a "strategic relationship" - with our friends at Hexus, where we point you to sections of reviews posted at its site. Here's the first.. µ

Introduction
NVIDIA's GeForce3 chip and cards based around it are now well established. Cost is still high compared to other competing cards, but despite such a limitation, the gaming and performance graphics world have taken to the GeForce3 in a big way.

There is no doubting its performance, that's for sure. The GeForce3 GPU is a 256-bit processor with 57 million transistors making it more complex than a Pentium 4. 128-bit memory interface to extremely fast DDR memory. Memory bandwidth of 7.36Gb per second at the default memory speed.

Then there's the programmable aspect. NVIDIA lets the programmer directly program the graphics pipeline of the GeForce3 to offer potentially unlimited effects. Called the nFiniteFX engine, it provides the support for DirectX 8's pixel and vertex shaders and lets OpenGL access the functionality via extensions.

The programmer has power to control the GPU and program his or her own effects into the graphics pipeline and isn't stuck with the effects the chosen API gives them. This opens up a whole new world of graphics programming and we'll see this programmability available in all new consumer parts in the future.

Next up is their Lightspeed Memory Architecture. Featuring in the new Xbox as well as the GeForce3, LMA is NVIDIA's attempt at solving the memory bandwidth problem facing graphics cards, much like ATi with the Radeon and HyperZ. Current graphics GPU's are incredibly fast. They can process massive amounts of data but current memory technologies still struggles to keep up. Increasing the memory clock speed increases the memory bandwidth. This helps a lot with current GPU's, something which Gainward sells this GeForce on the back of.

LMA decreases the amount of work the AGP bus has to do when shifting vertex and pixel data from main memory to the card memory. It attempts to do intelligent memory access on chip, reducing the strain on AGP. It's a combination of a new memory crossbar between the GPU and card memory, along with some clever tricks when working with data that achieve this problem and help keep memory bandwidth issues to a minimum. The 7.36Gb per second limit still exists, just the GPU does a lot more with it and is more efficient.

Lastly on NVIDIA's list of top features is FSAA (full scene anti aliasing) or HRAA as NVIDIA like to call it (high resolution anti aliasing). It is basically a way of quickly multisampling a scene, combining those samples and then displaying them in a way which eliminates pixel artefacts such as jagged edges.

All those features and more, combined give the GeForce3 the current edge over the competition. ATi's new r200 core is hot on the heels of the GeForce3 and offers performance improvements. I'm sure NVIDIA have a trick or two up their sleeves to keep the GeForce3 at the top, plus the GeForce3 Ultra is on the horizon.

Go to HEXUS.NET for the rest of the review including benchmarks and pictures.