One of the most significant events of the past year in gaming hardware was the collapse of onetime market leader 3dfx and subsequent absorption by their rival NVIDIA. With the only other rival for the high-end gaming market seemingly eliminated, NVIDIA seemed set to rule the roost – but overall market leader ATI has made a strong push to fill the void left by 3dfx. ATI’s Radeon has remained an inexpensive and feature-packed alternative to NVIDIA’s cards, and now the next generation of ATI hardware has arrived – the 8500. ATI’s answer to the GeForce 3, this card promises a superior frame rate and feature set while maintaining the stellar image quality and integrated capabilities of their previous line. With a promised performance increase of 2 to 3 times the original Radeon’s speed and a fully buzzword-compliant feature set that embraces all of the pixel and vertex shader features, this is ATI’s bold move to claim a space in the high-end gaming market for the first time. And it will ship in four to six weeks.
It is important to note that the facts and figures presented here come straight from ATI, not from a third-party source, and many claims are currently unverified; it is also important to note that the card being described here is the PC model, though the Mac version of the ATI Radeon 8500 is expected to have an identical feature set.
Let’s get right to the numbers. The Radeon 8500 is a 2x/4X AGP, single-GPU card with 64 MB of 275 MHz DDR RAM. The card GPU itself is clocked at 250 MHz, as is the RAM. Using “a combination of brute force and intelligence,” according to our interview with ATI staff, they have managed to enhance every area of this card’s performance over the previous architecture.
The fundamental areas of enhancement over the previous Radeon models are as follows: the 8500 will feature a faster clock speed with dramatically increased memory bandwidth and a 2x speed increase overall according to Winbench benchmarks. The 8500 GPU is built with a 15-micron process and is more complex than a Pentium 3 and Pentium 4 chip combined – over 60 million transistors. This monster processor is coupled with Hyper-Z 2 technology and 64 MB of DDR RAM for a staggering – nay, unbelievable! – 12 GB per second memory bandwidth between the GPU and the DDR RAM.
The result? A claimed fill rate of one billion textured pixels per second. Yes, that is “billion” as in nine zeroes. That translates to 62.5 million triangles processed per second, well over twice what the already-impressive Radeon card could deliver.