You all know and have been reading about R300, that was the codename for the Radeon 9700 Graphics core. What's in that name ? You likely also noticed the addition of 'Pro' on the videocard's name and yes there will be a 'normal' Radeon 9700 Graphics card also. It's exactly the same stuff yet lower clocked core and memory. Right now this Pro model is clocked at 325 MHz and 620 MHz for it's DDR memory. Next to pounding the NVIDIA GeForce4 in the face this card is the first on the market to have full support for DirectX 9.

The Radeon 9700 graphics core is based on a 0,15 micron fabrication. Quite daring as to achieve all that raw power you must get a lot of transistors into that graphics core. Not less than 110 million transistors are working hard on that silicon. So much transistors translate into a issues as it's large in size and thus yields will be less. The Radeon 9700 has over 1,000 pins due to a 256-bit memory bus and a lot of power delivery pins. When you would look at the core it would actually remind you of the Pentium III design, it's that big in size. As I mentioned, yields are another issue when you build a silicon with that many transistors. 110 million transistors should make the card faster then NVIDIA's GeForce4 and to accomplish that it should at least match the clock frequency of 300 MHz with the means to go higher. Somehow ATi managed that though, the silicon design seems to work out just fine.

Who would have thought ATi would be second with AGP8x, they got slapped in the face by SiS who was the first to manufacture the AGP8x Xabre. Hilariously enough SiS made a board that hardly even needs AGP4x so 8x is totally useless on their cards. It's nice marketing to have though ;)

For the ATi Radeon 9700 Pro this is a somewhat different ballgame though. AGP8x delivers a total of 2.1 GB/sec bandwidth between the Radeon 9700 and your mainboard's North bridge. Even here I doubt a little if we'll notice a big difference between AGP8x and 4x though. So if you have an AGP4x mainboard, don't worry .. the card is backwards compatible and I believe you wouldn't even notice it. All future mainboards and graphics cards will move to AGP8x though.

The Radeon 9700 meets the DirectX9 standard of 2.0 Vertex Shader which means flow control (loops, jumps and subroutines) and more instructions that can be executed per clock cycle. The new Vertex Shader allows up-to 1024 instructions per clock cycle.

Another item in the r300 design was eight 128-bit floating point pixel rendering pipelines. As comparison the Radeon 8500 and the GeForce4 make use of four 64-bit pipelines. This also explains the need for the huge transistor count. Again this is DirectX9 specification. The big advantage here translates itself into a very high fillrate and gives it a lead over the competition. Of course there is support for Pixel Shaders also and it's a version 2.0, again this is part of the DirectX9 specification. There are several changes/additions in the specification, however not very interesting stuff for the common readers. Basically things like instructions, data type, data precision and render targets have increased or improved.

Memory interface - The Radeon 9700 Pro incorporates a new high-performance 256-bit DDR memory interface, capable of providing over 20 GB/sec of graphics memory bandwidth. It includes four independent 64-bit memory channels, each of which can be simultaneously writing data to memory, or reading data back into the graphics processor. Sophisticated sequencer logic ensures that all four channels are being utilized for maximum efficiency.