Production of the Annihilator, which uses Single Data Rate (SDR) memory, slowed once the faster DDR (Double Data Rate) memory began to appear in abundant quantities. To ensure that a supply of DDR memory would be available to graphics card manufacturers, Hyundai Electronics closed a deal in November of 1999 to supply one million DDR synchronous DRAM chips to NVIDIA. A second deal is expected next year which will supply 1.5 million units of the chips per month to NVIDIA.
Before discusssing the difference between the Annihilator and Annihilator Pro, let's cover a few basics of NVIDIA's GeForce 256. This information is covered in detail in our GeForce 256 review. For a broad overview, I recommend reading the GPU Overview page.
The Graphics Processing Unit
The Annihilator and Annihilator Pro contain a Graphics Processing Unit, or GPU, which was designed by NVIDIA. The GPU functions as a specialized processor and its purpose is to accelerate transform and lighting calculations and does so at incredibly fast speeds.
While a PCs processor can execute these specialized calculations rather well, its performance is no match for dedicated silicon chips such as a GPU. In most cases, a GPU with an integrated transform and lighting engine can perform certain calculations 2 to 4 times faster than today's leading CPU's.
An additional benefit of the GPU is that is reduces the work load on a central processor. In the past, the CPU performed transform and lighting calculations, but now those chores can be offloaded to the GPU. There are applications that immediately benefit from the use of transform and lighing which include games developed in OpenGL, such as Quake and it's derivatives, and Computer Aided Design. A major feature of Microsoft's DirectX 7 application program interface (API) was the addition of support for transform and lighting.
The GeForce GPU Frees CPU Cycles
The bottom line is that game developers can increase the realism of their games, by adding greater levels of detail for objects and real-time lighting, knowing that a dedicated GPU is available to back their enhancements up.
SDR vs DDR Memory
I didn't spend much effort in this section since NVIDIA has already written a DDR Memory Technical Briefing. Put simply, DDR (Double Data Rate) is a new memory technology that improves graphics performance. Theoretically, DDR memory offers almost twice the peak memory bandwidth (4.8GB/s) over Single Data Rate (SDR) memory (2.6GB/s) which is used on the Annihilator (GeForce SDR).
Memory bandwidth refers to the speed at which memory exchanges data with the graphics processor and is measured in gigabytes per second. For example, the following chart illustrates the memory bandwidth requirements in a simple 3D application:
As a rule of thumb, the greater the graphics complexity is, the greater the memory bandwidth is, especially in 32-bit color.
Next Page: Installation