GeForce3 Up Close & Personal
We've already discussed the technical aspects of the GeForce3 chip in detail in High-Tech And Vertex Juggling-NVIDIA's New GeForce3 GPU . You can read up on the first benchmarks in the GeForce3 Performance article. Above and beyond the high performance, the chip also offers a broad array of features. At present, quite a few of them are dormant, since the current generation of games doesn't support all of the new features. Others, however, can already enhance your 3D-gaming experience. In the upcoming 8.1 upgrade of DirectX there will be significant improvements for pixel-shading so, all the more advanced features of the GeForce3 should eventually yield better 3D experiences.
Here is a list of the most important features:
- A total of 57 million transistors on the GeForce3 is more than double what its predecessor, the GeForce2, offers. This steps up the performance from 46 to 76 Gigaflops. The manufacturing process was also optimized thanks to the introduction of the 0.15 micron process . The die and power consumption were hardly affected. NVIDIA sets the clock speed of the chip at 200 MHz. The RAM clock operates physically at 230 MHz for DDR SDRAM, which NVIDIA and the manufacturers market as "460 MHz" due to double data rate.
- Vertex Shader
The GeForce3 is not only furnished with a 'static' Transform & Lighting Engine, which was also a feature of its predecessors GeForce 256 and GeForce2. Its functionality is improved by a programmable vertex processor, which NVIDIA dubs Vertex Shader. - Pixel Shader
Supplementary to the Vertex Shader, NVIDIA adds a texture combiner, in order to make texture operations programmable and thus more flexible. This module works hand-in-hand with the Vertex Shader, expanding the range of the GeForce3's new potential effects. - LightSpeed Memory Architecture
The GeForce3 uses a form of memory management that is a novelty among 3D cards. The crossbar memory controller makes up for problems caused by limited memory bandwidth - it uses the bandwidth provided by the onboard memory of the graphics card more efficiently than previous 3D designs. This allows the GeForce3 to obtain frame rates that are significantly higher than NVIDIA's older boards. The new chip still has the same memory bandwidth and, in theory, even higher fill rates. The GeForce3 comes furnished with an additional feature that aims to omit hidden objects from its calculations - the user doesn't see them anyway. This function, called Z Occlusion Culling, helps to further increase effectiveness. - Multi-Sampling Full-Scene Anti-Aliasing
NVIDIA's developers were painfully aware that full-scene anti-aliasing was a main cause of massive slumps in performance. As a solution to this, a compromise was worked out and titled Quincunx . This special version of multi-sampling anti-aliasing permits the GeForce3 to use FSAA at the high-res setting of 1024 x 768 without frame rates dipping below an acceptable level. Here are two example images with and without FSAA (each 1.1 MB in size).
There's a budget GeForce GPU selling in China that not even Nvidia knew it made — RTX 4010 turns out to be a modified RTX A400 workstation GPU
US to patch loopholes that allow China to buy banned AI GPUs from other countries — new regulations include national quotas on GPU exports and a global licensing system
