Nvidia, STM try consumer graphics ICs again

Electronic News, April 21, 1997 by Andrew Maclellan

Code-named the NV3, the device is a successor to Nvidia's NV1 chip and the second-generation NV2, the latter of which the companies shelved last summer after receiving disappointing feedback from OEMs. In addition to 2-D/3-D acceleration, the new device includes support for a variety of video processing functions such as DVD playback, TV-out, video phone and Intercast.

The RIVA 128 touts 20 billion operations-per-second (GOPS) dedicated to 3-D graphics, including a 5 GFLOPS floating-point set-up engine to offload 3-D processing from the CPU and what the company claims is a massively parallel pixel processor capable of sustaining 15 billion pixel operations-per-second.

The result, according to Nvidia, is a 128-bit engine capable of processing up to 5 million triangles-per-second for frame rates approaching 100 frames-per-second and a more than 100 million pixel-per-second fill rate for rendering real-time scenes. Incorporating a 100MHz synchronous graphics RAM (SGRAM), the RIVA 128 is said to achieve a bandwidth of 1.6 gigabytes-per-second.

The 3.5-million transistor chip, manufactured on STM's 0.35-micron process technology, supports a flexible texture memory which can be stored on-chip, in the graphics frame buffer or in main memory with no latency and a claimed hit rate of better than 90 percent

Priced at $30 in 10,000-unit quantities, the chip is available this quarter in volume and is shipping to board makers Diamond Multimedia Systems and STB Systems. The part will be jointly marketed and sold by the two companies, with Nvidia taking on the U.S. and STM handling the European, Japanese and Asia-Pacific markets.

"As real-time 3-D modeling experts, we have worked with many implementations of 3-D graphics, from Onyx to Genesis, said Josh White, president of Vector Graphics. "RIVA 128's performance gives us more artistic freedom than we expect from a PC platform and standard 3-D API. Freedom from texture memory limits and complete acceleration of Direct3D allows us much higher face counts and texture detail."