From what I remember their cards were just superior in both graphics power and power usage. 3dfx kind of dropped the ball and went a bad direction (ala, Intel with the Itanium) and Nvidia capitalized on it. ATI couldn't quite keep up and I believe they still can't to this day.
They pioneered the regular release cycle. Their chief competitor, 3Dfx, had feature-based product releases. When a feature like integrated transform and lighting became difficult to implement, their product got delayed. The GeForce, on the other hand, never missed a release date. Even if it was just a faster clock cycle tweak of the last product, something ALWAYS got released on the relevant conference date. People switched while they were waiting for 3Dfx’s Voodoo 4 and Voodoo 5 vaporware, and devs switched to OpenGL because Glide wasn’t supported by NVIDIA (Direct3D came later). OpenGL had poor performance on 3Dfx hardware. Within a year or two of delays, they just were no longer relevant.
Interesting! I remember that the field was very crowded in those days. It wasn't just Nvidia but many other companies trying to compete with 3Dfx. All of those companies have fallen by the wayside now.
What's surprising is that copying a product cadence seems like an easy thing to do. Any of their competitors could have made minor improvements / bug fixes to their products and released a new card at every conference as well, so why didn't anyone else do it?
Not really fallen by the wayside. PowerVR and ATI were the main competitors, and they very much still exist today (PowerVR powers arm chips, and ATI merged with AMD).
There was a tile-based rendering chip startup that was also making waves around that time. I forget their name. IIRC they got bought out for their patents, and their architecture is very similar to what the Apple Silicon gpu does.
The first video card with a programmable pixel shader was the Nvidia GeForce 3 (NV20)
---
These early efforts to use GPUs as general-purpose processors required reformulating computational problems in terms of graphics primitives, as supported by the two major APIs for graphics processors, OpenGL and DirectX. This cumbersome translation was obviated by the advent of general-purpose programming languages and APIs such as Sh/RapidMind, Brook and Accelerator.[8][9]
These were followed by Nvidia's CUDA, which allowed programmers to ignore the underlying graphical concepts in favor of more common high-performance computing concepts.[
6 comments
[ 3.4 ms ] story [ 23.1 ms ] threadWhat's surprising is that copying a product cadence seems like an easy thing to do. Any of their competitors could have made minor improvements / bug fixes to their products and released a new card at every conference as well, so why didn't anyone else do it?
There was a tile-based rendering chip startup that was also making waves around that time. I forget their name. IIRC they got bought out for their patents, and their architecture is very similar to what the Apple Silicon gpu does.
---
These early efforts to use GPUs as general-purpose processors required reformulating computational problems in terms of graphics primitives, as supported by the two major APIs for graphics processors, OpenGL and DirectX. This cumbersome translation was obviated by the advent of general-purpose programming languages and APIs such as Sh/RapidMind, Brook and Accelerator.[8][9]
These were followed by Nvidia's CUDA, which allowed programmers to ignore the underlying graphical concepts in favor of more common high-performance computing concepts.[