At Nvidia's GTC event, CEO Jensen Huang introduced the Vera Rubin platform, featuring their first custom-designed CPU 'Vera' and new 'Rubin' GPUs. These innovations are engineered to enhance AI inference performance, which is deemed critical as the demand for real-time data processing surges. Huang pointedly criticized past assumptions about AI model training, notably the DeepSeek phenomenon, asserting that computational needs for inference are now underestimated by a factor of 100, signaling a transformative shift in AI's computational landscape.
Huang emphasized that the computational requirements for AI reasoning have increased dramatically, claiming they are now 100 times what was previously anticipated.
The introduction of the Vera Rubin platform represents a significant leap in AI technology, focusing on enhancing performance during inference by processing vast data swiftly.
Collection
[
|
...
]