-
Book Overview & Buying
-
Table Of Contents
Modern Computer Architecture and Organization - Third Edition
By :
This chapter examined the computational demands of video scene generation and the expanding role of GPUs in machine learning and AI inference. It included a detailed review of the architecture of a current-generation GPU, the NVIDIA RTX 3090. The analysis focused on the structure of the GA102 processor, which performs highly parallelized computations for high-resolution video rendering, big data analysis, and AI applications such as neural network training and LLM inference.
We examined the GA102 in detail, starting at the chip's interfaces to other components, including the GDDR6X memory controllers and the computer's central processor. We examined the internal architecture of the GA102, beginning with the GPCs, the SMs within the GPCs, and finally, the Warp processors, CUDA cores, and ray tracing cores within each SM.
We also learned about tensors and how the tensor core within each SM partition accelerates matrix operations for video scene generation, big data analysis...