Image
SystemX icon

GPU is what happens to AI while you’re busy making other plans

Summary
Raja Koduri (Mihira)
Gates B12
Jan
25
Date(s)
Content

About the talk: A walk through the brief history of GPU architecture from 1990’s to now. The evolution from a display frame buffer, 2D drawing, 3d rasterization, programmable pipelines to massively parallel AI supercomputer. Contrasting computing approaches between CPUs, GPUs and TPUs with transportation analogies. What made the GPU architecture so resilient and what we can learn from that. Ending with thoughts on possible directions for future architectures.

About the speaker: Raja M. Koduri is the Founder, Chairman, and CEO of Mihira. He also serves on the Board of Directors for Tenstorrent Inc, a pioneer in Risc-V and AI silicon. In his previous roles at Intel Corporation, he held key positions such as the Executive Vice President and Chief Architect, focusing on driving Intel's High-performance computing and GPU efforts. Raja's extensive experience includes being the General Manager of Radeon Technology Group at AMD and serving as the Director of GPU Architecture at Apple, where he made significant contributions to Apple's silicon roadmap. His reputation for advancing visual and accelerated computing across various platforms is notable, and he played a vital role in advancing visual effects technology in India during his time with Makuta VFX Studio. Raja Koduri holds a master's degree in electronics and communications from the Indian Institute of Technology Kharagpur.