Tasawwur D310
Project Overview
Challenges
Category
Disciplines
Year
About
Tasawwur D310, a transformative transistor-based Graphical Processing Unit (GPU), drastically curbs Artificial Intelligence (AI)'s environmental impact. As AI models like Generative Pre-trained Transformer 3 (GPT-3) grow, traditional 2D GPUs emit significant CO2: training ChatGPT-3 produces emissions comparable to 123 cars driven yearly. Tasawwur D310, using its advanced 3D circuit design, reduces power needs by over 250%. Training ChatGPT-3 on Tasawwur cuts CO2 emissions by 220 tons, equaling the benefits of 9 million newly planted trees annually. This innovation represents a leap towards sustainable AI computation.
Team
Saravanan Yuvaraja
King Abdullah University of Science and Technology (KAUST)
I'm an excited research scientist eager to introduce our groundbreaking project. We're on a thrilling mission to revolutionize artificial intelligence (AI) by making it more energy-efficient. Imagine AI chips so smart that they could significantly reduce power consumption, making them the top choice for enhancing AI models like ChatGPT and powering data centers. We're at the forefront of exploration, deep in the research phase. Our initial proof-of-concept chip is already changing the game with an impressive 250% reduction in power usage compared to regular chips. But that's just the start. Our ambitious goal? A breathtaking 1000-fold reduction, reshaping AI chips and opening up new possibilities. This journey is filled with excitement, and we'd love for you to be a part of it!
Gallery
