Compute Substrate for Software 2.0
Jasmina Vasiljevic, Ljubisa Bajic, Davor Capalija, Stanislav Sokorac, Dragoljub Ignjatovic, Lejla Bajic, Miloš Trajković, Ivan Hamer, Ivan Matosevic, Aleksandar Cejkov, Utku Aydonat, Tony Zhou, Syed Zohaib Gilani, Armond Paiva, Joseph Lin Chu, Djordje Maksimovic, Stephen Chin, Zahi Moudallal, Akhmed Rakhmati, Sean Nijjar, Almeet Bhullar, Boris Drazic, Charles Lee, James Sun, Kei-Ming Kwong, James Connolly, Miles Dooley, Hassan Farooq, Joy Yu Ting Chen, Matthew Walker, Keivan Dabiri, Kyle Mabee, Rakesh Shaji Lal, Namal Rajatheva, Renjith Retnamma, Shripad Karodi, Daniel Rosen, Emilio Ángel Villanueva Muñoz, Andrew Lewycky, Aleksandar Knežević, Raymond Kim, Allan Rui, Alexander Drouillard, D. A. Thompson
Abstract
The rapidly growing compute demands of AI necessitate the creation of new computing architectures and approaches. Tenstorrent designed its architecture (embodied in Grayskull and Wormhole devices) to tackle this challenge via two fundamental and synergistic approaches. The first is via compute-on-packets fabric that is built from ground up for massive scaleout. The second is the ability to execute dynamic computation, built into the compiler, runtime software and hardware architecture. By combining these approaches, TensTorrent will enable continued scaling of AI workloads.