ALCF AI Testbed: Tenstorrent Training Workshop

Wednesday May 27, 2026, 9:00 AM → 4:00 PM America/Chicago

D173-T (B241)

Murali Emani (LCF), Varuni Katti Sastry (LCF), Venkatram Vishwanath (LCF)

This training workshop will introduce users to the Tenstorrent (TT) Galaxy systems deployed at the ALCF AI Testbed. This session will introduce users to TT Wormhole and Blackhole architectures and software stack to run AI and HPC applications.

Our current deployment includes two Galaxy Wormhole Servers each equipped with 32 Tenstorrent Wormhole processors, offering a massive pool of interconnected Tensix cores. One server (tt-01) is dedicated for AI inference workloads and another (tt-02) for TT-Mettalium, a low-level programming SDK for bare metal programming.

Agenda (in Central time):

Introduction to Tenstorrent	9.00 - 9:30
Hardware deep dive (Wormwhole and Blackhole)	9:30 - 10:30
Coffee break	10:30 - 10:45
Software deep dive Introduction to SW stacks Model bring-up	10:45 - 11:45
Debug and Visualization tools Performance Optimization tools: tracy, ttnn-visualizer Telemetry tools	11:45 - 12:15
Lunch	12:15 - 1:00
Inference server deployment TT-inference server introduction TT-console demo TT-home demo High-resolution on Galaxy demo	1:00-1:30
Scale out	1:30- 2:00
Matmul TT-metal kernel concepts in C++	2:00-3:00
Coffee break	3:00 - 3:15
Hands on + discussions TT-lang TT-Inference server running example	3:15 - 4:30

 

Please note that you will need a CELS account to access this system. To access the Tenstorrent Galaxy nodes, you will need to join the Tenstorrent Galaxy CELS project (ttgalaxy) https://accounts.cels.anl.gov.

 

To help you get up and running with the new hardware, please refer to the documentation at https://github.com/argonne-lcf/user-guides/tree/feature/Tenstorrent_docs/docs/ai-testbed/tenstorrent.