TensorWave

About

TensorWave provides a dedicated, high-performance cloud infrastructure specializing in AMD Instinct™ GPUs, including the MI300X, for demanding AI and High-Performance Computing (HPC) workloads. Their platform is designed to accelerate large language model (LLM) training and inference, demonstrated by successful fine-tuning of a 405 billion parameter model on a single 8-GPU node with 192GB of VRAM per GPU. Notably, TensorWave simplifies the deployment of AMD’s ROCm software stack – reportedly loading in 1.5 minutes compared to 20 minutes on comparable NVIDIA systems – and has been utilized by companies like Kamiwaza to showcase enterprise GenAI platforms.

Focus Areas

Technology Focus

Key People 3

• D Darrick Horton
• J Jeff Tatarchuk
• P Piotr Tomasik