Joint Training on AMD and NVIDIA GPUs

📅 2026-02-20
📈 Citations: 0
Influential: 0
📄 PDF

career value

237K/year
🤖 AI Summary
The surging computational demands of large language model (LLM) training exceed the capacity of homogeneous clusters from a single vendor, necessitating heterogeneous collaborative training across AMD and NVIDIA GPUs. This work proposes a compatibility-oriented hybrid communication architecture that integrates CPU-mediated forwarding and direct device-to-device connectivity, and introduces an innovative CPU-offloaded peer-to-peer transfer mechanism that enables cross-vendor GPU data exchange without host memory staging for the first time. By leveraging multi-NIC parallel transmission and differentiated communication backend scheduling, the system achieves 98% of the throughput of an NVIDIA-only homogeneous cluster on LLaMA-8B and Qwen2-7B models while preserving training stability and correctness.

Technology Category

Application Category

📝 Abstract
As large language models continue to scale, training demands on compute and system capacity grow rapidly, making single-vendor homogeneous clusters insufficient. This paper presents a technical solution for heterogeneous mixed training in AMD-NVIDIA environments. We first adopt a compatibility-oriented approach based on CPU-Forwarding Communication, with differentiated communication back-end selection across parallel groups and multi-NIC parallel data transfer. To achieve higher performance, we further propose another Device-Direct Communication approach, integrating a CPU-offloading P2P mechanism to enable direct cross-vendor GPU data transfer without host-memory staging. Experiments on LLaMA-8B and Qwen2-7B demonstrate that the proposed Device-Direct Communication approach achieves up to 98% of the throughput of an NVIDIA homogeneous system, while preserving training stability and correctness.
Problem

Research questions and friction points this paper is trying to address.

heterogeneous training
AMD-NVIDIA GPUs
large language models
cross-vendor communication
distributed training
Innovation

Methods, ideas, or system contributions that make the work stand out.

heterogeneous training
cross-vendor GPU communication
Device-Direct Communication
CPU-offloading P2P
multi-NIC parallel transfer
J
Jon Hu
Zettabyte AI, Inc.
T
Thomas Jia
Zettabyte AI, Inc.
J
Jing Zhu
Zettabyte AI, Inc.
Z
Zhendong Yu
Zettabyte AI, Inc.