BullSequana AI 642-H03: your high-performance AI accelerator
The Eviden BullSequana AI 642-H03 is an AI-optimized compute node designed to empower industries, national HPC centers, government agencies, academia, and startups to run accelerated, low-latency AI inference and precision fine-tuning for mid-sized language models (~100M–10B parameters) and multimodal AI workloads.
This turnkey solution provides flexibility and efficiency for both rapid brownfield rollouts and greenfield deployments.
BullSequana AI 642-H03 at a glance
Performance
& speed
Supports up to 48 AMD Instinct™ MI355X GPUs per rack delivering 64 TB of global GPU bandwidth for high-speed communication and direct memory exchange.
Advanced GPU architecture
The GPUs supports a wide range of precision formats for training and inference and offer 2.3TB HBM3e with 288GB onboard memory per GPU.
Hybrid
cooling
Features hybrid cooling for improved energy efficiency and sustained performance, with up to 70% direct liquid cooling (DLC) and 30% air cooling via fans.
Comprehensive software infrastructure
AI developers and system administrators can leverage AMD ROCm 7 and Eviden’s HPC Suite for optimized model performance, unified management, and sustainable, cost-effective AI operations.
Rapid
deployment
With its fully pre-integrated, pre-tested, and configured design, the BullSequana AI 642-H03 server offers fast deployment, streamlined installation, and reduces space and costs.
Turnkey AI infrastructure for fast deployment, energy efficiency, and reduced total cost
In partnership with
Eviden is your trusted partner for end-to-end compute solutions
European sovereignty
Only supercomputer manufacturer in Europe
Local expertise, global footprint
Our experienced team serves customers around the world
End to End Compute
We deliver systems from design to deployment
Unified Expertise Advantage
Holistic solutions through R&D, delivery and partner technology
Related resources
BullSequana AI 642-H03
A high-performance AI accelerator designed for training AI models and performing inference with large models
High-Performance Artificial Intelligence
Discover our midrange Servers for High-Performance AI
Get in touch!
Connect with us to unlock the potential of High Performance AI.