Introducing SupremeRAID™ HE: High Availability NVMe for HPC and Beyond

Are you concerned with protecting mission-critical operations, preventing costly downtime or meeting regulatory and compliance requirements in your high-performance computing environment?

If so, then SupremeRAID™ HE, our powerful new HPC-focused addition to the SupremeRAID™ product family, is the solution you’re looking for. Engineered specifically for high-performance computing, AI factories, and enterprise workloads that demand both speed and resilience at scale, this GPU-accelerated high availability solution helps safeguard data, maintain regulatory standards, and keeps demanding applications running at scale — maximizing both resilience and speed.

Designed for Distributed, Performance-Hungry Environments

SupremeRAID™ HE achieves breakthrough performance in clustered and distributed systems by shifting RAID operations from the CPU to the GPU. By leveraging GPU acceleration, it unlocks exceptionally high NVMe throughput — up to 132GB/s for reads and 83GB/s for writes after RAID processing — freeing up CPU resources for critical application workloads and improving overall system efficiency.

Built for environments where high availability is critical, SupremeRAID™ HE features integrated array migration across nodes and cross-node RAID protection without the need for data replication. That means more usable capacity, lower TCO, and simplified failover…even in the most demanding enterprise and HPC infrastructures.

Now Supporting Ceph, Lustre, MinIO, IBM SpectrumScale, and More

Originally showcased in a joint reference architecture with Supermicro and BeeGFS, SupremeRAID™ HE is now fully hardware- and file-system agnostic. The latest release supports popular parallel file systems including Ceph, Lustre, MinIO, and IBM SpectrumScale, as well as deployment across JBOFs with Broadcom BF3 switches. This flexibility makes SupremeRAID™ HE ideal for organizations looking to maximize NVMe efficiency across diverse storage environments.

Unlock the Full Potential of NVMe

Whether you’re scaling an AI cluster, building out an HPC node, or modernizing your enterprise storage architecture, SupremeRAID™ HE empowers your team to achieve:

• GPU-accelerated RAID performance with near-theoretical throughput
• Cross-node high availability without replication
• Increased CPU headroom for upper-layer services
• Seamless integration across a broad range of infrastructure

For sales or reseller inquiries, contact us at [email protected].

Explore Our Joint HPC Solution with Supermicro

To learn how SupremeRAID™ HE performs in a 2U dual-node all-flash system with BeeGFS and Supermicro’s Petascale High Availability Dual Node All-Flash platform, check out our full solution brief: Read it here.

👉 Get More Info: Streamlining Extreme Performance: SupremeRAID™ HE, Supermicro’s All-Flash SBB Platform, and BeeGFS for Unmatched NVMe Storage

SupremeRAID™ HE Technical Specs

SupremeRAID™ HE Linux Driver 1.7.2 Release Notes, Software, and Documentation

• Supports up to 32 drives
• Supports RAID 0/1/5/6/10
• Supports NVMe-oF
• Supports dual controllers
• Supports array migration
• Support for erasure coding coming soon
• Linux only
• OS: Compatible with Ubuntu versions 20-24.04 and RHEL versions 8.x-9.x

SupremeRAID™ HE is bundled as a GPU card + license: GPU options are NVIDIA RTX A1000 8GB or NVIDIA 2000 Ada 16GB, see physical specs below for more information.

NVIDIA RTX A1000 Workstation GPU (8GB)

• GPU Memory: 8GB GDDR6
• Display Ports: 4x mini DisplayPort 1.4a
• Max Power Consumption: 50 W
• Graphics Bus: PCIe Gen 4 x 8
• Form Factor: 2.7” (H) x 6.4” (L) single slot
• Thermal: Active

NVIDIA RTX 2000 Ada Generation GPU (16GB)

• GPU Memory: 16GB GDDR6 with error-correcting code (ECC)
• Display Ports: 4x mini DisplayPort 1.4a
• Max Power Consumption: 70 W
• Graphics Bus: PCIe Gen 4 x 8
• Form Factor: 2.7” (H) x 6.6” (L) dual slot
• Thermal: Active
• VR Ready: Yes

‍