NVIDIA HGX Rubin NVL8 platform with 8 Rubin SXM GPUs, 2.3 TB HBM4, sixth-generation NVLink, and support for Vera or x86 host CPU baseboards.
Fast Shipping
100k+ SKUs
8/5 Support
Certified
NVIDIA HGX Rubin NVL8 8-GPU Baseboard | HGX-RUBIN-NVL8
NVIDIA HGX Rubin NVL8 8-GPU Integrated Board (HGX-RUBIN-NVL8)
NVIDIA HGX Rubin NVL8 is the eight-GPU server platform for NVIDIA Rubin generation systems. The platform integrates eight Rubin SXM GPUs with sixth-generation NVLink and can be paired with either an NVIDIA Vera CPU baseboard or an x86 host platform, giving OEMs flexibility to build dense AI, analytics, and HPC systems around the same Rubin GPU complex.
Availability / RFQ: Rubin-based HGX systems are ramping into full production and shipping through the partner ecosystem. Exact OEM server configuration, CPU baseboard, cooling, support, and lead time vary by system manufacturer. Contact Catalyst Data Solutions to discuss current status, compatibility, configuration requirements, and sourcing options.
End-user / export compliance notice: Advanced AI accelerators and systems may be subject to applicable U.S. export controls, end-user/end-use restrictions, destination requirements, and OEM/channel policies. Catalyst may request purchaser, final end-user, intended-use, and deployment-location information before a quote or fulfillment can be approved. An RFQ does not guarantee allocation or shipment.
Technical Specifications
| SKU | HGX-RUBIN-NVL8 |
|---|---|
| Product Type | 8-GPU HGX AI Platform |
| GPU Configuration | 8x NVIDIA Rubin SXM GPUs |
| GPU Memory | 2.3 TB HBM4 total |
| GPU Memory Bandwidth | 176 TB/s aggregate |
| NVLink | Sixth-generation NVIDIA NVLink |
| NVLink Switch Bandwidth | 28.8 TB/s total |
| Host CPU Options | NVIDIA Vera CPU or x86 CPU baseboard |
| Networking Bandwidth | Up to 1.6 TB/s platform networking bandwidth |
| Availability Status | RFQ / Watchlist |
Deployment and Compatibility Considerations
- Choose the OEM server implementation first; HGX is a platform, not a finished rack server by itself.
- Confirm Vera versus x86 host CPU requirements based on workload, memory bandwidth, software, and fleet standards.
- Validate liquid cooling or OEM thermal design, power envelope, rack density, and service requirements.
- Size ConnectX-9 or equivalent validated networking and the scale-out Ethernet/InfiniBand fabric around the target cluster topology.
Common Use Cases
- Large-model training and post-training
- Agentic AI and reasoning inference
- Enterprise AI factories
- HPC and accelerated analytics
- OEM high-density eight-GPU Rubin servers
Frequently Paired Infrastructure
- OEM Rubin NVL8 servers from validated partners
- NVIDIA Vera CPU or supported x86 baseboards
- NVIDIA ConnectX-9 SuperNICs
- Spectrum-X Ethernet or Quantum-X800 InfiniBand
- AI storage, optics, cabling, power, and cooling
Official NVIDIA Documentation
NVIDIA HGX Rubin NVL8 Datasheet
Frequently Asked Questions
Can HGX Rubin NVL8 use an x86 CPU?
Yes, HGX Rubin NVL8 with x86 CPU baseboards and also supports a Vera CPU configuration called HGX Vera Rubin NVL8.
How much GPU memory is in an eight-GPU HGX Rubin NVL8 platform?
NVIDIA lists 2.3 TB of HBM4 memory across the eight Rubin GPUs.
| product type | HGX GPU Platform |
|---|---|
| Platform / Architecture | NVIDIA Rubin NVL8 |
| Accelerator Configuration | 8 Rubin SXM GPUs / 2.3 TB HBM4 |
| Host CPU Options | NVIDIA Vera or x86 Baseboard |
| Availability Status | Contact for Current Availability |
