Ingrasys IGS-HXR200 vs NVIDIA HGX Rubin NVL8
Accelerator configuration, aggregate memory, bandwidth and power, side by side.
Specifications
| Specification | Ingrasys IGS-HXR200 | NVIDIA HGX Rubin NVL8 |
|---|---|---|
| Vendor | Ingrasys | NVIDIA |
| Part number | IGS-HXR200 | — |
| Form factor | rack | baseboard |
| Released | — | 2026 |
| Accelerator | 8x Rubin | 8x Rubin |
| Aggregate memory | — | 2.3 TB |
| Aggregate bandwidth | — | 176.0 TB/s |
| System power | — | — |
| Interconnect | One NVIDIA HGX Rubin NVL8 platform, co-designed around NVLink 6, ConnectX-9, BlueField-4 and Spectrum-X; a single PCIe 5.0 x16 full-height half-length slot in the node itself | NVLink 6 Switch, 28.8 TB/s total |
Rows shown in grey are identical between the two.
Ingrasys IGS-HXR200
Ingrasys' implementation of NVIDIA's HGX Rubin NVL8, pairing eight Rubin GPUs with two Intel Xeon 6 processors in a 2U 19 inch chassis measuring 87 by 448 by 800 mm. Unlike the company's AMD and Intel nodes this is a conventional rack part rather than an Open Compute one. It is entirely liquid cooled, managed over DC-SCM 2.0 with an AST2600, and carries eight E1.S drives with two M.2 alongside a single full-height half-length PCIe 5.0 x16 slot, the networking being handled by the platform rather than by add-in cards. Ingrasys publishes no power, memory or throughput figure for the node; the linked HGX Rubin NVL8 entry carries NVIDIA's.
Vendor product pageNVIDIA HGX Rubin NVL8
Eight Rubin SXM on an HGX baseboard, listed by NVIDIA alongside HGX B300 and HGX B200. NVIDIA publishes NVFP4 twice and the two are different workloads, not a sparse/dense pair: inference 400 PFLOPS sparse, training 280 PFLOPS dense. FP8/FP6 training is 140 PFLOPS dense. System power is not published for the baseboard; the 24 kW figure on DGX Rubin NVL8 is the full chassis including CPUs. NVIDIA marks the specification preliminary.
Vendor product page