Inventec P5800G7 vs QCT QuantaGrid D75H-10U
Both are built on the NVIDIA HGX B300 8-GPU baseboard, so their accelerator performance is identical rather than merely similar. The choice between them is chassis, cooling, expansion and supplier, not FLOPS.
What actually differs
| Specification | Inventec P5800G7 | QCT QuantaGrid D75H-10U |
|---|---|---|
| Vendor | Inventec | QCT |
| Part number | P5800G7 | D75H-10U |
| Form factor | rack | rack |
| Released | — | — |
| Accelerator | 8x B300 | 8x B300 |
| Aggregate memory | — | — |
| Aggregate bandwidth | — | — |
| System power | — | — |
| Interconnect | One NVIDIA HGX B300 NVL8 liquid-cooled baseboard, with eight 800Gb OSFP connectors driven by NVIDIA ConnectX-8 SuperNICs for east-west traffic | One NVIDIA HGX B300 8-GPU baseboard; scale-out is eight ConnectX-8 OSFP ports carrying 800G Ethernet or InfiniBand at one port per GPU for GPU-Direct RDMA |
Rows shown in grey are identical between the two.
Inventec P5800G7
Inventec's densest Blackwell Ultra node: the same eight-accelerator HGX B300 NVL8 baseboard as its 10U air-cooled sibling, folded into 5U at 448 by 219.5 by 875 mm by moving to liquid cooling. The fan count tells the story, four 80 mm GPU fans against fifteen in the air-cooled machine. East-west networking is eight 800Gb OSFP connectors driven by NVIDIA ConnectX-8 SuperNICs, one per accelerator, and hosts are Intel Xeon 6700 series with E-cores or 6700 and 6500 series with P-cores. Power is 3,300 W 54 V supplies in 4+1 or 4+2 redundancy with a 2,000 W 12 V host plane. This is the only Inventec node whose GPU count is stated on the page, in the NVL8 designation itself. No throughput, memory or system power figure is published.
Vendor product pageQCT QuantaGrid D75H-10U
QCT's 10U Blackwell Ultra server, built on the NVIDIA HGX B300 platform with two Intel Xeon 6 processors. Networking is eight ConnectX-8 OSFP ports at 800G, one per accelerator, for one-to-one GPU-Direct RDMA, and there is room for four full-height full-length PCIe Gen 5 slots. QCT quotes 72 PFLOPS of FP8 and 144 PFLOPS of FP4 without saying whether either is dense or sparse. Both match NVIDIA's sparse figures for this baseboard exactly, and neither matches the dense figures of 36 and 108, so they are recorded as sparse. Worth noting QCT calls the FP8 number a training figure, which would normally imply dense; the number itself settles the basis, not the label. Power is twelve 3,200 W supplies in N+N redundancy, which is capacity rather than draw, so no system power is shown.
Vendor product page