Penguin Solutions Relion XE5318GTS vs QCT QuantaGrid D75F-7U
Both are built on the NVIDIA HGX H200 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 | Penguin Solutions Relion XE5318GTS | QCT QuantaGrid D75F-7U |
|---|---|---|
| Vendor | Penguin Solutions | QCT |
| Part number | Relion XE5318GTS | D75F-7U |
| Form factor | rack | rack |
| Released | — | — |
| Accelerator | 8x H200 | 8x H200 |
| Aggregate memory | — | 1.1 TB |
| Aggregate bandwidth | — | — |
| System power | — | — |
| Interconnect | Penguin publishes no interconnect figure; expansion is 12x PCIe Gen5 x16 low profile and 1x PCIe Gen4 x16 low profile | One NVIDIA HGX H200 baseboard connecting the eight GPUs at 900 GB/s over NVLink, which QCT notes is seven times the bandwidth of PCIe Gen 5; two BlueField-3 DPUs handle networking |
Rows shown in grey are identical between the two.
Penguin Solutions Relion XE5318GTS
The Penguin Solutions Relion XE5318GTS is a 5U GPU server carrying eight NVIDIA HGX H200 SXM5 modules of 141 GB each, hosted by 5th and 4th Gen Intel Xeon Scalable processors with 12x PCIe Gen5 x16 low profile and 1x PCIe Gen4 x16 low profile. The Intel-hosted Hopper machine, and the most expansion-heavy server in the range with thirteen slots. Penguin publishes only rack height, host processor, expansion slots and the accelerator for its servers, so no power draw, total memory, interconnect bandwidth or throughput figure is shown here; its full datasheets are hosted off its own domain and were not used as a source.
Vendor product pageQCT QuantaGrid D75F-7U
QCT's 7U Hopper machine, with two fifth or fourth generation Intel Xeon Scalable processors and eight NVIDIA HGX H200 GPUs linked at 900 GB/s over NVLink. Storage is eighteen small-form-factor all-NVMe bays for GPU-Direct Storage plus a boot drive, and two NVIDIA BlueField-3 DPUs offload networking. QCT gives 1.1 TB of aggregate high-bandwidth memory. It also quotes over 32 petaflops of FP8, which is not shown here: QCT does not say whether that is dense or sparse, it is close to double the dense figure NVIDIA published for this baseboard, and NVIDIA no longer lists the H200 on its HGX page, so there is no way to settle which column it belongs in without guessing. Power is four-plus-two 4,000 W Titanium supplies, which is capacity rather than draw.
Vendor product page