NVIDIA HGX H100 8-GPU vs Supermicro GPU SuperServer AS -8125GS-TNHR (HGX H100)

Accelerator configuration, aggregate memory, bandwidth and power, side by side.

Specifications

SpecificationNVIDIA HGX H100 8-GPUSupermicro GPU SuperServer AS -8125GS-TNHR (HGX H100)
VendorNVIDIASupermicro
Part numberAS -8125GS-TNHR
Form factorbaseboardrack
Released20222024
Accelerator8x H1008x H100
Aggregate memory0.6 TB0.6 TB
Aggregate bandwidth26.8 TB/s38.4 TB/s
System power8.0 kW
InterconnectNVSwitch + NVLinkNVIDIA NVLink with NVSwitch between GPUs; PCIe 5.0 x16 CPU-to-GPU

Rows shown in grey are identical between the two.

NVIDIA HGX H100 8-GPU

HGX H100 8-GPU configuration for hyperscalers

Vendor product page

Supermicro GPU SuperServer AS -8125GS-TNHR (HGX H100)

Supermicro's 8U AMD EPYC platform for the NVIDIA HGX H100 8-GPU baseboard, and the EPYC sibling of the Intel-based SYS-821GE-TNHR. Two SP5 sockets take AMD EPYC 9004 processors up to 128 cores and 256 threads at up to 400 W each, air cooled, with 24 DIMM slots supporting up to 6 TB of DDR5. GPUs connect to the CPUs over PCIe 5.0 x16 and to each other over NVLink with NVSwitch, and the chassis provides eight NICs for GPU-direct RDMA at a 1:1 ratio with the GPUs. The CSE-GP801TS chassis is 8U, weighs 102.1 kg gross, takes up to ten heavy-duty fans and six 3000 W Titanium supplies in 3+3 redundancy. Supermicro publishes no FLOPS figure for this system; the compute shown is NVIDIA's for the HGX H100 8-GPU baseboard it carries. Supermicro also publishes no system power draw, only the supply configuration, so total power is left blank rather than quoting 9 kW of installed supply capacity as though it were consumption.

Vendor product page