CARDINAL
EU
Hardware catalog / NVIDIA H100 / H100 PCIe 80GB
EXACT VARIANT REFERENCE

H100 PCIE 80GB

Choose between SXM modules, standard PCIe cards and NVL cards before comparing offers

SPECIFICATIONS

Memory per accelerator80 GB HBM2e
Memory bandwidth2000 GB/s
Form factorDual-slot PCIe card
Maximum accelerator power350 W

Manufacturer specifications describe the product, not the condition of an offered unit. Checked 2026-09-08

COMPUTE SPECIFICATIONS

Manufacturer peak arithmetic per accelerator. These values do not predict application throughput or latency.

PrecisionExecutionPeak rateSparsityReference
FP64Vector26 TFLOPSDenseSource ↗
Conditions

Manufacturer peak per GPU; not sustained workload performance

FP32Vector51 TFLOPSDenseSource ↗
Conditions

Manufacturer peak per GPU; not sustained workload performance

FP64Tensor51 TFLOPSDenseSource ↗
Conditions

Manufacturer peak per GPU; not sustained workload performance

  • Peak arithmetic rates are not token throughput, latency or a guarantee of application performance
  • Structured-sparse rates require eligible sparsity and supported kernels; do not compare them directly with dense rates
  • NVIDIA September 2023 article explicitly reports these theoretical PCIe rates; an older linked H100 datasheet reports different preliminary values and is not merged into this table
  • Low-precision tensor values omitted pending an exact PCIe source with unambiguous sparsity footnotes; no SXM or NVL figures substituted
  • Compute values are reported specifications, not derived from CUDA-core count or assumed boost clocks

PLATFORM AND SOFTWARE

  • Passive cooling requires suitable server airflow
  • Confirm power cabling and three-bridge set when requesting an NVLink pair
  • NVLink uses the 600 GB/s value in Table 6; the brief introduction contains a conflicting figure
  • Confirm CUDA, driver and framework compatibility against the intended server image
  • Ask whether any software subscription can transfer with the particular used offer

Manufacturer maximum accelerator rating; actual draw and supported limits depend on system

INTERCONNECT

NVLink

600 GB/s. Manufacturer aggregate bidirectional bandwidth; topology depends on platform

PCIe Gen5 x16

128 GB/s. Theoretical aggregate bidirectional interface bandwidth

MANUFACTURER SOURCES

NVIDIA H100 PCIe product brief ↗NVIDIA · Checked 2026-09-08NVIDIA H100 PCIe theoretical arithmetic rates in HPC system article ↗NVIDIA · Checked 2026-09-08

HOW CARDINAL HELPS

Share your quantity, destination and server model, if known. We help turn those requirements into a clearly specified request.

  • Help confirm the exact configuration and platform requirements
  • Request supplier records for condition, unit identifiers and available testing evidence
  • Clarify included accessories, delivery and proposed warranty terms before you decide

Any additional testing or guarantees must be agreed in the written offer.