1 1

Cisco

Cisco UCSC-GPU-H100-NVL | H100 NVL 94GB GPU Accelerator, 400W, 2-Slot FHFL

UCSC-GPU-H100-NVL

Stock Status: Enquire

Request Quote
Sale Sold out
Shipping calculated at checkout.
20-min response

Need a quote or bulk pricing? Answered within 20 minutes, Sydney business hours.

Description

The NVIDIA H100 NVL is a Hopper-architecture Tensor Core GPU accelerator for Cisco UCS C-Series rack servers. It is built for large language model inference, generative AI, deep learning and HPC acceleration in enterprise data-centre environments.

The card carries 94GB of GPU memory with memory bandwidth of nearly 4,000 GB/s, a 400W maximum thermal design power and a PCI Express Gen5 host interface. It is a dual-slot, full-height full-length (FHFL) card measuring 10.5 inches, with a passive heat sink, so the host server must supply sufficient airflow. Supported precisions include FP64, FP32, FP16, FP8 and INT8, and Multi-Instance GPU allows up to 7 GPU instances per GPU.

Two cards can be paired with NVLink bridges for up to 600 GB/s of bidirectional bandwidth, around 10x PCIe Gen4. The card is suited to servers such as the UCS C240 M7 with full-height, full-length x16 slots that support double-wide GPUs. Check the specific server spec sheet for supported riser and GPU-count rules.

Enquire with XS Network Tech for new or used pricing and availability.

Features

- Fourth-generation Tensor Cores based on the Hopper architecture
- Transformer Engine with FP8 precision for large language model workloads
- Second-generation Multi-Instance GPU (MIG) for secure partitioning in virtualised environments
- NVIDIA Confidential Computing support on Hopper
- NVLink bridge support to pair two cards for 600 GB/s bidirectional bandwidth
- Passive heat sink design for rack servers with directed airflow

Warranty

All products sold by XS Network Tech include a 12-month warranty on both new and used items. Our in-house technical team thoroughly tests used hardware prior to sale to ensure enterprise-grade reliability.

All technical data should be verified on the manufacturer data sheets.

View full details

specs-tabs

Technical Specifications

FAQs

Technical Specifications

- GPU architecture: NVIDIA Hopper
- GPU memory: 94GB
- Memory bandwidth: nearly 4,000 GB/s
- Maximum thermal design power (TDP): 400 W
- Form factor: PCIe dual-slot, full-height full-length, 10.5 inch
- Host interface: PCI Express Gen5
- Cooling: Passive heat sink (system airflow required)
- Interconnect: NVLink via three NVLink bridges, 600 GB/s bidirectional between two cards
- Multi-Instance GPU: up to 7 instances per GPU
- Supported precisions: FP64, FP32, FP16, FP8, INT8
- Operating humidity: 5% to 85% relative humidity

FAQs

Q: What power does the H100 NVL need?
A: The card has a maximum TDP of 400W and is powered via a PCIe 16-pin cable. It uses a passive heat sink, so the host server must supply sufficient airflow.

Q: Which servers is it used in?
A: It is sold for Cisco UCS C-Series rack servers such as the C240 M7, using GPU risers with full-height, full-length x16 slots that support double-wide GPUs. Check the specific server spec sheet for riser and GPU-count rules.

Q: Can two cards be linked together?
A: Yes. Two H100 NVL cards can be paired with NVLink bridges for up to 600 GB/s of bidirectional bandwidth. It uses the same NVLink bridges as H100 PCIe cards.

Q: How does this differ from other Cisco H100 NVL part numbers?
A: This SKU is the 94GB, 400W, 2-slot FHFL passive card for UCS C-Series servers. Other Cisco PIDs exist for different platforms, such as UCSX-GPU-H100-NVL for the UCS X440p PCIe Node.

Related Products

View all Graphics Card | GPU

Recently Viewed

  • Request a Quote

    Looking for competitive pricing? Submit a request, and our team will provide a tailored quote that fits your needs.

  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Call us Now  

  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Contact us