NVIDIA 900-9X7AX-003NMC0 | ConnectX-7 | 400GbE QSFP112 | PCIe Gen5 x16 | RoCEv2 | SR-IOV | GPUDirect | PTP 1588v2 AI The NVIDIA L40 GPU is

The NVIDIA L40 GPU is a professional data center graphics accelerator built on Ada Lovelace architecture

Designed for AI/ML training fabrics

reliable links are needed

providing low-latency

NVIDIA 900-9X7AX-003NMC0 | ConnectX-7 | 400GbE QSFP112 | PCIe Gen5 x16 | RoCEv2 | SR-IOV | GPUDirect | PTP 1588v2 AI The NVIDIA L40 GPU isUltra high performance Ethernet adapter delivering up to 400GbE over QSFP112 for AI clusters, storage, and cloud scale networking. PCIe Gen5 x16 host interface and extensive hardware offloads (NVMe oF TCP, OVS, TLS IPsec, RoCEv2) maximize throughput while minimizing CPU usage. Supports IEEE 1588v2 PTP timing, advanced congestion control, and breakout options such as 4x100G where supported. Interoperable with QSFP family optics cables rated for the