ConnectX-7 Ethernet Datasheet - Nvidia

1 Nvidia ConnectX-7 . 400G Ethernet . SMART ACCELERATION FOR CLOUD, DATA-CENTER AND EDGE. ACCELERATED NETWORKING AND SECURITY FOR THE PRODUCT SPECIFICATIONS. MOST ADVANCED CLOUD AND AI WORKLOADS Maximum Total 400 GbE. Bandwidth The Nvidia ConnectX -7 SmartNIC is optimized to deliver accelerated networking for . Supported Ethernet 10/25/40/50/100/. modern cloud, artificial intelligence, and traditional enterprise workloads. ConnectX-7 Speeds 200/400 GbE. provides a broad set of software-defined, hardware accelerated networking, storage, Number of 1/2/4. and security capabilities which enable organizations to modernize and secure their IT. Network Ports infrastructures. Network Interface NRZ (10/25G) / PAM4. Extending the tradition of Nvidia 's industry leading innovation for networking, Technologies (50/100G). ConnectX-7 , is available in 1, 2, or 4-port configurations and delivers up to 400Gb/s of Host Interface PCIe x16/ x32. bandwidth. With features such as Nvidia ASAP2 - Accelerated Switching and Packet Processing , advanced RoCE, Nvidia GPUD irect Storage, and in-line hardware Cards Form Factors PCIe FHHL/ HHHL, acceleration for TLS/IPsec/MACsec encryption/decryption, ConnectX-7 empowers agile SFF.

2 And high-performance solutions from edge to core data centers to clouds, all while enhancing network security and reducing the total cost of ownership. Network Interfaces SFP56, QSFP56, QSFP56-DD, Available in PCIe card and form factors, ConnectX-7 empowers solutions for QSFP112, SFP112. cloud, hyperscale, and enterprise networking. Accelerate Software-Defined Provide Security from Networking Edge to Core Nvidia ASAP2 technology Hardware engines in ConnectX-7 . accelerates software-defined offload and accelerate security, with networking, delivering line-rate in-line encryption/decryption of TLS, performance with no CPU penalty. IPsec, and MACsec. Enhance Storage Performance Enable Precision Timing ConnectX-7 enables high- ConnectX-7 provides extremely performance storage and data access accurate time synchronization for with RoCE and GPUD irect Storage and data-center applications and timing- accelerates NVMe-oF over both RoCE sensitive infrastructures. and TCP. Nvidia ConnectX-7 | Datasheet | APR21.

3 FEATURES. Network Interface > ASAP2 accelerations/actions: counters, > PTP-based packet pacing QoS, NAT, aging, mirroring, sampling, > Time-based SDN acceleration (ASAP2). > Up to 4 network ports supporting NRZ, flow tag PAM4 (50G and 100G), in various ports configurations: > Hairpin flow for full hardware offload Storage Accelerations > 1 x 10/25/40/50/100/200/400 GbE > Highly-scalable number of classifications > NVMe over Fabrics (NVMe-oF) storage and actions target offloads > 2 x 10/25/40/50/100/200/400 GbE. > Application access to hardware statistics > NVMe-oF over TCP/RoCE acceleration > 4 x 10/25/40/50/100/200 GbE. > Application access to crypto offloads > Storage protocols: iSER, NFSoRDMA, SMB. > Up to 400Gb/s total bandwidth Direct, NVMe-oF , and more Host Interface Cyber Security > 32 lanes of PCIe Gen , compatible with > IInline hardware IPsec encryption HPC / AI. and decryption > All-to-All engine PCIe Gen 2/3/4. > AES-GCM 128/256-bit key > Nvidia GPUD irect > Integrated PCI switch > IPsec over RoCE > Nvidia GPUD irect Storage > Nvidia Multi-Host (up to 8 hosts).

4 And Nvidia Socket Direct > Inline hardware TLS encryption and decryption Management and Control > MSI/MSI-X mechanisms > AES-GCM 128/256-bit key > Advanced PCIe capabilities > SMBus > Inline hardware MACsec encryption and > Network Controller Sideband decryption Networking Interface (NC-SI). > AES-GCM 128/256-bit key > RoCE, Zero Touch RoCE > NC-SI, MCTP over SMBus and MCTP. > AES-GCM-XPN 128/256-bit key over PCIe - Baseboard Management > ASAP - Accelerated Switch and Packet > Data-at-rest AES-XTS encryption Controller interface > Processing for SDN and VNF acceleration and decryption > PLDM for Monitor and Control DSP0248. > Single Root I/O Virtualization (SR-IOV) > AES-XTS 256/512-bit key > PLDM for Firmware Update DSP026. > VirtIO acceleration > Platform security > I2C interface for device control > Overlay network acceleration: > Secure boot with hardware root-of-trust and configuration VXLAN, GENEVE, NVGRE. > Secure firmware update > General Purpose I/O pins > Programmable flexible parser: user-defined classification > On-board flash encryption > SPI interface to flash > Connection tracking (L4 firewall) > JTAG IEEE and Advanced Timing and > IEEE > Flow mirroring, sampling and statistics Synchronization > Header rewrite > Advanced PTP Remote Boot > Hierarchical QoS.

5 > IEEE 1588v2 (any profile) > Remote boot over Ethernet > Stateless TCP offloads > Meets Class C standard > Remote boot over iSCSI. > PTP hardware clock (PHC) (UTC format) > UEFI support for x86 and Arm servers VNF Acceleration > 12 nanosecond accuracy > PXE boot > Hardware offload programmable pipeline: > Line-rate hardware timestamp > Packet classification on network layers L2 to L4 and tunneled traffic such as GTP. (UTC format) Form Factors and Options and VXLAN > SyncE > PCIe HHHL/FHHL. > Packet dispatching to multiple cores > Meets (eEEC) > OCP SFF. > Multi-threaded API for concurrent update > Configurable PPS In and configurable of offloaded rules PPS Out > Time-triggered scheduling To learn more about the Nvidia ConnectX SmartNICs visit 2021 Nvidia Corporation. All rights reserved. Nvidia , the Nvidia logo, ConnectX, ConnectX-7 , GPUD irect, Multi-Host, Socket Direct, and ASAP 2. - Accelerated Switch and Packet Processing are trademarks and/or registered trademarks of Nvidia Corporation in the and other countries.

6 Other company and product names may be trademarks of the respective companies with which they are associated. MAY21.