Datasheet

QLogic QLE7340 Single-Port 40 Gbps QDR InfiniBand Host Channel Adapte
r
2
Part number information
Table 1. Ordering part numbers and feature codes
Description Part number Feature code
QLogic QLE7340 Single-Port 40 Gbps QDR InfiniBand HCA 59Y1888 5763
0.5 m QLogic Copper QDR InfiniBand QSFP 30AWG Cable 59Y1892 3725
1 m QLogic Copper QDR InfiniBand QSFP 30AWG Cable 59Y1896 3726
3 m QLogic Copper QDR InfiniBand QSFP 28AWG Cable 59Y1900 3727
3 m QLogic Optical QDR InfiniBand QSFP Cable 59Y1920 3731
10 m QLogic Optical QDR InfiniBand QSFP Cable 59Y1924 3732
30 m QLogic Optical QDR InfiniBand QSFP Cable 59Y1928 3733
3m IBM Optical QDR InfiniBand QSFP Cable 49Y0488 5989
10m IBM Optical QDR InfiniBand QSFP Cable 49Y0491 5990
30m IBM Optical QDR InfiniBand QSFP Cable 49Y0494 5991
Features and benefits
The QLogic QLE7340 HCA has the following summary of features and specifications:
40 Gbps InfiniBand interface (40/20/10 Gbps auto-negotiation)

3400 MBps unidirectional throughput

Approximately 30 million messages processed per second (non-coalesced)

Approximately 1.0 microsecond latency that remains low as the fabric is scaled

Multiple virtual lanes (VLs) for unique quality of service (QoS) levels per lane over the same physical

port
TrueScale architecture, with MSI-X interrupt handling, optimized for multi-core compute nodes

Operates without external memory

Optional data scrambling in InfiniBand link

Complies with InfiniBand Trade Association 1.2 standard

Supports OpenFabrics Alliance (OFED) software distributions, WinOF support planned for 1Q/2010

Key benefits of the QLogic QLE7340 HCA are:
High performance: Quad Data Rate (QDR) InfiniBand (IB) delivers 40 Gbps per port (4x10 Gbps),

providing the necessary bandwidth for high-throughput applications. With one of the highest message
rates and lowest latencies of any IB adapter, the QLE7340 provides superior HPC application
performance.
Superior scalability: QLogic's TrueScale architecture is designed to deliver near linear scalability. As

additional compute resources are added to a cluster, latency remains low and the message rate
scales with the size of the fabric, resulting in maximum utilization of compute resources.