NVIDIA Quantum Firmware v27.2008.2102 Release Notes
NVIDIA Quantum Firmware v27.2008.2102 Release Notes

Changes and New Features History

For older versions history, please refer to their dedicated Release Notes.

Category

Description

Rev. 27.2008.1904

SHARP

Added support for SHARP SAT performance counters.

Adaptive Routing

Added support for a new Adaptive Router counter (PortARTrials) as part of the PortRNCounters group. PortARTrials is used to indicate the number of times the AR decision mechanism is used per port

Rev. 27.2008.1604

SHARP

Added support for MAD security policy for SHARP.

Cables

Enabled PLR with Low-Latency (LL)-FEC mode in optical cables up to 30m.

Rev. 27.2008.1300

EDR Link in ConnectX-6 100Gb/s cards

EDR link speed is now supported when using ConnectX-6 100Gb/s HCA and connecting with HDR optical cables.

Rev. 27.2008.0232

Link Speed

Added support for FDR link speed when connecting a ConnectX-3 HCA with an HDR Mellanox Quantum switch.

SHARP

SHARP stability improvements.

SHARP

Added support for legacy SHARP port and performance counters.

Cables

Enabled KP4RS FEC on Active Fiber cables (OPN: MFS1S00-V0xxE).

Cables

Disabled PLR on Active Fiber cables.

Bug Fixes

Bug Fixes.

Rev. 27.2007.1124

Bug Fixes

Bug Fixes.

Rev. 27.2007.0618

Bug Fixes

Bug Fixes.

Rev. 27.2007.0300

Performance

Added support for link-negotiated credit size.

General

Added support for DCQCN Congestion Control.

Rev. 27.2000.3276

Mellanox Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™

SHARP (SAT) is at GA level.

*SAT: Streaming Aggregation Tree

Rev. 27.2000.2708

SHARP

SHARP now supports running 2 flows in parallel.

Rev. 27.2000.2626

Speed Link

SDR link speed on InfiniBand systems is now available for all cables, including cables that do not advertise InfiniBand speed in their memory map.

Mellanox Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™

SHARP (LLT) is at GA level, whereas SHARP (SAT) is at Beta level.

*LLT: Local Latency Tree

*SAT: Streaming Aggregation Tree

General

Added support for Error Injection with PTER register.

Bug Fixes

See Bug Fixes.

Release 27.2000.2306

General

Bug Fixes

Release 27.2000.2182

General

Bug Fixes

Release 27.2000.2046

Mellanox Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™

[Beta] Mellanox Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)™ technology improves the performance of MPI operations by offloading collective operations from the CPU to the switch network, and by eliminating the need to send data multiple times between endpoints.

Link Speed

Added QDR/FDR support in Quantum switch systems when using optical cables of up to 30m.

Note: QDR speed is only supported when using the FDR cables. See Validated and Supported FDR Cables

Cables

Removed PLR from active cables longer than 30m.

Release 27.2000.1886

Physical Layer Retransmission (PLR)

Added support to the Physical Layer Retransmission (PLR) functionality for HDR speed.

Link Speed

Link-up time improvements. The link up time is up to 60 sec

Release 27.2000.1600

General

Bug Fixes

Release 27.2000.1400

Link Speed

HDR speed rate is at GA level.

Cables

Added support for Break-Out Cable auto-sensing.

Release 27.2000.1142

Link Speed

HDR link stability enhancements.

Note: HDR is at beta level.

Release 27.2000.1012

Link Speed

HDR speed rate is currently at Beta level.

Release 27.2000.1004

Link Speed

Stability improvements for HDR link.

Subnet Manager (Adaptive Routing)

Added support for Adaptive Routing. Adaptive routing (AR) allows optimizing data traffic flow. The InfiniBand protocol uses multiple paths between any two points. Thus, when unexpected traffic patterns cause some paths to be overloaded, AR can automatically move traffic to less congested paths according to the current temporal state of the network.

Adaptive Routing

Added support for Private Linear Forwarding Tables (PLFT). This capability helps SM to optimize traffic.

© Copyright 2024, NVIDIA. Last updated on Sep 8, 2024.