bandwidth downgraded for CX-5 and CX-6 Lx on Redhat9.1 with driver mlnx-en-23.07-0.5.0.0-rhel9.1-x86_64
This is a vague post though, my recommendation first is to make sure that the server(s) are properly tunned from the BIOS & OS level for our HCA/Driver.
We have various community posts which go over recommended tunning (ROCE, IB, IPoIB, TCP, VMA etc.).
Should you still observe a degradation, I would suggest opening a support case with our team so we can further investigate based on your deployment and assist you accordingly.
I have do following tune on my server :
- BIOS setting
- Hyperthread enable
- NPS 2
- L3 cache as NUMA
- x2APIC Mode - enable
- System Profile Settings - Custom
• CPU Power Management - Max Performance
• Turbo boost - enable
• C State - disable
• Memory Patrol Scrub - Standard
• Memory Refresh Rate - x1
• PCI ASPM L1 Link - Disable
• Determinism Slider - Performance
• ApbDis – enable
• ApbDis P-State - P0