The following is the `perf top` output of blue's CPU2 (which has been made to handle both tb0 and eno1) when running iperf3 from purple->red (the fast, ~930Mbps direction).

   PerfTop:    3689 irqs/sec  kernel:100.0%  exact: 100.0% lost: 0/0 drop: 0/0 [4000Hz cycles:P],  (all, CPU: 2)
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

     7.36%  [kernel]       [k] clflush_cache_range
     7.29%  [kernel]       [k] memcpy_orig
     4.61%  [kernel]       [k] ioread32
     3.83%  [kernel]       [k] pfn_to_dma_pte
     2.92%  [kernel]       [k] tbnet_start_xmit
     2.74%  [kernel]       [k] _raw_spin_lock_irqsave
     2.40%  [kernel]       [k] clear_page_erms
     2.40%  [kernel]       [k] intel_iommu_map_pages
     1.77%  [kernel]       [k] dma_pte_clear_level
     1.28%  [kernel]       [k] e1000_intr_msi
     1.26%  [kernel]       [k] cpuidle_enter_state
     1.19%  [kernel]       [k] menu_select
     1.18%  [kernel]       [k] dev_gro_receive
     1.18%  [kernel]       [k] e1000_clean_rx_irq
     1.15%  [kernel]       [k] csum_partial
     1.13%  [kernel]       [k] _raw_spin_lock
     1.09%  [kernel]       [k] e1000_irq_enable
     0.96%  [kernel]       [k] tcp_gro_receive
     0.88%  [kernel]       [k] kmem_cache_free
     0.85%  [kernel]       [k] memset_orig
     0.72%  [kernel]       [k] tbnet_poll
     0.66%  [kernel]       [k] __iommu_dma_unmap
     0.63%  [kernel]       [k] alloc_iova_fast
     0.62%  [kernel]       [k] __schedule
     0.62%  [kernel]       [k] br_handle_frame
     0.60%  [kernel]       [k] iommu_dma_free_iova
     0.58%  [kernel]       [k] e1000_xmit_frame
     0.58%  [kernel]       [k] e1000_alloc_rx_buffers
     0.54%  [kernel]       [k] skb_segment
     0.54%  [kernel]       [k] _raw_spin_unlock_irqrestore
     0.53%  [kernel]       [k] kmem_cache_alloc_noprof
     0.53%  [kernel]       [k] intel_iommu_unmap_pages
     0.51%  [kernel]       [k] inet_gro_receive
     0.51%  [kernel]       [k] skb_release_data
     0.49%  [kernel]       [k] fdb_find_rcu
     0.48%  [kernel]       [k] native_irq_return_iret
     0.48%  [kernel]       [k] iommu_pgsize
     0.47%  [kernel]       [k] ring_work
     0.47%  [kernel]       [k] get_page_from_freelist
     0.45%  [kernel]       [k] iommu_dma_map_page
     0.45%  [kernel]       [k] build_detached_freelist
     0.45%  [kernel]       [k] ring_write_descriptors
     0.44%  [kernel]       [k] cache_tag_flush_range_np
     0.43%  [kernel]       [k] iommu_map
     0.42%  [kernel]       [k] __iommu_map
     0.41%  [kernel]       [k] __iommu_dma_map
     0.41%  [kernel]       [k] iommu_dma_alloc_iova