Upload Speed Limitation on MR33 with IPQ4019

Setup

  • Device: Meraki MR33
  • Radio: IPQ4019 (2x2 embedded) - QCA9887 (1x1) disabled
  • Channel: 153 (5.765 GHz, VHT80)
  • TX Power: 20 dBm
  • Testing: WiFi speedtest.net (same town server)

Complete Test Results Across Versions

OpenWrt Version Download Upload Ping Notes
v21.02.7 (r16847-f8282da11e) 316.34 Mbps 25.80 Mbps 6-14 ms Best upload
v25.12.4 (r32933-4ccb782af7) 404.89 Mbps 23.15 Mbps 5-32 ms Best download
v24.10.7 (r29197-ab4c7d6af7) @ Ch.153 360.60 Mbps 23.27 Mbps 6-11 ms Channel optimized
v24.10.7 (r29197-ab4c7d6af7) @ Ch.36 321.24 Mbps 21.86 Mbps 6-11 ms Default channel
v23.05.6 (r24232-539228933c) 420.07 Mbps 21.62 Mbps 6-34 ms Highest download
v19.07.10 (r11427-9ce6aa9d8d) 310.69 Mbps 22.15 Mbps 7-12 ms Stable

Key Observation

Upload consistently limited to 21.62 - 25.80 Mbps range across all OpenWrt versions, despite switching to IPQ4019 (2x2 MIMO).

Optimizations Tested (Minimal Impact)

  • :white_check_mark: ath10k-ct driver (Candela)
  • :white_check_mark: Different firmware variants (full-htt, standard)
  • :white_check_mark: TX Power levels (20-30 dBm)
  • :white_check_mark: Kernel buffer tuning (128MB rmem/wmem)
  • :white_check_mark: Flow offloading modes
  • :white_check_mark: Channel selection (36 vs 153)
  • Result: None significantly improve upload beyond firmware version differences

Questions for MR33 Community

If you're running Meraki MR33 with OpenWrt, please share:

  1. Your OpenWrt version: cat /etc/openwrt_release | grep VERSION
  2. Which radio as primary AP? QCA9887 or IPQ4019?
  3. Your WiFi speedtest results:
    • Download: ___ Mbps
    • Upload: ___ Mbps
    • Ping: ___ ms
  4. Any special tuning you've applied?

Am I missing a configuration that unlocks higher upload speeds? :thinking:

What's the expected UL speed and how are you testing ?

Expected Upload Speed

Based on my ISP subscription and wired baseline tests, the expected upload speed is ~275 Mbps.

When a MacBook Pro M1 is connected via a Gigabit Ethernet adapter directly to the ONT, it consistently achieves 275 Mbps upload. The ~21–25 Mbps I am seeing on the MR33 represents a massive ~90% performance drop specific to the wireless upload path.

Testing Methodology & Environment

To isolate the issue and eliminate external variables, the test environment is configured as follows:

  • Physical Topology: [Internet] <--> [ZTE F6600R ONT/Router] <-- (Gigabit Ethernet) --> [Meraki MR33 AP]
  • Client Device: MacBook Pro M1 (2x2 MIMO).
  • Test Software: Tested exclusively via web browser on Speedtest.net, targeting the exact same local ISP server for both wired and wireless tests to ensure routing consistency.
  • RF Environment: * The MacBook and the MR33 are placed side-by-side on the desk to ensure a near-perfect RF environment with maximum signal strength, eliminating distance, walls, or attenuation as factors.
    • Wireless link-rate negotiation is verified as stable at max rates (VHT80, 2x2, MCS9).

The fact that download speeds scale beautifully up to 420 Mbps on wireless proves the Ethernet backhaul, local routing, and browser-based testing can easily handle high throughput—the bottleneck is strictly confined to the upstream wireless transmission.