Why Your Wi-Fi 7 Access Point Won’t Deliver Multi-Gig Speeds (Yet)

A CW-917x access point can utilize a 2.5G, 5G or 10G Ethernet port. That is not the speed your laptop will see. Most clients are 2 spatial streams. Offices use 80 or 160 MHz, not 320 MHz channel width. Many closets are still 1G PoE 802.3af, and high-tier APs require PoE+ 802.3at. The WAN or MX path is often under 1 Gbps. Test on the LAN with a known client. Fix the bottleneck that is actually in the path.

Why does the datasheet say multi-gig and the speed test does not?

Datasheets advertise every radio’s peak PHY under ideal modulation. A speed test measures one client’s TCP through air, AP CPU, Ethernet, switch and usually the internet under real life situations.

If any hop is 1 Gbps, TCP will not beat about 900 Mbps, Wi-Fi 7 or not.

  Why a Wi-Fi 7 AP may not show multi-gig on a speed test
  
      Bottleneck
      What people assume
      What usually happens
    

  
      Client radios
      Wi-Fi 7 AP = multi-gig to the laptop
      Most phones and laptops are 2 spatial streams. PHY is a fraction of the AP aggregate
    

      Channel width
      320 MHz everywhere
      Dense floors stay on 80 or 160 MHz so neighbours can reuse 6 GHz
    

      MCS / SNR
      4096-QAM at every desk
      Modulation drops as soon as the client leaves the high-SNR bubble
    

      Ethernet uplink
      AP has a 5G or 10G port
      Switch is 1G PoE+, or the cable will not train above 1G
    

      PoE class
      Any PoE switch is fine
      CW9176I / CW9178I need 802.3bt for full radios and full mGig. On 802.3at they cut streams and link speed
    

      WAN / firewall
      Office Wi-Fi is slow
      Internet path is 500 Mbps–1 Gbps. Speedtest.net is not a radio test
    

      Radio mode
      All bands on
      6 GHz off, dual-5 GHz when you expected 6 GHz, or a guest SSID in NAT
    

What throughput should you plan for?

Plan based on current RF interference patterns, uplink speeds, and user density.

  Practical indoor TCP throughput bands (planning only)
  
      Client
      Typical setup
      Honest TCP band
    

  
      Wi-Fi 6, 2x2, 80 MHz @ 5 GHz
      2 SS
      ~400–700 Mbps
    

      Wi-Fi 6E, 2x2, 160 MHz @ 6 GHz
      2 SS
      ~800 Mbps–1.4 Gbps if the uplink allows it
    

      Wi-Fi 7, 2x2, 160 MHz + MLO
      2 SS
      Often similar peak to good 6E, steadier under load
    

      Wi-Fi 7, 4x4, wide 6 GHz
      Rare in laptops
      Only then are multi-gig client tests realistic
    

MLO is the feature that usually improves consistency, not the headline number. MCS, streams and width are unpacked further in the planned speeds article.

How do you test without lying to yourself?

Note the client NIC (Wi-Fi 6 vs 6E vs 7, 2x2 vs 4x4).
Confirm the AP Ethernet negotiated speed and PoE class in dashboard.
Run iPerf (or equivalent) to a server on the same LAN, not through the WAN.
Repeat near the AP and at the far desk.
Only then run an internet test — that measures the circuit and the MX, not the radio.

If step 3 is ~900 Mbps and the AP port is 1G, you have finished the radio investigation.

What has to change before multi-gig is a fair KPI?

  What else you must buy if multi-gig is the KPI
  
      If the KPI is
      You also need
    
      Multi-gig internet
      WAN and firewall above 1G, then the AP path
    
      Multi-gig to a local server
      mGig / 10G switchports, cabling that trains, 802.3bt where the SKU requires it, a 4x4 or wide-channel Wi-Fi 7 client
    
      Zoom just works
      Enough APs, roaming, and mixed-client design — not a 10G drop to every ceiling

Buying CW9178I onto 1G PoE+ and calling the project “multi-gig Wi-Fi” is how tickets start. Size the AP to the closet, or size the closet to the AP. That mapping is in the AP choice guide.

FAQ

Is the AP broken if we only see 600 Mbps?

Not if the client is 2x2 on 80 MHz or the uplink is 1G. Check RSSI values and channel associated on client device.

Does MLO give us multi-gig?

MLO can use more than one band at once. It helps reliability. It does not invent spatial streams the laptop does not have.

Should we wait to buy Wi-Fi 7?

No, if you are buying new APs anyway. Buy the right tier and switch. Do not promise a multi-gig speed-test SLA on Wi-Fi 6 laptops.

Is measuring speedtest the best way to benchmark wireless?

No. Speedtest is one metric, but it's usually a poor standalone benchmark for wireless quality. Measuring RSSI, packet loss, SNR, re-transmission rate are better metrics to benchmark a wireless network.

Send the AP model, switchport speed / PoE class, and the client used for the test. Megatron Technology will say which hop is the cap before anyone orders a bigger AP.

Request a design review: www.megatron.com.sg/ciscomeraki · Contact

john franciscoSeptember 30, 2026