2.4Ghz IoT Network Speed Test Comparison

I performed comparative speed testing of the IoT network configured for 2.4 GHz only. I wanted to provide some controlled data because users have reported relatively low Internet download speeds—often around 25–50 Mbps—when connected to the 2.4 GHz IoT network.

 

For these tests I used my M4 MacBook Pro as the client. The Mac was approximately 8 feet from my third-floor RBE970 satellite, which is Ethernet-backhauled to the RBE971 router. I tested the regular network on 6 GHz, the IoT network on 5 GHz, and then the IoT network restricted to 2.4 GHz. All tests therefore used the same client, location, Internet connection and nearby RBE970.

 

The results are in a table in the attached image.

 

The wireless connection information shows a very substantial difference between the three connections. On the regular 6 GHz network, the M4 connected on channel 69 at 160 MHz with a Tx rate of 1,693 Mbps. On the 5 GHz IoT network, it connected on channel 48 at 160 MHz with a Tx rate of 1,667 Mbps. Those connections produced average Internet download throughput of approximately 1,480 Mbps and 1,308 Mbps respectively.

 

On the 2.4 GHz IoT network, the Mac connected on channel 11 using a 20 MHz channel, with a reported Tx rate of 229 Mbps. Actual Internet throughput was consistently much lower, at 62–64 Mbps download and 67–75 Mbps upload.

 

The 2.4 GHz test was performed under what should be close to ideal RF conditions. RSSI was −32 dBm and noise was −89 dBm, for approximately 57 dB of signal-to-noise margin. The M4 was also only about eight feet from the RBE970. Therefore, weak signal strength or excessive distance from the satellite should not explain the approximately 63 Mbps download result.

 

The comparison should also largely eliminate my Internet connection, Ethernet backhaul or general RBE970 throughput as explanations. The same Mac in the same location was immediately capable of approximately 1.3 Gbps over the 5 GHz IoT network and approximately 1.48 Gbps over the regular 6 GHz network. The large throughput reduction appears specifically when the client is connected to the 2.4 GHz IoT network.

 

Certainly, some reduction is expected. The 2.4 GHz connection is using only 20 MHz of channel width, compared with 160 MHz on 5 and 6 GHz, and its negotiated Tx rate is only 229 Mbps, compared with approximately 1.7 Gbps on the other two bands. A 229 Mbps PHY/Tx rate also should not be interpreted as 229 Mbps of available TCP or Internet throughput. Wi-Fi protocol overhead, acknowledgments, contention, retransmissions and half-duplex operation all reduce real-world throughput, and 2.4 GHz can be a particularly busy RF environment.

 

However, I think there is another aspect of these results worth examining. The actual download throughput as a percentage of the reported Tx rate differs considerably between the bands. The 6 GHz connection delivered approximately 87% of its reported Tx rate, and the 5 GHz connection approximately 78%. The 2.4 GHz connection delivered only approximately 28% of its 229 Mbps reported Tx rate.

 

I realize that these percentages cannot be directly compared as though PHY-to-application throughput efficiency should be identical across different bands and channel widths. A 20 MHz 2.4 GHz connection has substantially different airtime and overhead characteristics from a 160 MHz 5 or 6 GHz connection. Nevertheless, approximately 28% seems low enough to raise the question of whether the observed 62–64 Mbps throughput is actually what Netgear expects from this connection.

 

For perspective, even 45% efficiency from the reported 229 Mbps Tx rate would produce approximately 103 Mbps of actual throughput; 50% would be approximately 115 Mbps; and 55% would be approximately 126 Mbps. I am not suggesting that any particular one of those figures is necessarily the correct expected throughput. Rather, they illustrate why the approximately 63 Mbps result may deserve further investigation instead of being attributed solely to the narrower 20 MHz channel.

 

This also changes somewhat how I view the Community reports of 25–50 Mbps. On one hand, those speeds are certainly plausible for older or less capable clients, single-stream devices, devices farther from the router or satellite, or devices operating in a more difficult 2.4 GHz RF environment. My M4 represents a much more favorable case.

 

On the other hand, my test reproduces relatively low 2.4 GHz throughput even with a modern Wi-Fi 6 client, only eight feet from the RBE970, at −32 dBm RSSI and approximately 57 dB SNR. That makes the result more interesting. Rather than demonstrating that 25–50 Mbps must be normal, I think my testing establishes a useful best-case baseline against which those reports can be evaluated.

 

So my main question for Netgear engineering would be: Is approximately 60–65 Mbps of actual throughput expected from a capable Wi-Fi 6 client that has negotiated a 229 Mbps 802.11ax connection at 20 MHz, with −32 dBm RSSI and excellent SNR? Or, under these conditions, would Netgear expect the client to achieve something closer to or above 100 Mbps?

 

If approximately 60–65 Mbps is expected, then the 25–50 Mbps results being reported by users may indeed be quite reasonable once differences in clients, distance and RF conditions are taken into account. But if Netgear expects substantially more throughput from a 229 Mbps PHY connection under these conditions, then my test may be reproducing the same underlying 2.4 GHz performance issue that other users are reporting.

 

 

Article originated and authored by donawalt​