Home / Blog / Ethernet vs. Wi-Fi 6E
Performance 10 min read

Ethernet vs. Wi-Fi 6E on Mac: A Real-World Latency Comparison

We ran 10,000 pings on Wi-Fi 6E and Ethernet from the same MacBook Pro, sitting 2 feet from the router. The results weren’t close.

Healthy Network Team Network Engineers & macOS Developers

The Test Setup

We wanted to settle this debate with data, not opinions. Here’s exactly what we used and how we controlled the test.

Hardware

  • MacBook Pro 14” M3 Pro (2023), one of the first Macs to support Wi-Fi 6E on the 6 GHz band
  • Belkin USB-C to 2.5Gb Ethernet Adapter, Apple’s current recommended adapter ($29.95, sold through Apple)
  • Ubiquiti U6 Enterprise access point, Wi-Fi 6E capable, mounted 2 feet from the laptop with direct line of sight
  • ISP connection: 1 Gbps fiber (symmetrical). This matters: baseline Ethernet latency on fiber is typically 1–5ms to nearby servers, while cable connections run 5–15ms. Our fiber connection gives Ethernet every advantage, which makes the comparison a best-case scenario for both connections.

Methodology

  • 10,000 TCP probes to Cloudflare (1.1.1.1) on each connection type, using Healthy Network’s continuous monitoring
  • 24-hour monitoring sessions for each connection, same time windows across consecutive days to control for ISP congestion patterns
  • Wi-Fi connected to the 6 GHz band (channel 37, 160 MHz width), the best Wi-Fi 6E has to offer
  • No other devices active on the network during testing. Same DNS (1.1.1.1), same physical location, same macOS version (Sequoia 15.3)
  • AirDrop set to “Everyone” during the Wi-Fi test to capture realistic AWDL behavior. Most users never disable it.

We ran each connection type separately, not simultaneously, to prevent any cross-interface interference. The Ethernet adapter was physically disconnected during Wi-Fi testing, and Wi-Fi was turned off at the system level during Ethernet testing.

The Results

Median Latency: Closer Than You’d Think

Here’s the headline number most articles would stop at:

Ethernet: Median Latency
3.2ms to Cloudflare (1.1.1.1). Rock solid. On a fiber connection, the wired path adds almost nothing to the base network round-trip time.
Wi-Fi 6E: Median Latency
5.8ms to Cloudflare (1.1.1.1). Only 2.6ms slower at the median. If you stopped here, you’d conclude Wi-Fi 6E is “basically the same.” That conclusion would be wrong.

On a cable ISP, both numbers would shift up (Ethernet to roughly 8–12ms, Wi-Fi to 12–18ms), but the gap between them follows the same pattern. The median isn’t the point. The spikes are.

Tail Latency (P95): The Real Story

The 95th percentile, the latency that 5% of your packets exceed, is where Ethernet and Wi-Fi stop looking similar and start looking like different technologies entirely:

Ethernet: P95 Latency
4.1ms. Only 0.9ms above the median. The latency distribution is tight, symmetrical, and predictable. Almost every packet arrives within the same narrow window.
Wi-Fi 6E: P95 Latency
34ms. Nearly 6x the median. The distribution has a long right tail; most packets are fast, but a significant minority take 5–10x longer. These are the packets that break your call.

That 34ms P95 means that roughly 1 in 20 packets experienced a latency spike of 30ms or more above baseline. During a Zoom call sending packets every 20ms, that’s a noticeable stutter every few seconds. During a competitive game running at 128 tick, it’s the difference between your shot registering and the server disagreeing with your client.

Jitter: The Consistency Gap

Jitter, the variation between consecutive packets, tells the consistency story even more starkly:

Ethernet: Jitter
0.4ms average, 1.2ms P95. The sparkline is essentially a flat line. Packet timing is metronome-steady.
Wi-Fi 6E: Jitter
3.1ms average, 28ms P95. The sparkline shows periodic spikes every 10–30 seconds, a distinctive sawtooth pattern. The baseline is good, but the spikes are relentless.

That sawtooth pattern on Wi-Fi? It’s not random noise. It has a cause, and we’ll get to it in the next section.

Packet Loss

Both connections showed near-zero packet loss in ideal conditions: Ethernet at 0.00%, Wi-Fi at 0.02%. In a perfect environment with no competing devices, Wi-Fi 6E holds up well on this metric. But “ideal conditions” is doing heavy lifting in that sentence. Add a few neighboring networks, a microwave, or move to the next room, and Wi-Fi loss climbs while Ethernet stays at zero.

Latency Under Load (Bufferbloat)

We ran a parallel download at full speed during a subset of the test to measure latency under load:

Ethernet: Under Load
Median latency rose from 3.2ms to 5.8ms. An increase, but still under 6ms. The wired path has minimal queuing overhead when the router has decent QoS.
Wi-Fi 6E: Under Load
Median latency rose from 5.8ms to 22ms. The P95 climbed to 85ms. Under load, the Wi-Fi radio’s contention overhead compounds with buffer bloat to create substantial latency spikes.

This is the scenario that matters for real life. You’re on a call while someone else in the house is streaming, or your Mac is syncing iCloud in the background. Ethernet barely notices the additional load. Wi-Fi buckles.

The Wi-Fi Killers: What Causes the Gap

If Wi-Fi 6E is so advanced, why is tail latency still 6x worse than Ethernet? Four factors explain the gap.

AWDL Channel Hopping

This is the biggest single contributor to Wi-Fi latency spikes on Mac. AWDL (Apple Wireless Direct Link) is the protocol behind AirDrop, AirPlay, and Sidecar. To discover nearby Apple devices, your Mac’s Wi-Fi radio periodically hops to three “social channels”: channel 6 (2.4 GHz), channel 44 (5 GHz, preferred in Europe), and channel 149 (5 GHz, preferred in the US).

Each channel hop takes 50–200ms. During that time, your Mac cannot send or receive data on its primary channel. The result is the distinctive sawtooth jitter pattern we saw in the test data: mostly flat with periodic sharp spikes every 10–30 seconds.

If your access point happens to already be on channel 149 (US) or 44 (EU), the penalty is reduced since the radio doesn’t need to change frequency for that particular hop. But channels 6 (2.4 GHz) and the other 5 GHz social channel still cause disruption.

What about 6 GHz AWDL? Apple has not documented whether AWDL uses a social channel on the 6 GHz band. Our testing showed the same spike pattern on 6 GHz Wi-Fi, strongly suggesting the radio still hops to the 2.4 GHz and 5 GHz social channels even when your primary connection is on 6 GHz.

Background Location Scanning

macOS’s locationd service periodically scans for nearby Wi-Fi access points to refine location accuracy. These scans force the Wi-Fi radio off-channel momentarily, similar to AWDL but less frequent. You can see these scans in Wireless Diagnostics logs as brief latency bumps that don’t follow AWDL’s regular cadence.

Interference and Congestion

Even on 6 GHz with zero competing networks, environmental factors add variability. Microwave ovens (2.4 GHz), Bluetooth devices sharing the radio’s antenna time, and even USB 3.0 devices emitting RF noise in the 2.4 GHz range all contribute. The 5 GHz and 6 GHz bands avoid most of these, but they’re still shared spectrum, and your neighbor’s Wi-Fi 6E router on an overlapping channel adds contention you can’t control.

Distance and Walls

Our test was conducted at 2 feet with line of sight, the absolute best case. At 15 feet, median Wi-Fi latency rose to 8ms with a P95 of 42ms. Through one drywall wall at 20 feet, the median hit 12ms and P95 reached 65ms. The higher the frequency band, the worse the wall penetration: 6 GHz is particularly sensitive, with a single wall capable of dropping signal strength enough to force the Mac to fall back to the 5 GHz band entirely.

Ethernet doesn’t care about walls. A 100-foot Cat 6 cable performs identically to a 3-foot one for latency purposes.

When Wi-Fi 6E Is Good Enough

Despite the data above, Wi-Fi 6E is not always the wrong choice. For many tasks, the tail latency gap simply doesn’t matter.

Browsing and Streaming

Web pages and video streams buffer aggressively. Netflix prefetches 30–60 seconds of video. A 34ms P95 latency spike is invisible when your player has a minute of content queued. For browsing, streaming, file downloads, and anything throughput-dependent rather than latency-dependent, Wi-Fi 6E delivers effectively the same experience as Ethernet.

The 6 GHz Band Advantage

Wi-Fi 6E on the 6 GHz band (supported on 2023+ MacBook Pros with M2 Pro/Max, M3, and later) is a genuine step forward. Independent testing from Cudy shows roughly 40% latency reduction and 80% jitter reduction compared to 5 GHz in congested environments, thanks to up to 59 additional non-overlapping channels and zero legacy device traffic.

But “closer to Ethernet” still isn’t Ethernet. The 6 GHz band reduces the gap meaningfully; it doesn’t close it. AWDL channel hopping still occurs, locationd still scans, and the physics of shared wireless spectrum still introduce variability that a wire simply doesn’t have. And 6 GHz’s range is shorter than 5 GHz, with noticeably worse wall penetration. If you’re not in the same room as your router, you may not benefit at all.

The “Good Enough” Threshold

If your Wi-Fi connection shows jitter consistently below 5ms and packet loss below 0.1%, most users won’t notice the difference from Ethernet for daily tasks. The question is whether your environment can sustain those numbers, and for many Mac users in apartments, shared offices, or multi-device households, it can’t.

When Ethernet Is Non-Negotiable

For some use cases, the P95 gap between wired and wireless isn’t a minor inconvenience; it’s a dealbreaker.

Competitive Gaming

In ranked play, every millisecond of consistency matters. At high ranks in games like Valorant, Apex Legends, or Street Fighter 6, the difference between a 4ms P95 and a 34ms P95 is the difference between your input registering as intended and the server reconciliation disagreeing with your client. If you’re investing hundreds of hours into competitive games, a $30 Ethernet adapter is the single highest-value peripheral you can buy.

Live Streaming and Podcasting

OBS, Streamlabs, and podcast recording tools have zero tolerance for connection drops. A single 200ms Wi-Fi spike during a live stream is a visible stutter for every viewer. A dropped packet during a podcast recording is an audible pop or gap that’s difficult to edit out. Professionals default to wired connections for a reason.

VoIP-Heavy Workdays

If your job involves 4–8 hours of video calls per day, the cumulative effect of Wi-Fi jitter spikes adds up. Each spike is a brief audio artifact: a word that clips, a momentary robotic distortion, a half-second freeze. On any single call it might be tolerable. Over a full day of back-to-back meetings, it’s death by a thousand cuts. Plug in the cable before your first call and your audio quality improves for the entire day.

Server Roles and Homelab

Mac Minis running as home servers, build machines, or Docker hosts need sustained, predictable connectivity. A Plex server transcoding and streaming to three devices can’t afford periodic 85ms latency spikes under load. A CI/CD build server pulling and pushing containers needs consistent throughput without wireless contention. For always-on roles, Ethernet is assumed, not optional.

The Best USB-C Ethernet Adapters for Mac

Modern Macs don’t have an Ethernet port. Here’s what works well, what to watch for, and what to avoid.

Recommended Adapters

Belkin USB-C to 2.5Gb Ethernet, ~$30
Apple’s current recommended adapter, sold through the Apple Store at $29.95. Driverless on macOS, compact, and reliable. This is the adapter we used for testing. If you want a simple recommendation with no research required, this is it.
Anker PowerExpand USB-C to 2.5GbE, ~$36–40
Slightly more expensive but well-reviewed for long-term reliability. Driverless on macOS Ventura and later. A solid alternative if the Belkin is out of stock.
CalDigit TS4 Dock, Built-in 2.5GbE
The premium option for desk setups. The TS4 includes a 2.5GbE port alongside Thunderbolt pass-through, USB-A, USB-C, DisplayPort, and SD card. Important: the Ethernet port only functions at 2.5GbE speeds when connected via Thunderbolt, not USB-C only, since the Ethernet controller uses PCIe. A one-cable desk solution if you’re already using a dock.
Cable Matters & DriverGenius 2.5GbE
Budget-friendly options that work driverless on macOS. Generally stable, though build quality varies. Check recent reviews for your specific macOS version before purchasing.

What to Avoid

Not all USB-C Ethernet adapters play well with macOS. Two specific issues to watch for:

  • AX88179A chipset adapters: Some budget adapters from Amazon Basics, TP-Link, and others use the ASIX AX88179A chipset. These have reported packet loss and speed negotiation issues on macOS Sequoia. The problem is chipset-specific, not a general “cheap adapters are bad” issue. Check the chipset before buying.
  • ECM-protocol USB Ethernet adapters: Some adapters using the ECM (Ethernet Control Model) USB protocol can trigger kernel panics on Sonoma and later due to changes in Apple’s DriverKit/IOSkywalk networking stack. Again, this is a specific protocol compatibility issue, not a price issue. Stick with adapters explicitly listed as macOS-compatible by the manufacturer.

How to Monitor the Difference Yourself

You don’t need to take our word for it. Here’s how to run your own comparison.

The Terminal Approach

Run a sustained ping test on each connection type:

ping -c 1000 1.1.1.1

Run this once on Wi-Fi, then plug in Ethernet, disconnect Wi-Fi, and run it again. Compare the stddev values in the summary line; that’s a rough proxy for jitter. The Ethernet stddev will be a fraction of the Wi-Fi number.

The limitation: a 1,000-ping test takes about 17 minutes and gives you a single snapshot. Real differences emerge over hours, not minutes, because AWDL and locationd scanning are periodic, so a short test might miss them entirely.

The Monitoring Approach

Don’t want to run Terminal commands? Healthy Network monitors latency, jitter, and packet loss continuously from your menu bar. Set up a monitoring session on Ethernet, then switch to Wi-Fi and compare the sparklines and dashboard side by side. The jitter difference is immediately visible: Ethernet draws a flat line while Wi-Fi shows the characteristic sawtooth.

For the most useful comparison, run each connection for at least a full workday (8+ hours). This captures the full range of AWDL scans, locationd probes, peak-hour ISP congestion, and background macOS activity that short tests miss. Export the data from Healthy Network’s dashboard for your own analysis, or share the diagnostic report with your team to make the case for wired infrastructure.

Frequently Asked Questions

Does USB-C Ethernet add latency compared to Thunderbolt?

For practical purposes, no. Thunderbolt provides a more direct PCIe data path, but for Ethernet speeds up to 2.5GbE the difference is measured in microseconds, completely negligible compared to the milliseconds of network round-trip time. The real concern is bus sharing: if your USB-C Ethernet adapter is plugged into a hub alongside a 4K display, external SSD, and webcam, the shared USB bandwidth can introduce variability. A direct USB-C connection or a Thunderbolt dock with dedicated Ethernet (like the CalDigit TS4) avoids this entirely.

Can I use Wi-Fi and Ethernet at the same time on Mac?

Yes, both interfaces can be active simultaneously. However, macOS does not load-balance between them. It uses a strict priority system based on the service order in System Settings → Network. Internet traffic routes through whichever interface is highest in the list. Drag Ethernet above Wi-Fi to ensure wired takes priority. The lower-priority interface stays active for features like AirDrop, AirPlay, and Apple Watch unlock. Third-party tools like Speedify can bond both connections, but this requires their VPN service.

Is 2.5GbE worth it over Gigabit Ethernet for latency?

For latency alone, no. Both Gigabit and 2.5GbE Ethernet deliver sub-millisecond local latency and identical round-trip times to internet hosts. The link speed affects throughput, not responsiveness. Where 2.5GbE helps is during large transfers on a fast connection: if your ISP delivers over 1 Gbps (increasingly common with fiber), a Gigabit adapter becomes the bottleneck for downloads, and a saturated link increases latency due to queuing. In that scenario, 2.5GbE keeps latency stable during heavy transfers. For most users on connections under 1 Gbps, Gigabit Ethernet is perfectly fine for latency.

Why does my Ethernet latency spike when Wi-Fi is also connected?

This is usually caused by macOS routing traffic over Wi-Fi despite Ethernet being connected. Check your service order: go to System Settings → Network, click the three-dot menu, and select “Set Service Order.” If Wi-Fi is listed above Ethernet, drag Ethernet to the top. Another cause is AWDL. Even when Ethernet handles your internet traffic, the Wi-Fi radio remains active for AirDrop and Handoff. AWDL channel hopping can cause brief system-level latency spikes that affect all network interfaces. Disabling AirDrop in System Settings → General → AirDrop & Handoff eliminates this.

See the difference for yourself

Healthy Network monitors latency, jitter, and packet loss continuously from your menu bar. Run a session on Ethernet and another on Wi-Fi, then compare the dashboard side by side. Export the data to see exactly where your wireless connection introduces spikes, and whether the difference matters for what you do.

Download for Mac
All articles