Ever since its introduction as a privacy feature in Apple’s paid-for iCloud+, the Private Relay service has generated complaints that it impairs Internet performance. Last week’s brief look at network utilities available in Sonoma was in part my response to a question about its poor performance on a brand new MacBook Pro M3 Max. This article looks in more detail at whether Private Relay can have this effect.

Apple’s introduction to the Private Relay service recognises that some have complained about its apparent impairment to performance, but warns that popular Internet performance tests can give misleading results: “Private Relay uses a single, secure connection to maintain privacy and performance. This design may affect how throughput is reflected in network speed tests that typically open several simultaneous connections to deliver the highest possible result. While some speed test measurements may appear lower when Private Relay is enabled, your actual browsing experience remains fast and private.”

Reports claiming poor performance seldom make it clear what aspect of performance is impaired, and only rarely support their claim with meaningful measurements. Here I’ll distinguish between two objective measurements:

• Download speed, as the time it takes to download a 1.07 GB Zip file from a remote site, using Safari.
• Responsiveness under working conditions, in round-trips per minute (RPM), a measure that Apple’s experts have recently embodied in an IETF draft, and is implemented in the tests performed by the command tool networkQuality.

To investigate these further, I ran both tests on three different Macs, each running macOS 14.1.2, over wired Ethernet to a single ISP, with and without iCloud+ Private Relay. As you’ll see from the speeds reported, this is a ropy old connection that only enjoys fibre to the cabinet, and runs over copper from there to the local network.

Download speeds

With Private Relay turned off, all three Macs downloaded the test file at speeds of 7.3 MB/s. With Private Relay turned on, one (a Mac Studio M1 Max) achieved the same, but the other two (iMac Pro and MacBook Pro M3 Pro) were slightly slower, at 6.9 MB/s. In practice, that small difference is within the range of speeds delivered by the ISP, and probably wouldn’t be noticed by the user unless downloading more than 10 GB.

Responsiveness

The IETF draft defining and standardising a measure of ‘responsiveness under working conditions’ has been prepared by Apple staff and a professor at the University of Cincinnati, led by Stuart Cheshire, one of Apple’s Distinguished Engineers, Scientists and Technologists (DEST), who pioneered what became Bonjour. Apple has recently explained how to measure this responsiveness using the networkQuality command tool in macOS.

Results of responsiveness measurements differed markedly between the three Macs, although they were all connected to exactly the same network and testing was performed within a period of less than half an hour. Repeated tests using networkQuality confirmed that these differences remain consistent.

The iMac Pro and MacBook Pro M3 Pro both showed marked reductions in responsiveness when tested with the -p option to use Private Relay. The iMac Pro fell from 56 to 4 RPM, while the MacBook Pro fell from 86 to 5 RPM. The latter corresponds to a fall from just under 700 ms to 12 seconds when Private Relay was used.

However, the Mac Studio M1 Max showed no difference, with a responsiveness of 66 RPM without Private Relay, and 67 RPM when networkQuality was run with the -p option to use Private Relay.

Conclusions

iCloud+ Private Relay doesn’t appear to have any significant effect on real-world download speeds, but it may have substantial effects on responsiveness under working conditions. If results from testing using networkQuality are to be believed, then its effect varies between different Macs.

If you suspect that Private Relay may be reducing your Mac’s network responsiveness, then I recommend that you turn Private Relay off temporarily and run networkQuality and networkQuality -p to measure responsiveness in RPM, to see whether that supports your subjective impression. Unfortunately, the IETF draft doesn’t propose any simple solutions to address low responsiveness, but at least you will be better informed to judge whether the additional privacy provided by Private Relay is worth it.

Postscript

I’m delighted to report that you no longer need to run the networkQuality command tool in Terminal to perform these tests. The latest version of Bryan Christianson’s excellent WhatRoute network utility can perform them from the comfort of its pleasant GUI.