Why Free Mobile speeds vary so much by testing

Our first 5G barometer focused in Paris and released on September 13th provoked a lot of reactions from Free Mobile subscribers. Some were offended by the poor results obtained by their operator. In fact, its performance was lower than that of Bouygues Telecom, Orange and SFR, in both 4G and 5G. 5G didn’t even seem to bring any benefit compared to 4G when the measurement campaign was conducted in July.

A few days later, the RNC Mobile project published its Free Mobile observatory in the capital with much more favorable results.

It will be observed previously that it is impossible to compare two studies carried out in different places, conditions, time periods and with different perimeters.

Still, orders of magnitude are challenging. Most users who have used the RNC Mobile application have achieved speeds of several hundred Mbit / s. Far from the average 36.4 Mbit / s found by QoSi. Why so much discrepancy? Here are some explanations.

Different speed test protocols

As a reminder, our study was carried out by the company QoSi, a renowned specialist in the measurement of telecommunications networks that acts as an expert with Arcep. Its speed tests are performed on a single connection (single-thread), unlike those of RNC Mobile which are multiconnected (multi-threaded) like other speed test applications, such as nPerf.

The only thread is the standard adopted by the telecommunications police for their mobile service quality campaigns. That is why we use the same protocol.QoSi told us when we published our Parisian 5G barometer.

“A single thread is a single connection stream. For example, you’re on your smartphone and watching a live video. Multiple thread represents multiple connections in parallel. You’re watching a video online, playing music and doing a web search at the same time “a spokesman for Arcep reminds us.

Different representations

However, Free Mobile’s performance is consistently lagging behind its competitors when testing is on a single thread. This is especially the case every time Arcep publishes its annual 4G survey. Hence the misunderstanding of Free Mobile subscribers who get much better results when they do a speed test on their smartphone. The operator only highlights its multi-threaded results.

If the telecommunications police chose a single thread, would this completely disqualify the multi-wire speed tests? Not at all. “It’s not illegitimate or incorrect multi-line. You just have to know what it represents. Simplifying the stroke a bit, multi-threading allows you to measure the size of the pipe assigned to you by the operator, anything that can exceed the maximum. But when you download an app, your effective performance will always be what you get in a single thread. “further clarifies the telecommunications police.

In short, Free Mobile probably makes it easier to launch multiple connections simultaneously than its competitors. But the connection speed obtained with a single connection will be lower than average.

See also the video:

As close as possible to the uses of smartphones

For the authority, the measures in the mobile must continue to be a single thread. “We have no doubt that the single thread is the most representative of what people are doing today on their smartphones. The debate between the two has been revived with 5G, with some operators believing it will favor multi-threading It is worth discussing that on a day when multi-threading is representative of what people are doing on their smartphones, we will be able to evolve measures in this direction.concludes Arcep.

However, this would not explain all the poor performance of the Free Mobile network. It seems that the results also vary depending on the technical characteristics of the servers. More specifically, they would be sensitive to TCP congestion avoidance algorithms. The speeds would be lower with Cubic, installed on most servers, than with another protocol, called BBR. The phenomenon is already well documented in terms of Free’s fixed network, as evidenced by the forums on the site La Fibre.info and a note already published by Arcep. This could reveal a packet loss issue. But why? The mystery remains intact.

Free did not want to answer our questions about the results of their mobile speed when we were contacted.

Leave a Comment