Shanghai Unicom Caught Throttling Upload Speeds to 5 Mbps

A Gamer's Discovery Leads to a Broader Investigation

What started as a frustratingly slow game update over a WireGuard VPN tunnel has evolved into a detailed community investigation of bandwidth throttling by Shanghai Unicom, one of China's three major telecom carriers. The findings raise familiar questions about ISP transparency — issues that resonate well beyond China's borders.

A user on the Chinese broadband forum 'Kuandai Zhenghouqun' (Broadband Syndrome) documented a methodical investigation after noticing severely degraded upload speeds while remotely connecting to a home network. The case offers a fascinating technical breakdown of how modern ISPs can selectively limit traffic — and how savvy users can detect it.

Systematic Testing Reveals Targeted Throttling

The investigation followed a rigorous troubleshooting methodology that network engineers will find familiar. The user first ruled out local equipment issues by running speed tests against Shanghai Unicom's own Speedtest.net nodes, which returned normal results. This is a common pattern in ISP throttling — carriers often whitelist traffic to known speed-test servers to avoid detection.

The real picture emerged when testing against third-party infrastructure. Using iperf3 against a Tencent Cloud server in Shanghai, the user found upload speeds capped at just 5 Mbps across both TCP and UDP protocols, regardless of whether single or multi-threaded connections were used. The throttle persisted across multiple tests:

• Reconnecting the PPPoE session did not resolve the issue
• Changing the device's MAC address and re-establishing the connection had no effect
• Testing during off-peak hours (late night) showed the same 5 Mbps cap
• Upload tests to servers across multiple providers — Alibaba Cloud in Hong Kong, as well as China Telecom, China Mobile, and Tencent Cloud nodes — all triggered the same limit

HTTPS Traffic and SNI-Based Filtering

Perhaps the most technically interesting aspect of the investigation involved testing whether the throttle applied to encrypted HTTPS traffic and whether it was sensitive to Server Name Indication (SNI) values or port numbers.

Drawing on prior community experience with throttling on Shanghai Telecom's premium international routes, the user set up self-hosted speed test servers using the open-source LibreSpeed platform. By hosting multiple test endpoints on the same physical server with different domain names and ports, the investigation could isolate whether the ISP was making throttling decisions based on SNI headers, destination ports, or raw traffic volume.

Results above 50 Mbps were classified as 'no observable throttling,' while the 5 Mbps cap indicated active restriction. This approach mirrors techniques used by researchers worldwide to detect traffic shaping, including tools like the Electronic Frontier Foundation's former 'Switzerland' project and more recent efforts by groups like M-Lab.

A Global Problem, Not Just a Chinese One

While this particular case involves a Chinese carrier, ISP throttling remains a contentious issue globally. In the United States, the FCC's net neutrality rules — which were reinstated in 2024 — explicitly prohibit such practices, though enforcement and political support remain uncertain. European regulators under the Body of European Regulators for Electronic Communications (BEREC) maintain similar protections.

Major U.S. carriers including AT&T, Verizon, and T-Mobile have all faced accusations of throttling specific types of traffic in the past. The technical methods described in this Shanghai Unicom case — SNI inspection, protocol-based filtering, and speed-test server whitelisting — are capabilities available to ISPs worldwide.

What Users Can Do

The investigation highlights several practical takeaways for technically inclined users anywhere:

• Never rely solely on ISP-hosted speed tests. Always cross-reference with third-party servers and tools like iperf3.
• Test across protocols. If both TCP and UDP are throttled equally, the restriction is likely applied at a broad traffic-shaping level rather than targeting a specific application.
• Monitor SNI and port behavior. Self-hosted test servers with different configurations can reveal whether deep packet inspection is in play.
• Document everything. Systematic evidence is essential for filing complaints with regulators or building community awareness.

Outlook

The Shanghai Unicom case is unlikely to result in regulatory action within China, where net neutrality protections are minimal. However, the detailed technical documentation produced by the broadband community serves as a valuable resource for researchers and advocates globally. As encrypted DNS and ECH (Encrypted Client Hello) adoption grows, ISPs may find SNI-based throttling increasingly difficult — potentially shifting the cat-and-mouse game to new battlegrounds like traffic fingerprinting and volumetric analysis.

For now, the incident is a reminder: the pipe your ISP sells you may not always deliver what you are paying for.