How Our Speed Test Works

Transparency is at the core of N Speed Test. Here's exactly how we measure your internet connection to provide accurate, reliable results.

Test Protocol Overview

Our speed test follows a multi-phase approach to comprehensively measure your internet connection. The test sequence is designed to isolate each metric while minimizing interference between measurements.

Test Phases:

  • Phase 1: Unloaded Latency (Ping) — Baseline measurement with no data transfer
  • Phase 2: Download Speed — HTTP-based download from edge servers
  • Phase 3: Upload Speed — HTTP-based upload to edge servers
  • Phase 4: Loaded Latency — Ping measurement under traffic load

Each phase is calibrated to run for an optimal duration, balancing accuracy with user experience. The entire test typically completes in 20-30 seconds.

Download Speed Measurement

Download speed is measured by initiating HTTP GET requests to our globally distributed edge servers. We use optimally-sized test files and multiple parallel connections to saturate your bandwidth without overwhelming slower connections.

Calculation Method:

Speed (Mbps) = (Total Bits Transferred / Time in Seconds)
Where: 1 Byte = 8 bits, 1 Mbps = 1,000,000 bits per second

We account for network protocol overhead (TCP/IP headers, HTTP headers) to provide a realistic measurement of usable bandwidth. The test automatically adapts to your connection speed, using progressive file sizes to ensure accuracy across the full spectrum from dial-up to gigabit connections.

Upload Speed Measurement

Upload speed is measured by sending data to our edge servers via HTTP POST requests. Similar to the download test, we use multiple parallel streams to fully utilize your upload bandwidth.

Upload testing is inherently more challenging than download testing due to asymmetric internet connections (most ISPs provide lower upload than download speeds) and the need to generate data on the client side. Our implementation uses efficient data generation to minimize CPU overhead on your device.

The upload calculation follows the same principle as download, measuring total bits sent divided by elapsed time. We measure throughput at the application layer to reflect real-world performance for activities like video calls, cloud backups, and file sharing.

Latency, Jitter & Bufferbloat

Latency (Ping)

Latency is the round-trip time (RTT) for a packet to travel from your device to the server and back. We measure both unloaded latency (idle connection) and loaded latency (under download/upload traffic).

RTT (ms) = Time(response_received) - Time(request_sent)

Jitter

Jitter measures the variation in latency over time. High jitter causes inconsistent performance, which is particularly noticeable in real-time applications like gaming and video calls.

Jitter (ms) = Mean Absolute Deviation of RTT samples
Jitter = (1/n) × Σ |RTT(i) - RTT(i-1)|

Bufferbloat

Bufferbloat is the latency increase that occurs when your network is under load. We measure this by comparing unloaded latency (idle) to loaded latency (during download/upload). A large difference indicates excessive buffering in your router or modem.

Bufferbloat Grade = f(Loaded Latency - Unloaded Latency)
A: < 50ms | B: 50-100ms | C: 100-200ms | D: 200-400ms | F: >400ms

Our bufferbloat detection methodology is aligned with industry standards like the Bufferbloat.net project and DSLReports' speed test grading system.

Infrastructure: Global Edge Network

N Speed Test is powered by a globally distributed edge networkconsisting of 300+ data centers in cities worldwide. This architecture ensures that your test is routed to the nearest server, minimizing artificial latency and providing a realistic measurement of your connection.

Why Edge Computing Matters:

  • Low Latency: Testing from a nearby server (often <20ms away) provides accurate baseline measurements
  • High Bandwidth: Our infrastructure can saturate even gigabit+ connections
  • Geographic Accuracy: Your test reflects real-world performance to major internet hubs
  • Reliability: Redundant servers ensure the test works even during peak hours or outages

Each test automatically connects to the optimal Point of Presence (PoP) based on your geographic location and network routing. This is the same infrastructure that powers millions of websites worldwide, ensuring enterprise-grade reliability and performance.

Accuracy & Industry Standards

Our methodology is inspired by industry standards and best practices, including:

  • RFC 6349 — Framework for TCP Throughput Testing
  • ITU-T Y.1540 — IP Packet Transfer Performance Metrics
  • IETF RFC 2544 — Benchmarking Methodology for Network Devices
  • Bufferbloat.net — Community standards for latency under load testing

While we implement our own optimizations for user experience and cross-platform compatibility, the core principles align with these established standards to ensure our results are meaningful and comparable.

Important Note: No speed test is 100% accurate. Factors like WiFi interference, background applications, network congestion, and server load can all affect results. For the most accurate measurement, we recommend running multiple tests at different times and using a wired Ethernet connection when possible.

← Back to Speed Test
How Our Speed Test Works - Methodology | N Speed Test