# Y2KDASH

> An ambient, continuous-monitoring internet speed test that measures download, upload, idle latency, loaded latency, jitter, packet loss, and bufferbloat against Cloudflare's public endpoints. Designed to run on a second monitor as an always-on network dashboard.

Full-text variant with all pillars inlined for LLM ingestion: https://y2kdash.com/llms-full.txt

Y2KDASH is an open-access speed-test dashboard at https://y2kdash.com. Unlike Ookla, Fast.com, or speed.cloudflare.com (which are one-shot tests that run only when triggered), Y2KDASH samples every ~60 seconds in the background and plots results on rolling time-series charts. It measures loaded-latency (the key bufferbloat metric) by running parallel ping probes during saturated downloads.

The dashboard exposes an AIM (Ambient Internet Measurement) composite score — a single A-through-F letter derived from 8 component metrics: download throughput, upload throughput, idle latency, loaded latency, jitter, packet loss, protocol efficiency, and connection stability.

## Key facts

- Backend: Cloudflare speed.cloudflare.com endpoints (/__down, /__up, /meta) — no custom server
- Continuous sampling: every 60 seconds once consented
- Measurements are client-side; no user data is logged server-side
- Bufferbloat detection: loaded-latency probes run during saturated downloads, using median-per-scan smoothing
- Packet loss: 20 parallel probes per burst with 30ms stagger, 2s timeout
- Map origin reposition: updates from /meta colo on every sample (covers ~75 Cloudflare edge POPs)

## Important pages

- [Y2KDASH home](https://y2kdash.com/): the live ambient dashboard with hero speed readout, charts, AIM score, packet-loss panel, and connection map
- [The Speed Test Lie](https://y2kdash.com/the-speed-test-lie.html): ~1900-word explainer on why one-shot speed tests under-report reality — bufferbloat, TCP slow-start, loaded latency, and what a continuous monitor catches that Ookla misses
- [How to Actually Fix Bufferbloat](https://y2kdash.com/how-to-fix-bufferbloat.html): ~2500-word router-by-router fix guide for OpenWRT, ASUS Merlin, pfSense, Ubiquiti EdgeRouter, TP-Link, Netgear, and ISP-supplied gateways. Covers CAKE queue discipline setup, shaping ratio, and verification
- [Networking Glossary](https://y2kdash.com/glossary.html): canonical definitions of the 9 terms that determine real-world connection quality (bufferbloat, loaded latency, unloaded latency, jitter, packet loss, TCP slow-start, CAKE, SQM, FQ-Codel)
- [Y2KDASH vs Waveform Bufferbloat Test](https://y2kdash.com/compare/waveform.html): honest side-by-side comparison of the two remaining free bufferbloat tests after DSLReports went offline in 2025
- [Y2KDASH vs Speedtest by Ookla](https://y2kdash.com/compare/ookla.html): why Ookla and Y2KDASH answer different questions (throughput vs latency under load) and when to use each

## Key definitions

- Bufferbloat: latency that only appears under load, caused by oversized packet buffers in routers and modems. Fixed by enabling queue management (CAKE, FQ-Codel)
- Loaded latency: round-trip time during a saturated download. Idle ping + bufferbloat delta
- AIM score: Y2KDASH's composite A-F grade combining 8 weighted components
- CAKE: a modern Linux queue discipline that eliminates bufferbloat by combining shaping, fair queuing, and AQM in one policy

## Sibling sites (same operator, privacy benchmark network)

- [piholekiller.com](https://piholekiller.com/): 103-test ad-blocker audit covering ads, trackers, fingerprinters, DoH bypass, and malware
- [adbloat.com](https://adbloat.com/): ad-tech overhead benchmark — measures bytes blocked, handshakes prevented, tracker payload weight
- [snitchtest.com](https://snitchtest.com/): browser fingerprinting + leak audit (WebRTC, canvas, audio, TLS, fonts)
- [Bufferbloat & Latency Reference 2026](https://y2kdash.com/bufferbloat-latency-reference-2026): Threshold table for good ping (<30ms), bufferbloat (<30ms=A), jitter (<5ms), packet loss (0%). What actually decides if a connection feels fast.