9GG NETWORK LAB / LOCAL MODE
Test the path.
Not the plan.
Analyze latency, jitter, packet loss, loaded latency, route quality, and bitrate headroom before you blame the GPU—or buy a faster internet package.
01 Measurement protocol
Build a baseline you can reproduce.
A single best-case result hides route changes, busy-hour congestion, Wi-Fi interference, and queueing. Keep the device, endpoint, and test conditions visible.
- 01Use the actual client
Test from the device and connection you will use. A router-side test cannot reveal client Wi-Fi or decoder conditions.
- 02Target the real region
Measure the service or server region where possible. A nearby generic endpoint can make the route look better than it is.
- 03Compare idle and load
Run a clean baseline, then repeat while the connection is busy. The difference exposes queueing behavior.
- 04Repeat across time
Capture multiple runs at quiet and peak hours. Record the median and the worst repeatable condition.
Record Ethernet or Wi-Fi, VPN state, time, endpoint, active downloads, and whether another person was using the network. Change one variable at a time.
02 Connection analyzer
Turn six measurements into an investigation order.
Enter representative results, not the fastest run. The score is an editorial planning baseline—not a service guarantee.
- Latency reflects distance and route.
- Jitter reflects timing variation.
- Loss can force recovery or visible artifacts.
- Loaded latency reveals queueing under traffic.
03 Loaded latency lab
Fast throughput can still feel slow.
Compare idle latency with latency during download and upload load. Large increases can indicate queueing—often called bufferbloat—somewhere in the path.
Interpretation note: The delta bands used here are editorial troubleshooting guides. Measurement tools, endpoints, access networks, and workloads define loaded latency differently.
04 Route comparator
The closest label is not always the cleanest route.
Compare up to three regions or endpoints using the same device and test window. The lab weights latency, jitter, and loss; throughput is evaluated separately.
05 Bitrate headroom
Plan below the measured ceiling.
A stream competes with protocol overhead, retransmission, bursts, and other traffic. Reserve headroom instead of treating a speed-test peak as sustainable bitrate.
- Cloud services mainly depend on client download.
- Self-hosted remote play depends on host upload.
- Higher resolution and frame rate increase demand.
- Loss and Wi-Fi variability reduce usable headroom.
06 Symptom-first diagnostic
Investigate the failure you can observe.
Select the closest symptom. Use the checks as an order of operations, then preserve telemetry from both the client and host or service.
FIRST INVESTIGATION
Confirm distance, endpoint, route, and queueing.
High latency can come from geography, ISP routing, congestion, a VPN path, Wi-Fi, or measuring a different endpoint than the service uses.- Confirm the actual service or server region.
- Compare Ethernet and client Wi-Fi.
- Test idle and loaded latency at several times.
07 Test worksheet
Leave with evidence, not a vague impression.
Describe the test conditions and copy a compact summary. The report stays in the browser unless you choose to copy it.
NETWORK LAB SUMMARY
Device: Gaming laptop Connection: Wi-Fi Endpoint: Region A Condition: Peak hour, normal household traffic Observation: Repeatable baseline; upload load increases latency.
08 Methodology & privacy
Transparent by design.
The tools use browser-side arithmetic. They do not contact a measurement server, detect your IP, test your line, save results, or rank providers.
- Connection score
- Starts at 100 and applies visible editorial penalties for latency, jitter, loss, insufficient throughput, loaded-latency increase, and access variability.
- Route score
- Weights latency, jitter, and loss to compare the rows entered on this page. It is meaningful only when tests use comparable conditions.
- Loaded latency
- Subtracts idle latency from the values recorded during download and upload load. Negative deltas are shown as zero.
- Bitrate ceiling
- Measured throughput multiplied by the unreserved share. It is a capacity-planning limit, not a codec recommendation.
09 Apply the result
Test the network in context.
A healthy route is only one layer. Continue with the workload, hosting model, device, region, and operating responsibility you actually need.
Cloud Gaming
Evaluate catalog, region, queue, client support, stream quality, and subscription limits.
Explore the model → 02Cloud PC
Match persistence, GPU class, storage, remote protocol, billing, and application compatibility.
Plan a Cloud PC → 03Self-Hosted
Validate host upload, encoder, remote access, power, recovery, and the client decoder.
Build your own path → 04Game Servers
Compare region, tick performance, player load, mods, protection, backups, and operations.
Plan a server →10 FAQ
Read the signal correctly.
Network results are evidence about a route at a moment in time—not universal promises.
Does this page run a real speed test?+
No. It analyzes measurements you enter. A real test requires a measurement endpoint and data transfer to that endpoint; this page intentionally does neither.
Which number matters most for cloud gaming?+
No single number is enough. Latency affects response, jitter affects consistency, loss can trigger recovery, loaded latency reveals queueing, and throughput sets capacity.
Why test against the actual region?+
Distance and routing to a generic nearby server can differ from the path to the application. Use the real service or game-server region whenever it is exposed.
What is loaded latency?+
It is latency measured while traffic is using the connection. The increase above the idle baseline can reveal queueing that a clean speed test misses.
Is Wi-Fi always unsuitable?+
No. A strong, modern, uncongested Wi-Fi link can work well, but it adds a variable radio segment. Compare with Ethernet to isolate that segment.
How many tests should I run?+
Use enough repeated runs to see a stable pattern. Test quiet and peak periods, idle and loaded conditions, and every candidate region using the same setup.
Does a high score guarantee a good experience?+
No. The score is an editorial baseline. Host performance, codecs, client decoding, display processing, application limits, and route changes still matter.
Are my inputs uploaded or stored?+
No. The included JavaScript performs calculations in the browser. The report is copied only when you press the copy button.
NEXT TEST / SAME CONDITIONS
Measure. Change one variable.
Measure again.
Use the lab to create a reproducible baseline, then continue to the full tools hub for cost and platform decisions.