Measure your recording capacity
Table of Contents
There is no fixed number of streams that every CaptureGem machine can record. The source sites, bitrates, network path, storage drives, and post-recording processing determine the result.
Set up a repeatable test
Write down the CaptureGem version, source sites, selected quality, network connection, and paths for temporary and finished files. Use your normal storage and network setup, including Proxy Mode if you normally need it. Record the free space on each drive. The file locations guide helps you find the relevant folders.
Choose a small set of supported streams that you are permitted to record. Keep the source mix and quality as consistent as possible between trials. A live broadcast can end or change quality during a test, so note those changes before comparing results.
Test capture and processing together
Start with one active stream. A 15-minute trial is a useful first check. Record download throughput, CPU use, available memory, and free space at the start and end. Watch for brief spikes as well as averages. Task Manager on Windows or Activity Monitor on macOS can help you observe the recorder and its background processes.
Then end the recording and let processing finish. Watch how long that takes and whether the temporary drive approaches full. Capture and finalization can overlap in everyday use, so include a trial where some recordings finish while others continue.
If the first result is healthy, add a few streams and repeat. When you approach a limit, increase by one at a time. Once a promising count passes the short trials, run it for your usual recording duration. A short trial can miss slow storage growth or a processing queue that builds over several hours.
Keep a small results table
Use one row per trial and keep detailed notes alongside it. These rows show what to record, with blanks for your own measurements.
| Trial | Capture measurements | Finished-file checks |
|---|---|---|
| One stream | Throughput, CPU, memory, free space | Duration, picture, audio |
| Higher count | Same measurements plus any errors | Every output and processing time |
| Usual duration | Resource peaks and queue growth | All outputs and recovery needs |
Record the number of streams actually capturing media. A configured model that stays offline contributes no recording load during that trial.
Inspect every finished recording
Open each file and check the beginning, middle, and end. Seek to several points, listen where the source had audio, and compare the duration with the time the broadcast was actually available. Note freezes, missing sections, early endings, and any errors. Keep the associated source files while investigating a failure.
If you already use FFmpeg tools, this optional command reports the file’s duration in seconds and its audio and video stream types. Replace the filename with your finished recording’s path.
ffprobe -v error \
-show_entries format=duration:stream=codec_type \
-of json "recording.mp4"
The command reads media metadata. Continue the playback checks to assess the actual picture and sound. The available reporting options are documented in the official ffprobe manual.
Use the failure pattern to choose the next test
- Download stalls while CPU and disks have room. Try fewer streams with the same source mix and check the network path. A provider interruption can affect a single recording even when other streams remain healthy.
- Free space keeps falling or writes stall. Check both the temporary and finished-file locations. If they share one physical drive, they share its capacity and write workload.
- The processing queue grows across trials. Allow it to clear, then repeat at a lower count. Include overlapping capture and processing before accepting the result.
- Only one source fails repeatedly. Repeat that source alone before treating the failure as a machine-wide capacity limit.
A GPU encoder meter does not measure CaptureGem’s ordinary stream capture capacity. See resource scaling for the different workloads and bandwidth and storage limits for a sizing example.
Choose a working count below the point where errors or growing queues first appear. Keep headroom for source bitrate changes and other activity on the computer, and repeat the test after a significant change to storage, networking, or recording settings.
Loading comments...