Measured October 28, 2026
Benchmarks
Every number on this page is read from a published run, with the raw data in the repository. Nothing is estimated or rounded in BoltGet's favour. A difference is claimed only when the run-to-run spreads do not overlap.
Setup
- Machine
- AMD Ryzen 5 5500 (12 logical CPUs), 15.9 GiB RAM, MLD M300 NVMe SSD (500 GB)
- Operating system
- Microsoft Windows 11 Home Single Language 10.0.26200
- Network
- Realtek PCIe GbE Family Controller, 1 Gbps link, wired
- IDM
- v6.43b16 Trial, settings as installed
- BoltGet
- 0.1.0-dev, engine at commit 2266755, release build
Results
Wall-clock seconds, 5 measured runs per configuration after 1 warm-up run.
| Configuration | Median | q1 to q3 | min to max | MiB/s | Connections |
|---|---|---|---|---|---|
| IDM, default | 3.22 s | 3.19 to 3.22 s | 3.18 to 4.20 s | 3.1 | 8 (default) |
| BoltGet, adaptive | 0.22 s | 0.21 to 0.22 s | 0.21 to 0.22 s | 46.3 | 3 |
| BoltGet, 8 connections | 0.41 s | 0.41 to 0.44 s | 0.41 to 0.46 s | 24.1 | 3 |
| Configuration | Median | q1 to q3 | min to max | MiB/s | Connections |
|---|---|---|---|---|---|
| IDM, default | 4.06 s | 4.05 to 4.09 s | 4.04 to 4.10 s | 24.6 | 8 (default) |
| BoltGet, adaptive | 2.17 s | 1.04 to 2.23 s | 1.04 to 2.49 s | 46.1 | 4 |
| BoltGet, 8 connections | 2.19 s | 1.09 to 2.22 s | 1.07 to 2.24 s | 45.7 | 8 |
| Configuration | Median | q1 to q3 | min to max | MiB/s | Connections |
|---|---|---|---|---|---|
| IDM, default | 4.15 s | 3.42 to 4.28 s | 3.32 to 4.68 s | 24.1 | 8 (default) |
| BoltGet, adaptive | 3.96 s | 3.92 to 5.06 s | 3.27 to 5.23 s | 25.2 | 8 |
| BoltGet, 8 connections | 4.11 s | 3.86 to 4.29 s | 3.35 to 6.10 s | 24.3 | 8 |
| Configuration | Median | q1 to q3 | min to max | MiB/s | Connections |
|---|---|---|---|---|---|
| IDM, default | 7.13 s | 7.12 to 7.45 s | 7.09 to 7.47 s | 14.0 | 8 |
| BoltGet, adaptive | 6.78 s | 6.76 to 6.79 s | 6.74 to 6.81 s | 14.7 | 16 |
| BoltGet, 8 connections | 6.66 s | 6.65 to 6.66 s | 6.65 to 6.67 s | 15.0 | 8 |
| Configuration | Median | q1 to q3 | min to max | MiB/s | Connections |
|---|---|---|---|---|---|
| IDM, default | 4.66 s | 3.76 to 17.68 s | 3.42 to 21.31 s | 219.5 | 8 |
| BoltGet, adaptive | 2.47 s | 1.92 to 2.49 s | 1.78 to 2.54 s | 413.9 | 4 |
| BoltGet, 8 connections | 6.34 s | 1.86 to 9.37 s | 1.79 to 17.27 s | 161.6 | 8 |
Connections: for local targets, the most simultaneous connections the server saw; for public targets, BoltGet's own count (IDM's is not visible from outside).
What this run does not show
- One machine, one line, one afternoon. Public results depend on the route and the time of day. They show these hosts from this line, not the general case.
- Only IDM’s defaults. The plan also calls for IDM at 32 connections; that needs a manual change in IDM’s settings and was not run.
- No shaped network. The local server throttles each connection, but added latency, packet loss, server-side connection limits and HTTP/2 are not covered yet.
- External timing for IDM. The time covers everything IDM does between accepting a link and closing the file. There are no throughput curves, CPU or memory figures for either tool.
- Five runs. Enough to see large differences; not enough to rank configurations whose spreads overlap.
Since this run
The adaptive connection controller was changed after this run, for servers that limit the speed of each connection, where the adaptive default had been slower than a fixed 8 connections. The numbers on this page are the run as published and are not edited; the new measurements are in the engine specification until the next full run.
Run it yourself
boltget bench runs the same kind of measurement on your own network: repeated, interleaved runs with the median, interquartile range and min to max. The CLI docs show how.
Raw data
Every run, including warm-ups, with the environment and the target list used:
- Results folderbench/results/2026-10-28-idm
- Full report with method and readingdocs/BENCHMARK.md
- Benchmark harnessbench/