In Ryo Tanaka’s September 26, 2026 benchmark, boxr 0.1.44 had a 134 ms median warm start, compared with 217 ms for Podman 4.9.3, when each ran a cached Alpine container whose command was true. That is a 38% lower median in this particular test—not proof that boxr is generally faster. The measurements came from one restricted VM, with networking disabled and a mismatched privilege setup: boxr ran in its --rootless mode, while Podman ran rootful without a container user namespace. Tanaka’s benchmark report describes the test and its limitations.
What the benchmark measured
The test timed the full command-line invocation for an already-pulled image, ending when the command completed. Both engines ran Alpine and executed true, a command that exits immediately. This captures container startup and exit overhead for a minimal one-shot container; it does not measure image download, application readiness, a first response, networking, mounted volumes, or sustained workload performance.
As an Amazon Associate I earn from qualifying purchases.
The author reports that both sides used the same image bytes and an identical config digest beginning sha256:320994c3. The warm-run sample comprised 20 observations per engine, after five cold runs. Engines were alternated between iterations; the author reports no failed runs among 50 and says no outliers were removed.
Recommended Free Tools
Reported warm-run results
| Statistic | boxr 0.1.44 | Podman 4.9.3 |
|---|---|---|
| Median | 134 ms | 217 ms |
| Mean | 150 ms | 225 ms |
| Minimum | 107 ms | 181 ms |
| Maximum | 262 ms | 310 ms |
| p95, as reported | 207 ms | 288 ms |
These are Tanaka’s reported measurements, not independently validated results. The median difference is about 83 ms, or roughly 38% relative to Podman’s 217 ms median. The spread matters: boxr’s slowest reported warm run was 262 ms, while Podman’s fastest was 181 ms. The distributions overlap, so the medians alone do not show that every boxr run was faster than every Podman run.
#1 Best Overall
- SPACE-SAVING PERFORMANCE FOR HOME & OFFICE – The Dell OptiPlex 3070 Micro delivers dependable computing power in a compact footprint, making it ideal for desks with limited space or clean, minimal workstations.
- RELIABLE INTEL PROCESSING POWER – Equipped with an Intel Core i5 9th Gen Hexa-Core processor (i5-9500), this system offers smooth performance for everyday multitasking, web browsing, and business productivity.
- CONFIGURED FOR EFFICIENCY – Comes with 8GB DDR4 RAM and a 250GB SSD, delivering fast load times, responsive multitasking, and ample storage for files and applications.
- WINDOWS 11 PRO & WIRELESS CONNECTIVITY – Pre-installed with Windows 11 Pro, offering advanced features and security for business or home use. Includes a WiFi and Bluetooth adapter for convenient wireless connectivity.
- VERSATILE & ENERGY-EFFICIENT DESIGN – The ultra-small form factor is ideal for space-conscious users and supports a variety of mounting and placement options. Its low power usage and quiet operation make it perfect for professional environments.
For cold runs, with the image already pulled, Tanaka reports medians of 128 ms for boxr and 212 ms for Podman. “Cold” here does not mean measuring an image download; the report does not establish a general cold-cache result across machines or workloads.
Commands and environment
The reported commands were:
boxr run --rootless --rm --network none library/alpine:latest true
podman run --rm --network none docker.io/library/alpine:latest trueThe VM had 2 vCPU (AMD EPYC 9D64), 7.7 GiB RAM, Ubuntu 24.04.5, and kernel 7.0.0-38-generic. Boxr was the prebuilt Linux x86_64 release binary at v0.1.44. Podman 4.9.3 came from Ubuntu packages and used crun 1.14.1.
Why this is not a matched rootless comparison
Tanaka says the VM was itself a restricted user-namespaced container, preventing a normal unprivileged-user setup. Both engine processes therefore ran as root. Boxr was invoked with --rootless; Podman ran rootful and without a container user namespace. Those modes are not equivalent, so the result does not establish how the engines compare in a regular-user, rootless-versus-rootless test.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsThe benchmark also disabled networking on both sides because Podman’s default netavark networking failed in that environment. The tested commands therefore exclude the cost of setting up a working network. Tanaka separately reports that boxr’s default auto networking, which sets up pasta, added roughly 45 ms in his measurement; that figure belongs to his setup and is not part of the headline comparison.
Storage and cgroup choices also shaped the test. Podman used a tmpfs graphroot to avoid the unavailable /dev/fuse path and use native overlay with idmapped mounts; Tanaka says this likely helped Podman. The test selected --cgroup-manager=cgroupfs because D-Bus was unavailable. These are practical workarounds for that VM, not neutral settings guaranteed to represent either engine’s usual deployment.
What could change the result
Podman’s performance guide identifies configuration and CLI options, OCI runtime, host filesystem, and image as factors that affect performance. It recommends timing with /usr/bin/time -v and using a separate test account for experiments that may alter storage state.
Runtime, storage, and user mappings
The guide describes crun as probably the fastest OCI runtime and gives a typical storage-performance order of native overlayfs, fuse-overlayfs, then vfs. It also notes that rootless Podman with modified UID/GID mappings can incur a first-container creation penalty under particular conditions. That creation-phase cost is distinct from steady-state runtime speed, so a test should say which phase it includes.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #2
- POWERFUL MINI PC WORKSTATION - GMKtec K10 Mini PC is powered by the 13th Gen Intel Core i9-13900HK CPU, built on a 7nm process with 14 cores and 20 threads, reaching a maximum frequency of 5.4 GHz. It features a 24MB Smart Cache and a TDP of 45W, delivering exceptional performance for demanding tasks and multitasking. Upgraded performance compared to Core i5/i7 series.
- 32GB DDR5 RAM & 1TB STORAGE - With 32GB of DDR5 5600 MHz dual-channel RAM (CPU supports up to 5200MHz) and a 1TB PCIe X4 NVMe M.2 2280 SSD, K10 mini PC provides fast and efficient memory handling. It supports up to 3x M.2 2280 PCIE slots and up to 12TB of storage with expansion (3 *4TB), ensuring plenty of room for all your data needs
- COM PORT FOR INDUSTRIAL USE - The COM port enables applications in industrial automation, data acquisition, and embedded systems development, making it perfect for tasks like serial communication with machinery, POS systems, and programmable logic controllers (PLCs)
- QUAD-SCREEN 8K DISPLAY - GMKtec K10 Mini computer offering two HDMI 2.0 (4K @ 60Hz) ports, 1×DisplayPort 1.4 (8K @ 60Hz), 1× Type-C (DP/Data) 10 Gbps/s, this Mini PC supports ultra-high-definition display and multi-screen setups, making it ideal for professional work and business applications.
- 2.5G LAN WiFi6 & BT 5.2 - The Realtek RTL8125BG 2.5G Ethernet port ensures ultra-fast wired connections, while WiFi6 and Bluetooth 5.2 offer reliable and high-speed wireless connectivity for all your devices, ensuring smooth performance across various network-intensive tasks
Networking and readiness
The guide says rootless traffic normally passes through pasta and incurs a performance penalty. Socket activation or another network arrangement changes what is included in startup and connection timing. The guide identifies pasta as the default and only rootless network driver since Podman 6.0.0; that version-specific statement should not be projected onto this benchmark’s Podman 4.9.3 installation.
A process exiting successfully is not the same endpoint as a service becoming useful. A web server test could time process launch, readiness, or the first successful response, and those answer different questions. A comparison that excludes network setup cannot predict port-published service startup.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A cache detail in the report
Tanaka says boxr recognized the cached image under library/alpine:latest, but not under docker.io/library/alpine:latest; the latter spelling triggered repeated pulls and extraction in his test. This is an observation from his setup, not confirmation of a current, universal boxr defect. The commands therefore use different image-name spellings even though the report says the image bytes and config digest were identical.
How to read the numbers—and improve the comparison
The result supports a narrow statement: on Tanaka’s VM, in this particular cached-image true test with the described storage, cgroup, network, and privilege choices, boxr’s reported warm median was lower. It does not answer which engine is faster for a web app, a mounted development environment, a networked service, or a properly matched rootless deployment.
A useful follow-up should keep the variables visible rather than compressing them into a single “startup speed” number:
- Record exact engine and OCI runtime versions, OS and kernel, machine resources, privilege mode, image digest, storage driver and graphroot, cgroup manager, network mode, and full commands.
- Separate image pulls from container startup, and state whether timing ends at process launch, process exit, service readiness, or first response.
- Alternate test order, retain and report every sample, and show a distribution rather than only the best run.
- Where possible, test as a normal unprivileged user, then include realistic workloads as well as the minimal
truecase. - Run cold page-cache tests separately from warm runs; do not conflate either with downloading an image.
Tanaka identifies a non-root rerun, a real workload such as starting a web server and timing its first response, and cold-page-cache runs as useful next tests. Until results under those conditions are available, the reported medians are best treated as a specific benchmark result—not a broad engine ranking.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
