WUProp@Home

WUProp@Home
BOINC Server Default Water image
Project
Status	Active
Category	Distributed computing infrastructure / Benchmarking
Compute	CPU (Non-CPU-Intensive)
Development
Developer	Sébastien (AF>WildWildWest)
Author	Sébastien
Sponsor	Self-hosted; community-supported
Initial release	March 28, 2010  (16 years ago)
Repository	https://wuprop.boinc-af.org/
Software
Operating system	Windows, Linux, macOS, Android, ARM
BOINC statistics
Stats as of	May 25, 2026  (0 years ago)
Performance	~577 GigaFLOPS
Active users	691
Total users	9,465
Active hosts	4,600
Total hosts	250,533
Metadata
Website	https://wuprop.statseb.fr/

WUProp@Home is a volunteer distributed computing project that uses BOINC and needs your help to continuously collect and share performance data about the Berkeley Open Infrastructure for Network Computing ecosystem itself. Unlike most BOINC projects, which direct donated CPU cycles toward science problems like protein folding or pulsar searching, WUProp@Home turns that computing community inward: its lightweight agent silently observes the work units running alongside it on volunteers' computers and reports back statistics about computation time, memory consumption, checkpointing behaviour, and report deadlines for dozens of active BOINC projects. The result is a living, crowd-sourced benchmark database that helps crunchers the world over make informed decisions about which projects to join and how to configure their hardware.

Why WUProp@Home?

BOINC hosts dozens of independent volunteer-computing projects, each running its own applications with their own resource requirements. A volunteer who wants to add a new project faces a difficult question: will it play nicely with their hardware? A project that demands 4 GB of RAM per task may be fine on a modern workstation but ruinous on an older machine. A project whose tasks run for many hours without checkpointing can waste hours of work if the computer is shut down mid-task. A project with a short report deadline may produce invalid results for volunteers in regions with unreliable internet connections.

Before WUProp@Home, the only way to answer these questions was to try each project and observe by hand. There was no centralised, continuously-updated database of real-world workunit performance data drawn from actual volunteer hardware. The BOINC project's own wiki described WUProp@Home's niche concisely: it "collects Work Unit properties of BOINC projects such as computation time, memory and disk space requirements, checkpoint intervals and report times." That gap is exactly what WUProp@Home was designed to fill.

The project is also uniquely self-referential within the BOINC ecosystem: it is a BOINC project about BOINC projects, using volunteer computers to audit the performance landscape of volunteer computing itself.

Goal

The BOINC Manager showing multiple projects running simultaneously. WUProp@Home runs as an NCI task alongside other projects, never stealing CPU time from science work. (Screenshot from an earlier BOINC version.)

The central objective of WUProp@Home is to build and continuously refresh a shared, hardware-specific database of workunit properties for as many active BOINC projects as possible. The specific metrics collected include:

• Computation time -- how long tasks for each project actually take on real volunteer hardware, broken down by CPU and platform.
• Memory requirements -- how much RAM each application consumes at minimum, average, and peak, distinguishing 32-bit and 64-bit environments.
• Disk space requirements -- the download and upload volumes associated with each workunit.
• Checkpoint interval -- how frequently a task saves its progress to disk, which directly governs how much computation is lost if the computer is interrupted.
• Report deadline -- the window within which completed tasks must be returned to the project server, a crucial factor for volunteers with limited or intermittent internet access.

This data is aggregated across all participating hosts and updated continuously. Because it is collected from real hardware rather than from project documentation, it reflects the actual experience of volunteers rather than idealised estimates. The database is publicly queryable: anyone can browse results broken down by CPU model, GPU model, operating system, or BOINC project, giving prospective participants an evidence-based guide to choosing and configuring their project portfolio.

A secondary goal is lowering the barrier to entry for new volunteers. Someone planning a dedicated BOINC build can consult WUProp@Home's CPU and GPU result tables to estimate, for example, how much RAM to install or which projects make best use of a particular graphics card -- before spending any money.

Methods

A non-CPU-intensive (NCI) application

WUProp@Home's most distinctive technical feature is that it is classified as a non-CPU-intensive (NCI) BOINC application. In the BOINC scheduling model, applications labelled NCI use negligible processor time; the BOINC client runs them continuously at normal process priority alongside all other tasks, rather than time-sharing them against science work at idle priority. A single NCI job runs per client at all times; it does not compete with other BOINC projects for CPU slots. This design means that running WUProp@Home imposes essentially no performance cost on other science projects.

The application uses at most approximately 6 MB of RAM, polls running BOINC tasks for their resource consumption every 10 minutes, and saves its collected data to disk via checkpointing every 5 minutes. Completed data packages are returned to the server after 12 hours of uptime, or within 13 hours of the task starting -- whichever comes first. For computers that are not running continuously, credits are awarded proportionally to actual uptime, at a nominal rate of 50 points per day for a machine that runs 24 hours a day (equivalent to 18,250 points per year).

Data collection mechanism

When a volunteer attaches to WUProp@Home, the agent communicates with the local BOINC client to observe which other projects are attached and what workunits are currently running or have recently completed. It records resource measurements and, where the BOINC client exposes them via its RPC interface, retrieves additional properties declared by each project's workunit templates. The server-side pipeline (visible in the project's published server status) runs separate validators and assimilators for each application version (v4 for legacy clients; v5 for BOINC 7.18 and newer, including a dedicated RPC-localhost variant), as well as a data-processing daemon that computes rolling statistics over the last 72 hours of submissions.

As of May 2026, the project monitors workunits from over 40 active BOINC projects simultaneously, including major efforts such as Einstein@Home, MilkyWay@home, PrimeGrid, World Community Grid, LHC@home, NFS@Home, Asteroids@home, and Minecraft@Home, among others.

The BOINC platform logo. WUProp@Home runs on BOINC as a non-CPU-intensive (NCI) project.

Why BOINC?

BOINC was chosen as the platform because WUProp@Home's measurement agent must run alongside other BOINC projects on the same machine in order to observe them. Only within the BOINC ecosystem can a lightweight NCI application monitor co-resident tasks in real time. The open-source BOINC platform, developed at the University of California, Berkeley Space Sciences Laboratory and first released in 2002, also provides WUProp@Home with a ready-made infrastructure for work distribution, result validation, credit accounting, and volunteer management -- all at no cost to the project's solo developer. The combination of BOINC's scheduling flexibility (specifically its NCI slot) and its large global volunteer base makes it the only practical home for a project of this kind.

WUProp@Home's data is therefore inherently crowdsourced and self-improving: the more volunteers run it, the broader the hardware coverage, and the more accurate and representative the resulting statistics become. A result table with hundreds of data points drawn from diverse CPUs is far more useful than one built from a handful of developer machines.

Project team / Sponsors

WUProp@Home was conceived and developed by a volunteer known as Sébastien (forum handle AF>WildWildWest), a member of the French-speaking BOINC community L'Alliance Francophone. The idea originated in discussions on the Alliance Francophone forum, and the project went live on 28 March 2010. It has always been an independent, community-driven effort with no academic institution or corporate sponsor; Sébastien both developed and hosted the project personally.

In early March 2022, Sébastien announced that he lacked the time to manage the project correctly, stopped work generation, and indicated the server would go offline within days. The BOINC community rallied, with volunteers offering to take over hosting. After a few weeks of reflection, Sébastien announced a project revival in May 2022, committing to continue hosting while simplifying and documenting the data-processing pipeline to make a future transition easier. A new application for Android and BOINC 7.18 was released at the same time. The project's source code was also published at that point to facilitate community involvement.

In August 2024, the project underwent a server migration and upgrade -- the primary URL for active participation shifted to wuprop.statseb.fr -- completing by 10 August 2024. The original address at wuprop.boinc-af.org remains accessible but the statseb.fr instance is the current canonical server. As of May 2026, all server daemons are running normally.

The Alliance Francophone (L'AF), while not a formal sponsor, has been the primary community home of the project since its founding, hosting forum discussions and promoting WUProp@Home as a recommended companion project for all BOINC crunchers.BOINC Manager Screenshot.jpg

Statistics

As of 25 May 2026, WUProp@Home reports the following from its server status page:

Metric	Value
Registered users with credit	9,465
Active users (recent credit)	691
Registered hosts with credit	250,533
Active hosts (recent credit)	4,600
Tasks ready to send	11,990
Tasks in progress	3,506
Current GigaFLOPS	~577 GFLOPS

The project's primary application, Data collect version 4, is the workhorse; a version 5 series supports newer BOINC clients (7.18 and above), including Android devices. Average task runtime for the main application is approximately $1.01$ hours, with a range of $0.01$ to $141.13$ hours depending on host behaviour.

Results

The primary output of WUProp@Home is its publicly queryable results database, available at wuprop.statseb.fr/results.py. This database allows any visitor to look up, for a given BOINC project and application:

• The average, minimum, and maximum computation time across all reporting hosts, filterable by CPU family and operating system.
• The average, minimum, and maximum RAM consumption in MB, distinguishing 32-bit and 64-bit environments.
• The checkpointing interval -- how many seconds typically elapse between progress saves -- so volunteers can judge the risk of lost work.
• The report deadline in hours, indicating how promptly tasks must be returned.

These statistics are recomputed continuously from data collected in the most recent 72-hour window, so they remain current as project software versions change. The results represent one of the only sources of systematic, hardware-resolved, real-world BOINC application benchmarks available to the public.

Beyond raw performance data, the Active Projects page provides a live feed of which BOINC projects and applications have reported workunits in the past 24 hours, offering a useful snapshot of what the broader BOINC ecosystem is computing at any given moment. As of 25 May 2026, this list included over 40 distinct projects and more than 80 individual application types.

WUProp@Home has been described by community members as "quite possibly one of the most significant and important projects of BOINC" because of its role in helping volunteers optimise their hardware choices and project configurations.

External links

• WUProp@Home official site
• Server status
• Results database
• Active projects (last 24 hours)
• GPU models tracked
• CPU models tracked
• WUProp@Home on the BOINC wiki
• Alliance Francophone project page (French)