What Kinds of Pre-Release Game Content Are in The Data Dungeon Archive?

Updated

The Data Dungeon archive holds pre-release video-game material across two top-level collections: Games (568 entries) and Software (3 entries). At the time of the site snapshot, the archive reported 16,444 files totaling 1674.79 GiB. Content is organized two ways: by file type (distributions, artwork, screenshots, videos, sounds, archives, documents, executables) and by game-level labels (prototypes, demos, source code, patches, and cancelled titles). The site describes itself as dedicated to preserving pre-release content such as screenshots, videos, and prototypes, and notes that much of it is still under construction.

File-type categories and what they contain

The archive's own counters break the collection down as follows:

Category Count What it typically means here
Distributions 982 Packaged builds or release-style distributions tied to a game entry
Artwork 4,147 Concept art, promotional art, and similar visual material
Screenshots 5,654 In-game and pre-release screen captures
Videos 676 Gameplay or promotional footage
Other 4,787 Material that doesn't fit the named categories
Unknown 342 Unclassified files
Images 12,995 The broad image total (includes artwork and screenshots)
Videos 614 A second video tally reported alongside the file-type breakdown
Sounds 172 Audio assets
Archives 1,724 Compressed or bundled file sets
Documents 423 Text and documentation files
Executables 174 Runnable binaries

Two practical notes: the site reports video counts in two places (676 and 614), so treat those as separate tallies rather than one figure. And the large "Other" and "Unknown" buckets mean a meaningful share of the archive isn't neatly labeled, so browsing by game entry is often more reliable than browsing by type.

Game-level labels: how entries are tagged

Each game row carries one or more labels describing what kind of pre-release material exists. The labels visible in the archive include:

  • Prototypes — early or unfinished builds. Examples: MDK2 Prototypes Demos, Black9 (Black 9) Prototypes, Minimum Prototypes.
  • Demos — playable sample builds. Examples: Grim Fandango Demos, Nocturne Demos, Driver Demos.
  • Source code — released or leaked code. Examples: Turok: Dinosaur Hunter Source Code Demos, Unreal Tournament 2003 Source Code Prototypes Demos, Heavenly Sword Source Code.
  • Patches — update files, as with Codename Eagle Demos Patches.
  • Cancelled — titles that never shipped. Examples: Duality, TORN, Maddrax, Sonic Xtreme, Stars of Blood, Zenith (Climber), Arena AD.

Labels combine, so a single entry can be both a prototype and a demo, or carry source code alongside demos. That combination is the fastest way to tell what a given game actually offers before opening it.

Using release year and last-updated fields

Each row includes a Release Year and a Last Updated date, and these serve different purposes:

  • Release Year places the material in context — a 1997 source-code entry (Turok) versus a 2014 prototype (Minimum) versus a 2009 entry (Final Fantasy XIII). Cancelled titles often show "Cancelled" instead of a year, and some show "N/A" (e.g., Sonic Xtreme).
  • Last Updated tells you how fresh the archive entry is. Most rows in the snapshot show 2026-05-03, with a few later dates such as 2026-05-25 (TORN), 2026-07-26 (Halo 2), and 2026-08-15 (Epic Mickey 2: The Power of Two). A recent update date is a reasonable signal that an entry was added to or revised recently, not that the underlying game content is new.

Per-game counts tell you what's actually there

The Games table also lists counts for Distributions, Images, and Videos per title, which is the most concrete way to judge an entry. For example, Black9 shows 1 distribution and 259 images; Duality shows 277 images and no distributions; MDK2 shows 2 distributions and 11 images; TORN shows 29 images and 1 video. A title with images but zero distributions is an image/art archive rather than a playable build, so check these columns before assuming a download exists.

Practical starting points

  • Looking for playable early builds: filter mentally for entries labeled Prototypes or Demos with a nonzero Distributions count.
  • Looking for code: scan for Source Code labels (Turok, Unreal Tournament 2003, Army Men, Heavenly Sword, Unreal Championship 2).
  • Researching lost games: the Cancelled label plus image counts is where most of that material sits.
  • Browsing broadly: the site's navigation includes Archive, Library, Blog, Donate, and rsync, and it points to a Discord server for feedback and a contact page for legal concerns. The rsync mention suggests bulk mirroring is part of the site's model, though the snapshot doesn't document the exact command or access terms.

Because the site states it is still under construction, counts and entries should be treated as a moving snapshot rather than a fixed catalog.