Jim Allchin demonstrating preliminary WDDM in pre-release Windows Vista


Recommended Posts

While searching through local files I discovered a video of former Windows chief Jim Allchin demonstrating a preliminary version of the Windows Display Driver Model (WDDM), then known as the "Longhorn" Display Driver Model (LDDM). Recall that WDDM was introduced in Windows Vista, and brought with it numerous benefits including display graphics fault-tolerance, the ability to update graphics drivers without restarting the system, and GPU virtualization and scheduling. Note that the latter two capabilities are required by the Desktop Windows Manager, and were so important during the development of Windows Vista that Microsoft's Greg Schechter stated that one "cannot overstate the importance of virtualization of the GPU in the same way the CPU has been virtualized."

Video link: http://1drv.ms/1WnaqHe

This video features Jim Allchin demonstrating XPDM on one machine and WDDM on another machine. Allchin opens several concurrent GPU-intensive applications on each machine, only for the XPDM machine to stutter and eventually run out of video memory. However, the WDDM machine has enough memory—in spite of eventually having additional applications open—and is also able to operate at better performance.

LDDM.thumb.png.8dc22fb998692e1e613c9c008
Image courtesy: Paul Thurrott

This video was originally uploaded by former Neowin writer Tom Warren, also known as "creamhackered."

WDDM's flaws were on integrated GPU chipsets and extremely low-end chipsets (pretty much any non-ATI or non-nVidia graphics chipset); with Intel and S3 being the biggest offenders (both were also quite commonplace in low-end XP boxes in the consumer-stable and corporate-stable areas) .  nVidia discrete GPUs were fine; the same applied to ATI (AMD hadn't acquired them yet); the later forays into integrated graphics (by both companies) were also fine; however, it DID take Intel longer to fix those woes.  In short, the issues were with specific companies and chipsets - not integrated graphics as a whole.

  • Like 2

Yep, despite the fact that Vista is considered a market failure, it was great they continued to build on it for Win7, rather than going back to base, nice little improvements like this lived on. 

Thanks for sharing the blast from the past

  • 1 month later...

For those interested, the earliest known references to the WDDM can be found here. This set of slides is from WinHEC 2002 and is so early that it does not even refer to the driver model as its preliminary name "LDDM," but instead refers to it as a set of significant improvements over XPDM.
Longhorn_Display_Features.thumb.png.a515
Several features are listed in these slides, such as display dimming, driver upgrades without reboots, and timeout detection and recovery. While this list of features does not go into the level of detail that one would expect when compared with later documentation, it is nice to have insight into Microsoft's thought process during this time period. Note that WinHEC 2002 is also the source for the earliest known references to the Desktop Window Manager.

WHy bring this thread back from the dead?

Yep, old technology, cool, move on...

Technology for it's own sake is a nice hobby but is not what most people care about (unless your in the tech industry).

Let's say you by a refrigerator. 4 years from now your refrigerator is still working fine. Is that refrigerator  "old technology" and you should go out and buy a new one? 

No. Why? Because the refrigerator you have is doing everything that you want it to. Probably will do virtually everything a new one will.

Most people don't upgrade because what they already have does everything they want it to. It's really not because their "stuck in their ways".

WHy bring this thread back from the dead?

Yep, old technology, cool, move on...

I wouldn't call it old technology, it's still the current driver model in Windows 10 (Upgraded of course)

WDDM is the reason we can do crazy things like run a Direct3D app and a OpenGL app side by side, and gives us stuff like Direct2D (That can share surfaces with classic GDI rendering codepaths, etc.)

This topic is now closed to further replies.
  • Posts

    • I imagine that was a review or something? My reviews mostly contain a lot of images and galleries, but these are all webp too, but yeah it all adds up on the page load. Would help if you were more helpful with your critique instead of bitching and moaning like a Karen 😂 Because then we might be able to fix it for you.
    • If Valve refused to let them make the case, I wonder if they've already partnered with someone else to do it? The fact that they didn't seek permission/licence before diving straight in is incredible though
    • OpenClaw now has native mobile apps on iOS and Android by Karthik Mudaliar OpenClaw, the viral open-source personal AI agent, now has its own mobile app, available on both Android and iOS. Users can pair the app with an existing OpenClaw gateway and can start using new mobile-native features that are now available on the app. The app supports all the existing features you'd already have seen on OpenClaw's TUI, as well as some more, such as real-time and background Talk mode, action approvals, sharing from iOS, and optional access to device capabilities such as camera, screen, location, photos, contacts, calendar, and reminders. These features are available on both the Android and iOS versions of the app. What's important with these apps is that they don't run OpenClaw on your phone, but are actually just companion apps that require a running OpenClaw Gateway on an existing device, on macOS, Linux, or Windows via WSL2. To pair the app with your existing OpenClaw gateway, users need to run the command "/pair qr" on the TUI or existing chat interface, which brings up a QR code. Users can then scan this QR code to pair it up with the mobile app. There's also an option to manually pair the app by entering the host and a port. Previously, OpenClaw had been available on phones via WhatsApp, Telegram, Slack, Discord, Microsoft Teams, Matrix, and others. Now, with a native mobile app, the interface is much cleaner and more focused on just the OpenClaw, of course, with the added support for camera, screen, location, and more. It's important to note that OpenClaw comes with its own security warnings. There's always a chance of prompt injection with these tools, so users are recommended to double-check authentication, tool policy, sandboxing, and execution approvals rather than prompts alone. For users well-versed with the AI harness, a native mobile app makes it easier to approve an automation, share a link, use voice, or let an agent react to phone-side context.
    • Google pitches Spanner as one database for all AI agents with these new featues by Karthik Mudaliar Google Cloud is introducing new features within Spanner, its distributed database, as a place where enterprises should keep their data, using which AI agents could make smarter and better decisions. In a detailed blog post, Google highlighted quite a few features coming to Spanner, including relational data, graph relationships, vector search, key-value access, full-text search, and operational analytics together in one database architecture. Google says that today's systems aren't well-made for AI agents. There could be data that is present in one system, search indexes in another, embeddings in a vector database, and relationship data in a graph database. This fragmentation isn't great for AI agents to do their jobs because they don't have access to all of this data in one place. This is where Google is positioning Spanner as a solution. Spanner is already a globally distributed relational database with strong consistency, and Google wants its customers to see it as a broader data layer for AI applications. The company introduced something called Spanner Graph, along with integrated vector search, full-text search, a Cassandra-compatible key-value endpoint, and a columnar engine for analytical queries on operational data. Google also added that its ScaNN-powered vector search can support indexes with more than 10 billion vectors, while the columnar engine can make some analytical scans up to 200 times faster. All of this isn't just exclusive to the Google Cloud Platform, and there's support for multi-cloud as well. This comes via Spanner Omni, which Google says is a downloadable, containerized version of Spanner that can run on Kubernetes and in environments outside Google Cloud, including Microsoft Azure and AWS, and even on-premises infrastructure as well as edge deployments. Google says that customers who are interested in the full-featured edition should contact the company, and there's no word on commercial availability or separate pricing. Those interested can read the full blog by Google Cloud, which details these features individually.
  • Recent Achievements

    • First Post
      rosiecharles earned a badge
      First Post
    • Reacting Well
      Juan Dela earned a badge
      Reacting Well
    • Week One Done
      Collagen Project earned a badge
      Week One Done
    • Reacting Well
      Wakeen1966 earned a badge
      Reacting Well
    • Rookie
      Almohandis went up a rank
      Rookie
  • Popular Contributors

    1. 1
      +primortal
      515
    2. 2
      +Edouard
      273
    3. 3
      PsYcHoKiLLa
      143
    4. 4
      Steven P.
      99
    5. 5
      macoman
      54
  • Tell a friend

    Love Neowin? Tell a friend!