MediaTek Genio upstream update: from HDMI 2.0 to booting on U-Boot and TF-A. Julien Massot August 12, 2026 Reading time: 6 minutes Since its last roundup, Collabora has continued working side by side with MediaTek to expand upstream support across both current and upcoming Genio IoT platforms. And there is a lot to talk about. Native HDMI 2.0 output is now available for several Genio boards in the upstream Linux kernel, support for the new Genio 520 and Genio 720 platforms is progressing across the kernel, and its work on the upcoming Genio Pro 5100 now reaches beyond Linux into U-Boot and Trusted Firmware-A. Together, these developments bring the Genio IoT family closer to an out-of-the-box embedded Linux experience built around widely used upstream projects: Trusted Firmware-A and U-Boot for key parts of the boot flow, the mainline Linux kernel, and Debian as the userspace distribution. This makes MediaTek Genio IoT boards a strong choice for developers looking for a solid, secure, and maintainable embedded platform built around upstream open source software. The result is a growing family of boards that embedded developers can evaluate, integrate, and maintain using a familiar upstream Linux stack, without having to rebuild their product around large sets of platform-specific kernel patches. Let's take a deeper dive into the latest milestones! MediaTek Genio 510, 700, and 1200 get HDMI 2.0. One of the biggest updates for the MediaTek Genio IoT lineup is the introduction of native HDMI 2.0 support in the upstream Linux kernel. This applies not only to MediaTek's own evaluation kits, including the Genio 510 EVK, Genio 700 EVK, and Genio 1200 EVK, but also to boards from MediaTek partners such as the Radxa NIO-12L, Grinn GenioBoard, Ezurio Tungsten, and others. In practical terms, you can now connect an HDMI 2.0 monitor or television and enjoy high-resolution video, high refresh rates, and audio output using a vanilla Linux kernel. The Genio 510, Genio 700, and Genio 1200 platforms are based on the MT8370, MT8390, and MT8395 SoCs. After an extensive review process with the upstream community, the MediaTek HDMI 2.0 controller driver was merged into Linux 6.19. The corresponding device-tree changes enabling HDMI display and audio output on the Genio boards followed in Linux 7.0. Alongside HDMI support, Collabora has continued contributing bug fixes and device-tree improvements to keep these platforms working as the upstream kernel evolves. Genio 520 and Genio 720 upstream support moves forward. Upstream support for the Genio 520 and Genio 720 EVKs, based respectively on the MT8371 and MT8391 SoCs, has made significant progress over the past months through the joint work of Collabora and engineers from MediaTek's IoT and Kompanio teams. Several of the building blocks needed to get these platforms running are now available in the mainline Linux kernel, including support for eMMC and SD storage, I[2]C, eFuse access, and the watchdog. MediaTek engineers have also contributed the pin-controller and power-domain support, two essential parts of the initial SoC bring-up. The next set of features is already moving through upstream review. This includes the system and basic clock controllers, Ethernet GMAC, display and DSI, PCI Express, and UART support, while work on UFS and USB is planned for the second half of the year. On the board side, the first device-tree patches for the Genio 520 EVK and Genio 720 EVK have also been submitted. They already allow both boards to boot a headless Linux system with serial and connectivity support. Collabora has also been working on the Airoha AN8801 Ethernet PHY driver used by both EVKs, preparing the way for upstream Ethernet support. Display, GPU, and multimedia enablement are under active development, so there is still plenty more to come for these two platforms. Genio Pro 5100: going beyond the Linux kernel. Another major part of its recent work is the upcoming Genio Pro 5100 EVK, powered by the MT8894 SoC. This project marks an important first for its collaboration with MediaTek: its platform enablement work now covers not only the Linux kernel, but also U-Boot and Trusted Firmware-A. For this platform, U-Boot is used both for the initial boot stage through its Secondary Program Loader, or SPL, and as the main bootloader. Trusted Firmware-A provides the platform runtime services. Upstreaming the U-Boot and Trusted Firmware-A support brings a larger and more visible part of the Genio Pro 5100 boot software into established open source projects. Bringing Genio Pro 5100 support to U-Boot. Getting a new SoC to boot with U-Boot is much more than a matter of adding a board configuration. For the Genio Pro 5100, Collabora first improved the existing UFS and USB controller drivers, then added the missing low-level support needed during the early boot stages, including pin control, clock management, SPMI, and the platform's power-management ICs. With those pieces in place, the board can now boot directly from its internal UFS storage, while also supporting microSD cards, USB flash drives, and USB-to-Ethernet adapters. Collabora also added support for reflashing the internal UFS storage over USB using Fastboot. This makes the platform much easier to bring up, install, recover, and iterate on during development, without requiring an already working operating system on the board. Linux support built on the MT8196 foundation. On the Linux side, Collabora were able to reuse much of the existing support for the closely related MT8196 SoC. Most of the initial work therefore focused on adding the device-tree descriptions for the MT8894 SoC and the Genio Pro 5100 EVK. Some changes were still needed in the pin-controller, clock, and power-domain drivers to support the Ethernet and PCI Express controllers. With those pieces in place, both wired and wireless networking are working on the board. The Linux changes are currently the first public part of this work. Collabora also plan to publish and upstream the corresponding U-Boot and Trusted Firmware-A improvements, increasing the amount of upstream open source software used in the platform's boot flow. Follow the progress. You can follow the upstream status of the different MediaTek Genio platforms through the following resources: These pages contain board and SoC mappings, feature status, development resources, kernel trees, images, and links to the relevant upstream work. What's next. There is still plenty happening across the Genio IoT family. For the Genio 520 and Genio 720, work is continuing on connectivity, display, GPU, storage, USB, and multimedia support. For the Genio Pro 5100, the next major step is publishing and upstreaming the U-Boot and Trusted Firmware-A work. Together with MediaTek and the upstream community, Collabora will keep expanding upstream support across the Genio software stack. Stay tuned for more updates! 18/08/2026 Linux kernel 7.2 adds broad core, Rust, and hardware-support updates, with Collabora contributing 102 patches across graphics, media, virtualization,... 12/08/2026 Collabora and MediaTek continue expanding upstream support across the Genio IoT family, with HDMI 2.0 now enabled for existing boards, Genio... 30/07/2026 Kraid, the new compiler for the Panfrost driver stack, now passes all Vulkan CTS compute shader tests. This marks a major milestone, with...

Valve backs RADV Vulkan driver port to Windows. Collabora is bringing Mesa's open-source RADV Vulkan driver to Windows, with the experimental port already capable of launching Counter-Strike 2. Collabora is working on an experimental Windows port of RADV, Mesa's open-source Vulkan driver for AMD graphics hardware, with Valve sponsoring the latest phase as the effort has reached a milestone, with Counter-Strike 2 now able to launch on Windows using RADV. RADV has become the de facto Vulkan driver for AMD GPUs on Linux, where it forms a key part of the Mesa graphics stack. Windows users, however, still rely on AMD's proprietary graphics driver, making the possibility of running RADV natively on the platform particularly significant. The project builds on earlier work by graphics developer Faith Ekstrand, who demonstrated at XDC 2024 that RADV could communicate with AMD's proprietary Windows kernel-mode driver through the Windows Display Driver Model 2 interface. Her initial implementation reverse-engineered enough of the private communication between AMD's user-space and kernel drivers to allocate memory, create queues, submit commands, and render a basic 3D scene. Collabora has expanded that PoC into a more capable native Windows implementation. The new work removes the dependency on WSL, reduces hard-coded hardware assumptions, improves command-stream handling and synchronization, and adds support for sparse bindings, tessellation, task shaders, and dynamic GPU capability detection. Although the driver does not yet pass the full Vulkan conformance test suite, its success rate has improved considerably. More importantly, the port can launch Counter-Strike 2 by starting the game with the -vulkan command-line argument. However, that achievement does not mean RADV is ready to replace AMD's official Windows Vulkan driver. The current implementation remains experimental, and several major technical obstacles must be addressed before it could become stable enough for general use. Collabora says that a production-ready implementation would require either a stable and documented interface to AMD's kernel driver or a compatibility library supplied by AMD that hides changes to the private communication format. Presentation performance remains another unfinished area. RADV currently uses a slower CPU-based presentation path on Windows instead of native DXGI swapchains. Supporting DXGI properly would require importing images from D3D12. Achieving zero-copy presentation would need additional cooperation from AMD and probably Microsoft. According to Collabora, zero-copy swaps could deliver performance improvements of up to three times in applications not limited by GPU processing power. However, that is still a future optimization rather than a capability of the current driver. Despite these limitations, the project demonstrates that Mesa's open-source AMD Vulkan driver can run beyond Linux and support genuine Windows gaming workloads. The Windows RADV work is currently available in a development branch rather than Mesa's main codebase, and Collabora has not announced a release schedule for an end-user driver. For now, it is still considered a technical experiment. Tell others: Bobby borisov. Bobby, an editor-in-chief at Linuxiac, is a Linux professional with over 20 years of experience. With a strong focus on Linux and open-source software, he has worked as a Senior Linux System Administrator, Software Developer, and DevOps Engineer for small and large multinational companies.

A framework to share analytics data in GStreamer. Daniel Morin February 13, 2024 Reading time: 4 minutes GStreamer has long been the best framework to build pipelines to handle video streams, and in particular, live ones. It's no coincidence that it has been adopted widely by engineers wishing to build video analytics pipelines. What do Collabora Ltd mean by analytics? Within computers, Collabora Ltd represent media data as a series of discreet samples over time, and in the case over images, over space. Collabora Ltd generally don't care about the meaning of those samples, as the goal is to display them back to humans. This data is unstructured. Sometimes, Collabora Ltd instead want to structure the content of this data to extract a meaning. For example, instead of just reddish pixels, Collabora Ltd want to know that it's a strawberry. There exist a number of different type of algorithms to do this, from traditional computer vision to the latest trends in deep learning. But they all have in common that they produce some structured data describing the content of the input. | A typical example of object detection and classification using strawberries and leaves. More examples available here: https://col.la/gstanalyticsexamplesmodels. | GStreamer is a natural choice to handle this kind of metadata describing the underlying media data. It has a flexible system to attach arbitrary bits of data to a media buffer. Many companies have built their machine learning analysis framework around GStreamer, but no one had made the effort to contribute upstream, until now. Introducing the analytics metadata framework. Its goal was to create an analytics framework for GStreamer that decouples analysis steps from each other, leverages platform-specific acceleration where available, defines generic elements that function across platforms, and scales to large amounts of data and detections. GStreamer has a feature called a GstMeta which is a way to attach an arbitrary structure to a buffer (such as a video frame). In particular, there is also a region of interest meta that allows defining a rectangle in the image and attaching some data to it. Its first idea was to extend this, but Collabora Ltd realized that it couldn't scale. For example, in a wide shot of a crowd, you could detect hundreds of people. The other thing Collabora Ltd wanted make it easier to do the analysis in multiple steps, for example by having one step that detect objects, then further steps that find more information about specific objects. Collabora Ltd defined a new GstAnalyticsRelationMeta that stores an array of metadata structures along with a graph of relations between those. This enables Collabora Ltd to have an object at a specific location, then define a class of objects and have a "this object belongs to this class" type of relationship. For example, we can have a "car" class and a "tire" class, so Collabora Ltd can define a relationship between object 1 as a car and object 2 as a tire. Furthermore, Collabora Ltd can include a relationship between objects, such as object 2 being part of object 1 - the tire is part of the car. | In this example, there are 2 types of metadata, classification and object dectection. The classification further describes the objects. | Collabora Ltd has also defined some base classes of metadata: objects, classification and tracking. But more classes can be defined in the future, and plugins can even define their own. Collabora Ltd hope that this will be a first step to foster more collaboration between everyone using GStreamer as a common language for video analysis. Please don't hesitate to contact Collabora Ltd if you want to discuss your GStreamer projects, or want help building media analytics into your products. 28/07/2026 RADV's experimental Windows port brings Mesa's open source Vulkan driver for AMD GPUs closer to running real workloads, while highlighting... 17/07/2026 The Holo Core aarch64 preview provides early binaries, sources, and containers for an Arch Linux-based aarch64 port, backed by tooling and... 16/07/2026 Weston 16 expands HDR and color management support, improves Perfetto debugging and DRM backend performance, and adds new protocol and renderer...