The Mesa 26.0-devel main development branch included important bug fixesrelated to High Dynamic Range (HDR) support in drivers VulkanThese fixes are also expected to be back-ported to the upcoming Mesa 25.3 release, making them available to a wider range of Linux users in the near future.
The patches were submitted by Mario Kleiner and affect the general windowing system integration code. Vulkan (WSI/display) Although the fixes are general, they have been tested and especially relevant for the AMD RADV driverIn total, three key changes have been made that should significantly improve the user experience when working with HDR content.
First of all, now HDR will be turned on, even Vulkan- the application does not call a specific function vkSetHdrMetadataEXT(), but simply selects the HDR color space for its swapchain. This will improve compatibility with many games and applications.
Secondly, the driver has learned correctly accept HdrMetadata metadata with a brightness level of 0 nits. According to the standard, such values are acceptable and mean "undefined/unknown" for the receiving device (monitor or TV).
Thirdly, and this is especially important for some drivers such as amdgpu, SDR mode recovery has been fixed when the swapchain is destroyed (for example, when closing the game). This driver requires a full atomic modeset to switch between SDR and HDR, and previously, SDR mode might not be restored correctly.
Nvidia CEO Jensen Huang made a series of important announcements regarding Elon Musk's plans, the Chinese market, and the AI race. Commenting on Elon Musk's ambitious plans to build his own chip factory, Huang expressed deep skepticism. According to him, Building an advanced chip manufacturing facility is extremely difficultHe emphasized that this is "not just building a factory," but also the complex "engineering, science, and true art" of what TSMC professionally does.
Regarding the Chinese market, Nvidia's CEO took a tough stance against export restrictions. He stated that at the moment there are no active negotiations about deliveries. “At the moment we We don't plan to supply anything to China."," he stated. He also added that the resumption of supplies depends on China itself and its desire to return Nvidia products to its market.
Huang also took the opportunity to clarify his previous comments about China's supposed "winning in AI." He denied making such a statement. "I said that China very good AI technologies" he clarified. As evidence, he cited the fact that there is a 50% of the world's AI researchers, who develop advanced artificial intelligence models.
ASUS has released a new update to its proprietary graphics card monitoring and overclocking utility. GPU Tweak III version 2.0.4.5The 105.89 MB update, dated November 5, 2025, brings not only bug fixes and stability improvements, but also a brand new exclusive feature.
The key innovation was Level Sense functionHowever, it won't be available to everyone. This feature was specifically added for the limited-edition anniversary graphics card. ROG Matrix Platinum GeForce RTX™ 5090 - ASUS Graphics Card 30th Anniversary EditionFor other users, the update focuses on improving overall usability.
A significant portion of the fixes affected owners of AMD graphics cards. The developers improved monitoring accuracy GPU AMD and fixed the discrepancy in the memory clock display between the "Monitor" and "OSD" tabs. A rare but annoying issue was also fixed. black screen error, which could occur with S.I. S-detection (S-VGA signal), which could occur with S-detection and S-VGA signals (S-VGA signal), which could occur with S-detection and S-VGA signals.
Among the system-wide fixes, the solution to the problem with duplicating the icon GPU Tweak III in the system tray. In addition, there was Fixed a bug with the Fan 1 curve, a percentage of which was blocked in 0dB mode. Other issues were also resolved: a non-functional "Browse" button when exporting a profile, translation errors in the monitoring section, incorrect interface width on ultrawide monitors, and reset after switching from two GPU for one.
Company Gigabyte announced its new flagship X870E AORUS XTREME X3D AI TOP motherboard for the AM5 platform. This model is positioned as the ultimate solution for gamers and enthusiasts using AMD Ryzen processors with 3D V-Cache technology. The board is equipped with a range of proprietary technologies, many of which are AI-focused.
The key feature was X3D Turbo Mode 2.0 — a technology that, according to the company, uses an AI model to dynamically optimize the parameters of the X3D processor in real time. Gigabyte promises Increase gaming performance by up to 25% All with the push of a button. Also included is the D5 Bionic Corsa AI memory overclocking technology, which provides support for modules DDR5 with a frequency of up to 9000 MT/s.
The power supply system is built according to the scheme 24+2+2 phases (Twin Digital VRM), and the board itself has an 8-layer server-grade structure with double the copper content. The cooling is "comprehensive thermal": it includes massive heatsinks (CPU Thermal Matrix), active memory cooling (DDR Wind Blade XTREME) and advanced M.2 Thermal Guard XTREME heatsinks. The board features five M.2 slots, two of which are PCIe 5.0.
The board also features numerous DIY-friendly features, including a 5-inch "Edge View" LCD screen for monitoring, magnetic M.2 EZ-Match mounts, reinforced PCIe EZ-Latch Plus Duo latches, and pre-installed Wi-Fi drivers (DriverBIOS). Networking is handled by Dual 10G LAN, Wi-Fi 7 and two USB4 40 Gbps ports.
An extremely disturbing rumor has appeared online from a UNIKO's Hardware insider, which could completely change the graphics card market in the coming year. According to his data, Nvidia has canceled the release of its updated "Super" lineup. for desktop PCs (RTX 5080 Super / 5070 Super) due to serious supply issues 3-gigabyte GDDR7 chips.
The reason for the cancellation is given as acute shortage of 3-gigabyte (24 Gb) GDDR7 chipsThis new, denser memory was intended for updated cards. However, Nvidia is now rumored to directs with priority all available volumes of these chips for the production of much more profitable products - flagship notebook GPU (RTX 5090 Laptop) and professional RTX Pro 6000 acceleratorsThere are simply no "Super" chips left for the consumer market.
The bad news doesn't end there. The insider also claims that prices for existing models (such as RTX 5090 and RTX 5080) will grow in the near futureThis is due to the fact that not only 3-gigabyte cards were in short supply, but also more common 2-gigabyte (16 Gbit) GDDR7 chips, the cost of which for producers has increased sharply.
Thus, instead of the expected lineup update with improved characteristics, the market may see the cancellation of long-awaited products and a simultaneous increase in the price of already released video cards.
Samsung Electronics has unveiled LPDDR6, the next generation of its low-power RAMThis type of memory is primarily intended for mobile platforms, edge computing, and embedded AI devices. It's important to note that this is memory that is soldered onto the board, not Not about desktop DIMM modules.
The new LPDDR6 standard is being produced 12nm process technologyAccording to the specifications, the memory supports data transfer rates up to 10.7 Gbit/s and uses an increased number of input/output (I/O) channels to increase overall throughput. This should speed up data-intensive applications on smartphones and laptops.
Particular attention was paid to energy efficiency. Implemented dynamic power management system, which regulates consumption depending on the load. This is claimed to ensure approximately 21% better energy efficiency compared to LPDDR5.
LPDDR6 also adds enhanced security mechanisms to protect data integrity. This allows the memory to be used not only in consumer electronics but also in industrial and embedded systems. This news is especially relevant for desktop PC users. has no practical significance, since this standard will not be used in their segment.
NVIDIA is actively developing Nova is a new open source graphics driver., written in Rust and designed for the company's modern graphics cards. The project is being created as an alternative to the outdated Nouveau and the official proprietary driver, and is already included in the Linux kernel. It will be added to Nova in the near future. support GPU Hopper and Blackwell architectures, including GeForce RTX 5090.
NVIDIA engineer John Hubbard published the first series of patches preparing the Nova kernel to support Blackwell and HopperAccording to him, the updates are based on changes made by colleagues and previous work on the boot0/boot42 components. Internal tests have already shown Nova booting successfully on GPU Blackwell and Ampere in real conditions.
At the same time, another NVIDIA engineer, Timur Tabi, is preparing support for the Turing architecture, which will also appear in Nova soon. However, the driver itself focused only on modern GPU, supporting GSP (GPU System Processor), which excludes older models.
Nova is developing as Rust-based DRM driver, integrated into the mainline Linux kernel. It is being built gradually, module by module, and at its current stage is not yet intended for everyday use. However, it is planned to become a standard by 2026. A complete and reliable driver for modern NVIDIA graphics cards. open source and highly secure.
Mesa 26.0 adds support for the extension VK_KHR_pipeline_binary for Intel and NVIDIA graphics cards. It allows game and application developers manually manage caching of graphics pipelines instead of using the built-in system VulkanThis may speed up loading, reduce lags and improve stability when launching games.
VK_KHR_pipeline_binary was added to the specification Vulkan back in August 2024, but for a long time remained active only in Radeon drivers. Now it Support appeared in Intel ANV and NVIDIA NVK - open Vulkan-drivers in Mesa.
In normal work Vulkan Automatically creates and caches "graphics pipelines"—sets of instructions used by the video card to render an image. However, this caching isn't always efficient. With the new extension, developers can save and load the required binary data from pipelines yourself, directly controlling the process. This is especially important for large games, where pipeline creation takes a significant amount of time.
Intel engineer Lionel Landwerlin worked on the implementation of the extension. Along with ANV support, he also made changes to the Mesa infrastructure, which allowed enable the extension for NVK as well, the new open source NVIDIA driver.
If you are not using the Git version of Mesa, then VK_KHR_pipeline_binary will appear for you Mesa 26.0 release next quarterThis is an important step for performance improvements, reduced loading times and control over graphics in games and applications.
Premium graphics card ASUS ROG Matrix Platinum RTX 5090 will go on sale later than planned - November 27 2025 years, as indicated on the German Proshop. The price will be €4099, which is almost double the European price of the reference RTX 5090 Founders Edition (€2099). The card is being created as exclusive anniversary version to commemorate the 30th anniversary of ROG and will be released in a limited edition - only 1000 pieces worldwide.
The device is built on the architecture NVIDIA Blackwellincludes 32 GB GDDR7 memory, takes up four slots in the case and is intended more for collectors, enthusiasts, and reviewers than for the mass market. The key feature is dual power system: standard 12V-2×6 provides up to 600W, and with a BTF-compatible motherboard, the card can consume up to 100W through a hidden slot. up to 800 W, opening about 10% additional performance headroom for extreme overclocking.
Cooling system includes four fans, copper steam chamber, liquid metal on a crystal, a massive heatsink, and additional copper in the PCB. ASUS also added a number of enthusiast-friendly features: Memory Defroster for extreme cooling, Power Detector+ for monitoring power instability and other tools.
Despite the impressive specifications, the high price makes the card inaccessible to most gamersFor this money, you could theoretically buy two RTX 5090 Founders Edition, which will limit interest from regular buyers. Matrix Platinum is a showpiece for one-off builds, nothing more.
ASUS is preparing to launch an exclusive graphics card. GeForce RTX 5090 ROG Matrix Platinum, created for the ROG brand's 30th anniversary. According to European retailers, the model will go on sale November 27 at the price of €4099, which is almost twice as expensive as the standard one RTX 5090 Founders EditionThe card is not intended for the mass market, but for enthusiasts, collectors, and extreme overclockers.
The flagship is built on a chip NVIDIA GB202-300 с 21760 CUDA cores, base frequency 2017 MHz and boost to 2730 MHzThe memory capacity is 32 GB GDDR7 with throughput 28 Gbps and a 512-bit bus. The interface is supported PCIe Gen5 and food standard 12V-2×6, and when connected to boards ASUS BTF energy consumption increases with 600 to 800 W, providing up to 10% additional performance.
The cooling system includes four fans, copper evaporation chamber, liquid metal and a massive radiator with a thick copper layer. Additional features are provided Memory Defroster и Power Detector+ for operation under extreme cooling conditions. A total of 100 units will be produced. 1000 copies, making the ROG Matrix RTX 5090 one of the rarest and most expensive Blackwell-generation graphics cards.
