and in the background successful launch desktop video cards Radeon RX 9070The company AMD is actively preparing a mobile series of graphic accelerators Radeon RX9000MAccording to leaks, the new lineup will include six models covering a wide range of performance.
The flagship RX 9080M will receive 64 computing units on the architecture RDNA4 (4096 stream processors) and 16 GB of GDDR6 video memory. The RX 9070M XT model will be equipped with 48 compute units and 12 GB of memory, similar to the desktop RX 9070 GRE.
The middle segment will be occupied by RX 9070M and RX 9070S with 32 blocks (2048 stream processors) and 8 GB of memory. The younger solutions RX 9060M and RX 9060S will receive 1792 cores and also 8 GB of memory. Versions with the "S" index will be optimized for power consumption, like Max-Q at NVIDIA.
According to reports, official announcement mobile series Radeon RX9000M will take place on Computex 2025 in early June. In addition to mobile cards, the desktop RX 9060 XT is expected to debut, which will intensify the competition in the affordable gaming segment.
Despite high expectations from the new processors meteor lake и Lunar Lake, sales of notebooks based on them have been extremely weak, Michelle Johnston Holthouse, Intel's general manager of products, said on the latest earnings call.
According to her, Buyers prefer older processors — rulers raptor lake (13th generation) and even earlier models, as the cost of laptops with the new chips turned out to be too high. Intel faced a situation where Demand for old processors exceeded expectations, and Meteor Lake and Lunar Lake stocks are growing.
The reasons for the failure are obvious: expensive cost structure, high prices for ready-made devices, unstable operation of new platforms and a general decline in interest in PCs with "artificial intelligence". This is especially true for the line Copilot+, which was never able to become a sales driver.
The situation is aggravated by the rapid growth of prices for components and new US trade duties, due to which buyers are trying to get their hands on them raptor lake until the next price increase.
Intel has acknowledged that demand for new laptops failed, and the main focus is now on future architecture Panther Lake, with first deliveries scheduled for late 2025. However, with interest in new platforms waning, the prospects for Panther Lake are also questionable.
AMD has introduced a new patch AGESA 1.2.0.3C for AM5 motherboards that fixes a serious vulnerability in processors Ryzen 9000. The problem associated with the verification of the microcode signature was previously discovered by specialists Google, which has caused concern among users of high-performance systems.
Company MSI was the first to start distributing the BIOS update with the new patch, which is already available for many 600 and 800 series motherboards. On the official MSI support page, you can find firmware for various models, including both premium solutions and more affordable boards.
The vulnerability affected the microcode verification mechanism in processors, which could potentially allow attackers to execute unauthorized code at the hardware level. The new patch fixes the issue without affecting performance systems: tests show no differences between the new BIOS version and previous builds.
Users are advised to install the BIOS update as soon as possible to protect their PCs. This is especially important for owners of Ryzen 9000 processors using AM5 systems. More motherboard manufacturers are expected to release firmware with the patch soon.
Company NVIDIA confirmed the release date of the new video card GeForce RTX 5060 - official sales and publication of reviews will begin simultaneously 19 May. Partners have already received updated embargo conditions and have also begun contacting journalists and testers.
The RTX 5060 will be the most affordable Blackwell-based solution and will feature chip GB206 with 3840 CUDA cores. The video card will receive 8 GB of GDDR6 memory on 128-bit tire, but with increased bandwidth thanks to the transition to the PCIe 5.0 standard (while maintaining eight connection lines).
NVIDIA priced the RTX 5060 at $299, like the previous RTX 4060 at launch. The new product offers 25% more cores and 65% higher memory bandwidth compared to the previous generation. However, users will have to rely on the company's official tests, as independent reviews will only appear on launch day.
Company GALAX announced the release of three models video GeForce RTX 5090D in the Hall of Fame (HOF) series, aimed at overclockers and enthusiasts. All versions of the map will be available exclusively on Chinese market, and there are no plans to release the model outside of China.
The line includes three options: OC LAB XOC, OC LAB Plus-X и OC LAB. The most powerful version of the XOC features dual 12V-2×6 power connectors, making it the only RTX 50 card with this configuration. The other models use a single connector but are based on a common PCB with a different number of components installed.
RTX 5090D HOF is built on a chip GB200-2000 with 21 CUDA cores and 760 GB of GDDR32 memory with speeds up to 7 Gbps. Boost frequencies vary: 28 MHz for the XOC Limited Edition, 2655 MHz for the OC LAB Plus-X, and 2610 MHz for the regular OC LAB version. All cards have a power limit of 2580 W, but the XOC version can access even higher numbers under certain conditions.
The cost of the models at launch will be:
OC LAB XOC LE – 29 yuan (~$999)
OC LAB Plus-X - 26 yuan (~$199)
OC LAB - 26 RMB (~$099)
There is no information yet on the number of copies available or exact points of sale.
TSMC announced plans to start production chips by process technology 1.4 nm in 2028. The new node will be called A14 (or 14-angstrom process technology) and will be the Taiwanese giant's first solution for sub-2-nm technologies. The company's representatives spoke about this at a symposium in North America, emphasizing that customers are already preparing for the transition to the next level of performance.
According to Nikkei Asia, the A14 process technology will provide an increase productivity up to 15% and a 30% reduction in energy consumption compared to previous nodes. TSMC did not disclose the names of the customers, it is expected that Apple will be the first major partner, given the close ties between the companies and Apple's focus on volume deliveries.
In parallel, in 2027, TSMC will begin implementing a new packaging technology that allows chiplets with different functions to be combined in a single package. This solution is aimed at the development of high-performance computing and AI.
TSMC's main rival, Samsung, has reportedly cancelled its own 1.4nm process technology, but is now focusing on 1nm chips scheduled for 2029. For now, TSMC remains the leader in the subnanometer technology race, strengthening its position through collaborations with the world's largest brands.
Intel is working on a new generation of automotive SoCs within the platform Frisco Lake - Is second generation SDV chips, built on Panther Lake architecture. According to recent leaks and kernel patches, these processors will use the same core as the upcoming desktop Panther Lake, which is expected to be released by the end of 2025.
The new chips will offer TDPs from 20 to 65 watts and in comparison with the previous generation, based on raptor lake, will receive 10x Boost in AI Performance and a 61% improvement in power efficiency. It was also confirmed that the graphics part will be built on the Xe3 Celestial architecture, which will replace the current Xe-LP and will be a major leap in capabilities.
In addition, there is support for up to 12 cameras and 280 audio channels, which indicates the automotive or multimedia focus of the platform.
Next generation - Grizzly Lake - will appear later and will be based on the architecture New Lake. Up to 32 power-efficient cores and integrated graphics with up to 7 TFLOPs of performance are expected, as well as AEC-Q100 Grade 2 certification and ASIL-B support.
Grizzly Lake is currently scheduled to launch on 2027 year, but the timing could change. Intel plans to speed up the rollout of automotive solutions to strengthen its position in the electric vehicle space.
Company AMD officially confirmed the holding of a press conference within the framework of Computex 2025To be held 21 May. The event will take place two days after a similar presentation by NVIDIA, and will be hosted by Jack Huin — Senior Vice President and head of the Computing and Graphics Group.
At the event, AMD plans to to announce several new products at once, including video, products for the AI segment and solutions for the enterprise sector. The description says that the presentation will cover the areas of gaming devices, AI-PC and enterprise technologies. New devices based on AMD's "vision in the field of AI" are also expected to be demonstrated.
Particular attention may be paid to the Radeon RX 9060 XT graphics card, rumors of which have been circulating for several months. This is also supported by the activity of the official AMD Radeon account on social networks, where hints of future announcements have already begun to appear.
Let us remind you separately that on June 12, AMD will hold another major presentation, which will be personally conducted by the company's head Lisa Su. It will be devoted to areas related to artificial intelligence, including chips and computing platforms.
Thus, within the framework of Computex 2025 we can expect several significant announcements from two competing companies - AMD and NVIDIA. Considering the excitement around the new GPU and AI solutions, May could be a key month for the PC market.
According to DigiTimes, AI companies в China started a mass sale video cards NVIDIA RTX 4090D. Reason - sharp drop in demand on computing power. Many servers are now running at only 20% capacity, forcing companies to shed excess hardware to remain profitable.
Amid a chip shortage due to US sanctions, Chinese companies have been modifying consumer RTX 4090D, installing additional VRAM modules to increase the power. These video cards have become an alternative to professional solutions for artificial intelligence tasks. However, today the market is cooling, and companies are trying to secure profits while the price of the cards remains high.
According to DigiTimes, price RTX 4090D with 48 GB of memory reaches $6000 in China's domestic market. However, active sales of surplus stock may reduce the price. This opens a window of opportunity for ordinary users, although the "pre-AI" prices are still far away.
The situation is similar to the AI cluster market: it has recently become known that more than 80% of all available AI resources in China are idle. In these conditions, manufacturers and integrators are looking for any way to monetize their stocks.
In parallel, rumors have emerged that NVIDIA has warned AIB partners about a possible ban on the RTX 5090D for the Chinese market. This could be another incentive for accelerated sales of 4090D and other cards in warehouses.
AMD has changed the release dates of its graphic cards on architecture RDNA4According to Board Channels, Radeon RX 9060 XT will be released earlier - on May 18, while RX 9070 GRE postponed to the end of 2025.
The RX 9060 XT will come in two variants - with 16 and 8 GB memory. Previously, the release was expected in June, but now it will take place a few days before the Computex 2025 exhibition. Presumably, this is an announcement, and not a full-fledged launch of sales.
Both versions of the RX 9060 XT are built on a 128-bit bus and will use GDDR6 memory with a speed of 20 Gbps, which will provide a bandwidth of up to 320 GB/s. In terms of performance and configuration, the card will compete with NVIDIA RTX 4060 Ti and the upcoming RTX 5060 Ti, but will be inferior to them in memory speed (GDDR6 vs. GDDR7).
Expected price new products will be $ 300 – 400, which makes them an interesting option in the mid-range segment.
At the same time, the Radeon RX 9070 GRE with 12 GB of memory, which was supposed to be presented on May 8, has been postponed until the end of the year. Now its launch is scheduled for the fourth quarter, presumably during the Chinese sales season on 11.11. This fits into AMD's tradition of releasing the GRE series mainly for the Asian market. The RX 9070 GRE will be based on a cut-down Navi 48 chip with a 192-bit bus.
