A new Intel graphics processor with the identifier has been discovered in the PCI-SIG database 99CMLN. Judging by the description, this is Arc B series graphics accelerator, working on PCI Express 5.0 x16 bus, that makes him first GPU from Intel with a full 16-lane PCIe Gen5.

Previous Intel Arc Battlemage cards such as B580 and B570, chip based BGM-G21, officially support only PCIe 4.0 x8, despite rumors of the presence of a Gen5 interface. Arc Pro B-seriesThe cards shown at Computex do use PCIe 5.0, but they are limited to the professional segment.

New entry from 9 May 2025 years may indicate either future consumer GPU BGM-G31, which was mentioned in leaks under the name Arc B770, or a unique Pro version with two GPU (e.g. MaxSun Arc Pro B60 Turbo). However, unlike OEM products, the recording officially belongs to Intel, and not partners, which increases the likelihood that we are talking about reference graphics card from Intel itself.

If it really is G31, he will become the first representative of the new, high-performance Arc class, capable of fully utilizing the PCIe 5.0 bandwidth. This is especially relevant given the competition with NVIDIA and AMD, which are actively implementing Gen5 support in future generations of their accelerators.

Add a comment

A new AMD graphics ID has been spotted in the latest Linux kernel version — GFX1250, which may indicate the upcoming release of a new generation of accelerators based on the architecture cDNA. This entry appeared in the official codebase and caused a wave of discussion in the industry: what exactly is hidden behind this designation?

Identifier GFX1250 is related to the CDNA architecture, which means, It's not about RDNA or graphics for gamers, and about computing accelerators - rather all, MI400 series, which should replace the current MI300 and MI350. It is important to note that the ID GFX1250 is following GFX1200 and GFX1201, which, according to insiders, are related to RDNA4And, therefore, GFX1250 is the next step after RDNA 4, but with an emphasis on computing and professional tasks.

CDNA is a line of AMD architectures, focused on servers, AI and HPC. Before that, we saw GFX90a (MI200), GFX942/950 (MI300) и The GFX1250 will apparently be a logical continuation of this line. But its location on the scale raises questions - it was previously assumed that the next generation after CDNA will be UDNA (Unified DNA), which combines RDNA and CDNA capabilities to compete with NVIDIA's CUDA. However, the GFX1250 is still does not have the GFX13xx marking, under which UDNA is usually expected to appear.

Add a comment

Palit has announced an international competition in which a prize will be awarded unique version of the GeForce RTX 5080 GamingPro video card, made in the style of the game DOOM: The Dark Ages. The special design of the device was created using the proprietary platform #PalitMaker, which allows you to customize the appearance of graphic adapters.

The draw is held exclusively through the company's official Instagram account - @Palit_Global. To participate, users must complete the simple steps specified in the competition publication. It is specified that the promotion is open only to residents Australia, Malaysia, Thailand, Indonesia and Singapore. The application period will end. 25 June 2025 year, after which the winner will be chosen.

The dedicated video card is a fully functional one. GeForce RTX 5080 with a custom DOOM: The Dark Ages-themed case. Judging by the image provided, the case GPU decorated with DOOM symbols and a characteristic texture that refers to the style of the game. This makes the device attractive not only to gamers, but also to collectors.

Interestingly, this model will not go on retail sale - Custom cases are developed through PalitMaker, the company's internal platform. The initiative allows fans to create unique graphics card designs that combine the aesthetics of games with the power of flagship GPU.

Thus, the card being played is not just productive hardware, but also exclusive item with limited design, available in a single copy. This highlights Palit's commitment to strengthening its presence in the creative segment of the market and focusing on user customization.

Add a comment

The Chinese Academy of Sciences (CAS) has officially unveiled QiMeng — a fully automated chip design system based on artificial intelligence. According to the developers, it is capable of replace engineering teams and reduce the time it takes to create chips from weeks to days. QiMeng was developed by the State Laboratory of Processors, the Center for Intelligent Software, and CAS University.

QiMeng uses large language models — the same technologies that underlie modern AI chatbots. The architecture includes three levels: a specialized model for processor chips, a design and optimization agent, and a set of ready-made applications for various tasks — from generating HDL code to setting up compilers and operating systems.

As part of the demonstration, the system designed two processors: QiMeng-CPU-v1, comparable to Intel 486and QiMeng-CPU-v2, similar to Cortex A53. Researchers claim that QiMeng is as efficient and quality as experienced professionals in terms of design. The system is already publicly available on GitHub.

Amid pressure from the US, which limited the supply of EDA software (automated design systems) to China, QiMeng is becoming an important part of the course towards technological independence. According to research by Morgan Stanley, Synopsys, Cadence and Siemens EDA accounted for 82% of the EDA market in China in 2024.

The development is aimed at reducing costs and accelerating the creation of specialized processors, which is critical in the context of sanctions and growing demand for AI accelerators. The developers have already announced plans to expand the system's capabilities, including self-learning functions and scaling to more modern architectures.

Add a comment

During the ASUS presentation at Computex 2025, the following was presented: unique video card ProArt GeForce RTX 5080, which stands out for two key features at once: an unusual 90 degree rotation of the graphics processor и the presence of a built-in slot for SSD format M.2.

In the photos published during the ASUS broadcast, you can see the back of the board with unusually oriented mount for the graphics chip, which indicates a rotated installation GPU. This is in stark contrast to traditional solutions in the RTX 50 series graphics cards, where GPU is installed strictly in the center in a fixed orientation. Such rotation requires complete redesign of the layout of the elements on the board, including power supply, cooling and video memory layout.

Additionally, the ProArt RTX 5080 is equipped with full-fledged SSD slot under M.2 drive and USB Type-C port, which makes it extremely attractive for users working with multimedia content and professional applications. No similar solutions with similar functionality have been noticed among other RTX 5080 yet.

Experts speculate that this design may be driven by the need to optimize cooling or tightly pack components inside SFF cases and workstations.

The ProArt RTX 5080 could be a new direction for graphics cards with expanded storage and interface options. It is not yet known whether the model will go on retail sale or is intended for OEM assemblies and the professional sector.

Add a comment

The portable gaming market is booming, with Logitech, ASUS, and Lenovo all releasing new models. But the Nintendo Switch 2 still holds the spotlight thanks to its unique balance between portability, exclusives, and a hybrid form factor. Let's compare it to its main competitors, including one of the most popular alternatives, Valve's Steam Deck.

Form factor and portability

The Nintendo Switch 2 is a 2-in-1 hybrid console that seamlessly switches between handheld and docked mode. Unlike the larger and heavier Steam Deck, the Switch 2 is more compact and easier to use on the go. The Steam Deck offers a full PC interface and more customization options, but it loses out on all-day versatility.

Performance and autonomy

The new chip in the Nintendo Switch 2 is optimized for gaming. Steam Deck has more powerful hardware, especially in terms of CPU and GPU, but it's less stable on high settings when not connected to the network. At the same time, the Switch 2 can work for up to 7 hours without recharging - a significant advantage on long trips.

Screen and graphics

The OLED display is expected to be 1080p in the handheld and up to 4K via the dock. The Steam Deck uses an IPS display with lower contrast. The OLED in the Nintendo Switch 2 delivers richer colors and deeper blacks, making the visual experience especially comfortable.

Game library and exclusives

Switch is Zelda, Mario, Pokémon, Metroid — exclusives that aren’t even on Steam. Steam Deck has access to Steam’s massive library, but it’s missing key Nintendo titles. Everyone chooses according to their preferences, but Nintendo’s unique style attracts millions of players.

Controllers and convenience

Joy-Con is all about local multiplayer, adaptability, and freedom. Steam Deck has a gamepad, touchpads, and tracks, which gives flexibility but takes some getting used to. In terms of ease of use, the Switch 2 remains more user-friendly for a wider audience.

Final World

• Portability + autonomy = mobility
• OLED + 4K = next level graphics
• Nintendo Exclusives = unique experience
• Joy-Con = flexibility and comfort

If you are looking for a balanced portable console with a hybrid format, powerful autonomy and a rich library of games - nintendo switch 2 will be an excellent choice. Especially for those who value simplicity, exclusives and multi-mode.

Add a comment

The PCI-SIG organization has officially approved PCI Express 7.0 specification, which provides throughput up to 128.0 GT/s и up to 512 GB/s in both directions when using the x16 configuration. The new standard is aimed at the most demanding scenarios, including AI/ML, 800G Ethernet, Cloud Computing and Quantum EngineeringAt the same time, preliminary work on PCIe 8.0 has begun.

PCIe 7.0 uses PAM4 (Pulse Amplitude Modulation with four levels) modulation и Flit Mode Coding, which provides increased energy efficiency compared to previous generations. Also, full compliance is observed backward compatibility with PCIe 1.0–6.0, making it easy to integrate into your existing ecosystem.

Additionally, the industry's first optical interface specification was announced - Optical Aware Retimer ECN. It introduces a standardized way to deploy PCIe over fiber, including between switches, hosts, and end devices. Supported multiplexing, mapping of data between electrical and optical channelsand increasing the connection range beyond the racks in data centers.

PCI-SIG notes that all major ASIC manufacturers are already working with IP providers, and interest in PCIe 7.0 turned out to be higher than even 6.0. Industry adoption is expected to begin in 2027, with the first list of verified devices expected by 2028.

Add a comment

Samsung has officially begun pre-production of the new Exynos 2600 mobile processor, which will be the company's first chip released 2nm process technology. This model will replace the unsuccessful Exynos 2500 and is expected to be used in the flagship lineup. Galaxy S26, which is scheduled to launch in the first quarter of 2026.

According to industry data, test assembly of the chip started in June, after the limited edition Exynos 2500 was completed. In the early stages, the 2nm process technology showed about 30% yield of suitable crystals, but now Samsung has succeeded achieve a level of more than 50% without loss of performance, which allowed us to move on to the next phase - "risky" production, in which a batch of several thousand chips is released to assess stability.

If the tests are successful, the company plans to begin Full-scale serial production of Exynos 2600 as early as early 2026, just in time for the Galaxy S26 launch. However, if things go wrong, Samsung could delay the launch again or return to Qualcomm solutions – as happened with the Exynos 2500 and the Galaxy S25 series, where the decision was made to abandon the chip entirely due to low yields of usable crystals.

Exynos 2600 to be First Commercial Solution on Samsung Foundry's 2nm Process, and its success will determine not only the position of System LSI, but also the future of the company's entire ecosystem of mobile solutions. After almost a decade of dependence on Qualcomm, Samsung wants to ensure complete autonomy in key components - and the Exynos 2600 should be the basis of this plan.

Add a comment

At the ISC 2025 conference, the company AMD Announces Future Power Consumption Forecasts for Zettaflop Supercomputers, and they were alarming. According to ComputerBase, such systems will require up to 500 megawatts of energy, which is comparable to the power of the whole nuclear power plant.

AMD has presented a graph of the growth of supercomputer energy efficiency from 2010 to 2035. In 2010-2015, the performance was allgo 3.2 GFLOPS per watt, but by 2035 it should reach 2140 GFLOPS/W, while the total capacity of the system can increase to half a gigawattThe forecast is based on the assumption that chip efficiency will double every 2.2 years.

The key limiting factors remain memory bandwidth and cooling systems. The higher the performance of accelerators, the more energy is required to transfer data and remove heat. This “snowball” effect is exacerbated by the need to support different precision formats - from FP128 and FP64 to FP16 and FP8, each of which requires different architectural support.

Against the background of already available chips, the situation looks realistic: Nvidia B200 consumes up to 1000 watts, and the new one AMD MI355X - up to 1400W... For comparison: Nvidia A100 5 years ago had TDP of only 400 W, less than the regular RTX 5090 graphics card. With the move to zettaflop data centers such as Oracle's 2.4 ZFLOPS cloud cluster, the issue of energy supply becomes critical.

 

Add a comment

Exclusive video card Asus ROG Astral GeForce RTX 5090 Dhahab OC Edition, which has a suggested retail price (MSRP) of $6,806, was originally intended exclusively for the Middle Eastern market. However, this model is now actively sold on eBay, where its value is significantly inflated.

On eBay, sellers, most of whom are based in China, are offering the ROG Astral GeForce RTX 5090 Dhahab OC Edition for prices ranging from $8,500 to $22,990. In this case, The average price on eBay is $14,998. This means that the resellers' markup is between 1.2 and 3.4 times the official price. In addition to this amount, buyers will also have to take into account additional costs for customs duties.

The appearance of such an exclusive card on eBay points to the existence of supply chains that allow products intended for a specific region to be redirected to the secondary market, creating a situation where those wishing to purchase the model are forced to deal with artificially inflated prices.

When comparing the Dhahab OC Edition to other versions of the RTX 5090, its price is particularly striking. It’s almost twice as expensive as the regular ROG Astral GeForce RTX 5090 32GB GDDR7 OC Edition ($3,499) and more than three times the price of the NVIDIA GeForce RTX 5090 Founders Edition ($1,999).

 

Add a comment