AMD has reportedly informed its GPU add-in board (AIB) partners that a price hike is imminent, according to a report from Chinese Board Channels. The report claims that AMD is preparing a 10% price increase on its GPU kits, which include a GPU die and GDDR6 memory, for July of this year. This increase will affect its AIBs that are producing custom Radeon RX GPUs with specialized PCBs. Typically, AMD sources GDDR6 memory from one of the three major memory makers like SK hynix, Samsung, or Micron. This approach allows GPU AIBs to avoid sourcing the memory themselves, and AMD can secure better contracts for volume purchases that surpass any single GPU AIB. However, since GDDR memory is also in short supply, AMD is forced to raise the price of its GPU kits supplied to AIBs like Sapphire, ASUS, XFX, and others.

For consumers, this doesn't necessarily mean an immediate 10% price hike, as the GPU kit is only part of the cost associated with the development of a GPU for third-party companies. While it is a significant expense, it may not account for the majority of the GPU cost, meaning that the final price increase consumers see might be less than 10%. However, this will also depend on whether AIBs pass these costs onto consumers, which they likely will, and exactly when the new GPU pricing appears at retailers. While GPU AIBs might honor pricing for models made before the price increase, retailers and smaller resellers might not, as they could use this information for an immediate price increase. Hence, we are yet to see how this plays out.

Earlier this year, we learned that NVIDIA might be planning to introduce a 9 GB edition of the GeForce RTX 5050 graphics card. However, later rumors suggested that the development was paused, and the latest reports claim that the project has been entirely canceled. According to MEGAsizeGPU, the GeForce RTX 5050 9 GB edition has been canceled because NVIDIA is currently relaunching its GeForce RTX 3060 12 GB edition GPU, which would compete with its own revived product. To avoid this overlap, the company has reportedly canceled the RTX 5050 9 GB edition launch. Initially, NVIDIA aimed to introduce this 9 GB version to address memory shortages in the lower-end segment. Instead of using four GDDR6 modules with 2 GB capacity each, NVIDIA planned to switch to three modules of GDDR7 with 3 GB capacity each, reducing the number of memory modules needed for this GPU.

For example, the current RTX 5050 uses 8 GB of 20 Gbps GDDR6 memory on a 128-bit bus, providing 320 GB/s of bandwidth. The 9 GB model was expected to use three modules of GDDR7 memory, each with 3 GB of capacity, resulting in a total of 9 GB across a 96-bit memory bus. While the narrower bus decreases the interface width, the switch to the new 28 Gbps GDDR7 memory would increase total memory bandwidth to 336 GB/s, a roughly 5% improvement, along with a 12.5% boost in VRAM capacity. Earlier leaks also claimed that NVIDIA was using the GB206 as a GPU base, while the older RTX 5050 8 GB with GDDR6 used the GB207 die. However, the core count remained at 2,560 CUDA cores. Since the GeForce RTX 3060 12 GB GPU is already on a path toward revival, launching this GPU wouldn't make sense.

Apple has reportedly been negotiating with the U.S. government to get clearance to use sanctioned Chinese memory maker CXMT in its products. According to a report from the Financial Times, Apple is in talks with the U.S. government to clear the previously blacklisted Chinese entity. The U.S. government had earlier placed Chinese memory maker CXMT on a sanctions list, preventing American companies from purchasing products like the now highly sought-after DRAM. This has created a challenging situation given the severe supply shortage the entire industry is facing. Since the big three—Samsung, SK hynix, and Micron—are ramping up capacity at an insufficient rate, sourcing DRAM from another supplier would help ease the pressure on electronics makers by providing better pricing, more supply, and facilitating product shipments.

CXMT is capable of manufacturing the latest DDR5 and its low-power sibling LPDDR5X memory. Late last year, the company showcased DDR5-8000 and LPDDR5X-10667 DRAM modules that are in mass production. These memory modules are available in 12 Gb and 16 Gb LPDDR5X capacities, while DDR5 scales to 16 Gb and 24 Gb module formats. However, a significant problem remains. Even if Apple gets clearance from the U.S. government to use CXMT memory, the company likely can't satisfy domestic demand first before shipping the quantity that Apple would require. Some estimates suggest CXMT could reach over $50 billion in revenue by 2026, according to SemiAnalysis, indicating that domestic demand is so strong that CXMT is rapidly selling out wafers.

On Monday, AMD released its Adrenalin Edition 26.6.2 WHQL Drivers to provide day-one game support for Assassin's Creed Black Flag Resynced and DOOM: The Dark Ages | Revelations across all RDNA generations. An interesting feature of this driver is the introduction of FSR Upscaling 4.1 for the RDNA 3 GPU generation. However, users have reported several issues when running this driver on Microsoft's Windows 10 operating system. For users of Windows 10 under the ESU program, or those simply running the operating system that has reached end of life, Radeon RX 7000 series and Radeon RX 9000 series GPUs have not been working properly. This means that the drivers for these GPUs weren't installing at all on Windows 10, which is problematic for enthusiasts still using the older operating system.

In response, AMD has released a hotfix to address this issue. The Adrenalin Edition 26.6.3 Hotfix includes a fix for the "intermittent install issue seen when installing AMD Software: Adrenalin Edition 26.6.2 on Windows 10 systems for Radeon RX 7000 series and above graphics products." For those running Windows 10 and experiencing issues, the Adrenalin Edition 26.6.3 Hotfix can be downloaded directly from AMD's website.

Intel's next-generation "Nova Lake" CPU microarchitecture is shaping up to be quite power-hungry. According to the latest leak, confirmed by well-known Intel leaker Jaykihn, Intel's top-end, overclocking-focused, dual compute die "Nova Lake-S" CPU with 52 cores will feature a PL2 mode with a power level of 474 W. This is an unprecedented number for an Intel consumer desktop platform, typically expected from a high-end desktop (HEDT) design. However, since this is a 52-core design that can be overclocked, it represents the newest high CPU PL2 setting that Intel will be shipping with "Nova Lake-S." Previous rumors suggested that the Intel Z990 platform would be a power-hungry consumer design, but now we understand just how power-hungry it is.

Reportedly, Intel's reference Z990 motherboard design, which it provides to partners like ASUS, MSI, GIGABYTE, ASRock, and others, now includes the possibility of three 8-pin CPU power connectors. This is an intriguing design choice, considering that most enthusiast boards typically ship with two 8-pin EPS CPU connectors. The three 8-pin connectors are mainly intended for the highest-end overclocking designs, rather than being required to run the dual compute die designs. Motherboards equipped with either two or three EPS connectors are still expected to support Intel's upcoming high-end 44-core and 52-core processors on the 175 W platform, meaning that only the overclocking-focused models will feature three connectors for additional overclocking headroom.

A very grim picture is being painted over the consumer PC space as Lenovo is now claiming that memory price increases will become the new norm. According to the latest report from Wall Street News China, Lenovo stated at ISC 2026 that the upward trajectory for memory prices has just begun, meaning that these increases will become a standard expectation. This is disappointing news for the majority of PC enthusiasts who were hoping for stabilization in DRAM and NAND Flash pricing, as these increases are pushing the cost of consumer electronics beyond previous levels of affordability.

Lenovo, one of the largest PC OEMs, frequently conducts channel checks with its DRAM and NAND Flash suppliers. The company is adept at projecting demand, allowing it to understand where the demand and supply ratio is heading. According to the company's analysis, even though Samsung, SK hynix, and Micron are accelerating the construction of new semiconductor fabs to expand capacity, this is simply not enough to meet demand. It is unclear whether this means that the new capacity coming online is insufficient or if the current capacity projections are inadequate. Either way, Lenovo projects that shortages of DRAM and NAND Flash will last longer than originally anticipated.

Microsoft has confirmed that it will offer Extended Security Updates (ESUs) for the Windows 10 operating system until late 2027. Specifically, ESUs will cover Windows 10 until October 12, 2027, after which Microsoft will stop providing security updates for this operating system. This extension gives those unwilling to upgrade a chance to extend the life of the legendary OS that has powered millions of desktops since 2015. On October 14, Microsoft officially ended general support for Windows 10, halting feature updates, routine quality-of-life improvements, and general technical support. Devices running this OS can still boot and run existing software, but the regular system-level enhancements and new capabilities that users have come to expect have stopped.

Originally, Microsoft designed its ESU program as a temporary remedy for the end-of-life status, allowing users to transition to the newer Windows 11 operating system, as migrations to a new platform can be complex. Organizations running Windows 10 can have tens of thousands of PCs, and managing an upgrade on this scale can be challenging. Therefore, organizations can often delay this transition by purchasing ESUs, which extend security features to keep their systems protected. In the European Economic Area, users receive the ESU package free for a year before Microsoft ends support, but they must sign in with a Microsoft Account. While ESUs can buy time for complex migrations, they are explicitly a temporary measure. Relying on ESUs long-term leaves systems exposed to unsupported components and growing incompatibilities with newer software.

Apple has reportedly canceled its M6 Pro and M6 Max processors, according to a recent note from Bloomberg's Apple insider Mark Gurman. Behind the scenes, there is a significant shift in Apple's silicon rollout strategy. The company is postponing the launch of its higher-end M6 SoC variants, like the M6 Pro and M6 Max, and is instead moving directly to the M7 processor family. Currently, Apple is testing the M6 SoC in the base MacBook Pro model, with many improvements in CPU microarchitecture and NPU performance for local AI processing. Reportedly, memory is also being upgraded for higher bandwidth, now targeting around 200 GB/s, up from 123 GB/s in the current base M5 SoC. The integrated GPU will see a 20% increase in core count, featuring 12 GPU cores in the M6, which will be a noticeable improvement from the current 10-core configuration in the M5.

Regarding the more powerful versions of the M6, such as the M6 Pro and M6 Max, this will be the first generation of Apple M-Series chips to lack the Pro and Max versions, as these will appear in the next-generation M7 series coming next year. Apple is planning a complete overhaul of the SoCs to bring major performance increases for on-device AI processing, which will first appear with the M7 SoCs. Next year's base M7 processor is expected to arrive in the first half of 2027 and will also upgrade memory bandwidth to about 240 GB/s. As a reminder, this is supposed to be the first Intel-manufactured Apple Silicon design, utilizing 18A-P node. The M7 Pro and M7 Max will arrive in late 2027 with much higher bandwidth and performance. An M7 Ultra version will follow in 2028 for Mac Studio.

Apple temporarily halted its online store today to implement a pricing update, bringing the store back online shortly after the changes were made. Even a leading consumer electronics company like Apple, known for having one of the best component supply chains in the world, couldn't ignore the impact of NAND Flash and DRAM price changes. As a result, the company has adjusted the prices of all products containing these components, reflecting an upward trend. Starting with its most affordable Mac option, the MacBook Neo, the price has increased from $599 to $699 for the base configuration. This pricing does not include the student discount, which has become more stringent. Apple now requires student verification for college students and other forms of verification for educational staff to access these discounts.

The average increase in today's price hike is $246.67. The most significant change is seen in the Mac Studio powered by the M3 Ultra SoC, which is now priced at $5,299, a $1,300 increase from its original $3,999 price. On the other end of the spectrum, the HomePod mini has experienced a $30 increase, going from $99 to $129. You can view the entire pricing update, compiled by MacRumors, below. Additionally, the 256 GB version of the Mac mini is back in stock, now priced at $799, up from its previous listing of $599. No price changes have been applied to any iPhone, Apple Watch, AirPods, Studio Display, and accessories.All price changes follow.

Yesterday, Qualcomm announced its High Bandwidth Compute (HBC) product, a memory-compute solution hybrid designed to replace traditional High Bandwidth Memory (HBM) and offer greater performance, efficiency, and throughput. HBC utilizes LPDDR memory, stacking them across multiple layers in a 3D vertical space and interconnecting them with through-silicon vias (TSVs). This approach provides better energy efficiency than traditional HBM, which stacks regular DDR in vertical layers, as the low-power DDR (LPDDR) chips consume less power while delivering similar bandwidth and capacity. Qualcomm has successfully stacked multiple LPDDR layers on top of each other, with the base of the HBC being a compute die that performs near-memory computation and offloads some processing from the main processor.

This technique is similar to what HBM4 memory currently does, with the base die being a logic die to better integrate compute solutions underneath for packet tracing and data preparation for entering and exiting HBM. However, Qualcomm aims to achieve higher efficiency and more optimized data movement, with a 6x increase in bandwidth per watt compared to the current HBM specification, likely HBM4. HBC Gen 1 achieved 133 TB/s of bandwidth on the AI250 accelerator card, a 18x increase over the LPDDR5X currently used in the AI200 card from Qualcomm. The company also claims there is a viable roadmap for using HBC across many Qualcomm AI accelerators, with HBC Gen 1 shipping with the AI250 AI accelerator in mid-2027. The next-generation HBC Gen 2 is expected to yield further significant gains in bandwidth.

AMD has released its latest Software Development Kit (SDK) version 2.3, bringing updates to several AMD FSR Redstone DLL-based machine learning technologies. According to AMD, the most significant change from the FSR SDK 2.2, released in March, is an update to Ray Regeneration, now at version 1.2. In Ray Regeneration v1.2, AMD has implemented quality improvements and optional add-on ambient and specular denoising. This means that with Ray Regeneration, games can deliver much better results compared to analytical denoising using traditional machine learning. Since FSR Ray Regeneration denoising is decoupled from FSR Upscaling, the visual fidelity should be significantly improved when running on AMD GPUs.

Additionally, in this SDK release, AMD has introduced FSR Upscaling 4.1.1 to the AMD Radeon RX 7000 series of GPUs, based on the RDNA 3 GPU architecture. Beyond this specific RDNA 3 feature, the FSR Frame Generation version has been updated to 4.0.1, which is a minor update for machine learning-based frame generation and the frame generation swapchain. This update provides an easy way to handle dispatching the workloads required for frame interpolation and frame pacing. Since AMD GPUs don't have hardware-based flip metering for frame pacing, AMD uses software pacing presentation to maintain consistent frame generation.

In late Q1, ASUS implemented a 30% price increase for PCs in Taiwan, with the rest of the world following closely. ASUS Joint Technology Systems Division General Manager Yi-Hsiang Liao announced this planned 30% price hike across the company's entire product line, explaining that the extremely high costs of DRAM and SSD storage, combined with a shortage of CPUs, were driving this increase. However, today we are facing another increase, but a much milder one. According to Money UDN, ASUS now expects single-digit PC price increases for Q3, which is a much better situation for the PC market than the continuously rising prices. While these price increases hurt consumers, they also impact PC makers' sales, meaning that signs of stabilizing prices present a good opportunity for the PC market to rebound in the coming months.

As PC price increases are now expected to be in the single digits compared to Q1, we can conclude that component pricing has exhausted its room for further increases from the mid-double digits seen since Q4 of 2025. We can hope to see this trend evolve into price stagnation in the coming quarters, as well as possible contractions down the line. Main cost drivers, including NAND Flash, DRAM, and CPUs, have made any price cuts impossible for the time being. However, there is hope that the supply chain will stabilize next year. DIY PC enthusiasts might see better availability in the coming quarters as major PC OEMs work with the supply chain to gain access to more DRAM and NAND Flash, but as of now, this remains a distant hope.

According to the June update of the TOP500 list, which tracks 500 of the world's most powerful supercomputers, China has taken the number one spot with its LineShine supercomputer. This system is capable of running 2.198 ExaFLOPS when measured on HPL, making it the world's first system to exceed 2 ExaFLOPS in dual-precision FP64 data processing for sustained workloads. It has a theoretical peak performance of 2.736 ExaFLOPS, meaning the system can sustain about 80% of its peak performance. This setup is powered by a massive network of traditional CPUs, with no GPUs or other accelerators contributing to the high FP64 figure. While traditional data centers have shifted toward 8-bit FP8 data formats for AI workloads, scientific HPC systems and supercomputers continue to use the traditional FP64 data. The Chinese LineShine is the first-ever system to sustain over 2 ExaFLOPS of compute capability, surpassing U.S.-based El Capitan, Aurora, and Frontier.

The LineShine comprises 13.79 million ARMv9 cores, spread across 20,480 nodes in a non-uniform memory access (NUMA) configuration. Each node contains two LX2 CPUs, with 304 cores and 32 GB of HBM memory. The CPU uses HBM for fast memory access, then offloads to system DDR memory, which has about 256 GB per CPU. The CPU is split across two dies, each divided into four NUMA domains. Each NUMA domain contains 38 cores running at 1.55 GHz and has 4 GB of HBM assigned to it. Four of these NUMA domains make up a die, and each die is assigned 128 GB of DDR memory. This is likely DDR5, but we haven't confirmed those details yet.

Earlier this year, the Munich Regional Court banned ASUS and Acer from directly selling their notebooks in Germany due to a patent dispute with Nokia. However, the two major notebook OEMs have reached an agreement with Nokia that allows them to sell their notebooks through agreements or by circumventing the rule. Readers may recall that Nokia's patent claims, which included patents related to core HEVC/H.265 encoding and decoding technology, forced both ASUS and Acer to halt notebook sales in Germany. Since their products use GPUs, processors, and other digital decoding/encoding engines for video, they must be licensed, and sales can only proceed with Nokia's approval. The judges concluded that the two OEMs had not demonstrated the behavior of willing licensees under FRAND rules, allowing the court to grant injunctive relief instead of limiting the remedy to damages.

However, Acer has now confirmed to its partners at a conference in Hamburg that it will resume shipping the disputed notebooks, though it did not provide specific details. In a statement obtained by ChannelPartner, Acer said, "Acer will ship products both with and without the HEVC codec. For products without a pre-installed HEVC codec, customers can activate support by installing the necessary software separately via official third-party channels." This could effectively be a circumvention of the rules, but a way to continue sales in Europe's largest economy.

Valve announced the availability of its long-awaited Steam Machine yesterday, bringing its mini home gaming cube to the masses. However, some interesting details about SteamOS—the operating system behind the Valve machines—have surfaced thanks to an interview with Valve's Pierre-Loup Griffais by The Verge. According to him, Valve has been working on expanding SteamOS compatibility with other hardware, even collaborating with NVIDIA to bring support for NVIDIA GeForce RTX GPUs to SteamOS, alongside numerous optimizations. Reportedly, Valve has been rolling out SteamOS improvements and optimizations to achieve much wider hardware compatibility than the current AMD-centric stack, meaning that the future landscape is expected to be much broader for Linux gaming enthusiasts.

For instance, SteamOS is Valve's fork of Arch Linux, with additional customization and optimization added to the operating system. Technically, NVIDIA GPUs can work under that OS with the installation of regular drivers, but Valve's modifications might hinder smooth usage due to many underlying OS changes. As a result, Valve has reportedly hired a "growing team" of engineers and developers to work on achieving proper support for NVIDIA drivers, which is expected to be available in the coming months. Griffais noted that this support might not arrive this year as Valve is optimizing the software stack for its current hardware like the Steam Deck and Steam Machine, but it will be available next year at the latest.