ASUS and Galax have become the latest NVIDIA board partners to raise prices on RTX 50-series cards, with increases landing September 1. According to a report from Channel Gate, ASUS is bumping RTX 5070 prices by 150 to 300 Yuan, or roughly $22 to $44, while Galax is applying a 200 Yuan (about $30) hike to the same card. Galax's RTX 5070 Ti is going up 300 Yuan (around $44), and its RTX 5080 sees the steepest jump at 500 Yuan (about $74). Budget cards like the RTX 5060 and RTX 5050, along with the flagship RTX 5090, are apparently untouched this time. The increases here are smaller in dollar terms than some earlier hikes, but they add up, and mid-range buyers eyeing the RTX 5070 family are in the crosshairs this time.

This is likely the third or fourth 2026 price hike for NVIDIA's board partners, and it results from the GDDR7 cost pressure we have tracked for months. NVIDIA reportedly raised the price of its GPU-memory kits for partners back in July, and that increase still seems to be trickling down one vendor at a time. NVIDIA had already raised prices on the RTX 5090 specifically in May for the same reason. With ASUS and Galax moving this week, other partners like MSI, Gigabyte, and Colorful will probably follow with adjustments of their own soon, since that's been the pattern all year. Right now, there seems to be no plausible end in sight either.

Micron is reportedly exploring the idea of developing high-endurance NAND Flash modules that would be positioned closer to the GPU, rather than being located far away in the storage pool behind multiple protocols. The company is said to be investigating ways to bring NAND Flash closer to handle workloads that don't require the latency and bandwidth of traditional HBM or regular DRAM modules. This architecture, currently referred to as "near-GPU NAND," would serve as a middle ground between density and durability. This concept would allow GPUs to access a pool of NAND Flash that has a lower density than conventional TLC or QLC NAND, but instead focuses on improving I/O speeds, bandwidth, and read times.

For instance, this would enable a GPU to access a potential storage pool with hundreds of gigabytes of storage directly on the GPU package or PCB, similar to what DRAM like HBM or GDDR7/LPDDR6 already does. This tier would sit between GPU memory and regular system storage, acting as a very fast buffer, which would be ideal for the inference of massive LLMs. At that point, LLMs wouldn't be memory-bound and could run on systems with fewer GPUs, but with more storage and memory tiers to compensate for the required inference size. Additionally, this NAND would be significantly cheaper than HBM or regular DRAM and would be sized according to system specifications. Images below are HBF slides from SK hynix and Sandisk.

NVIDIA has reportedly revived its "Rubin CPX" AI accelerator project, which had previously been put on hold. Today, well-known supply chain analyst Ming Chi Kuo reported that NVIDIA is reintroducing the "Rubin CPX," now featuring HBM4 memory instead of the previously planned GDDR7. Late last year, NVIDIA announced its plans to develop a specialized accelerator derived from the "Rubin" GPU family, designed for large-scale agentic AI workloads. After the announcement, the project was reportedly paused, but NVIDIA has now redesigned the memory architecture. Previously, the "Rubin CPX" was slated to have 128 GB of GDDR7 memory, but it will now include 168 GB of HBM4 memory. This GPU SKU is housed in a separate rack with configurations ranging from 64 to 256 standalone CPX GPUs per rack.

This GPU is dedicated to prefill workloads, powering racks that manage input context for LLMs and KV cache processing. Essentially, NVIDIA is separating prefill from decode tasks, with CPX GPUs handling prefill and regular "Rubin" GPUs managing decode. A rack tray with eight CPX GPUs can handle 1.34 TB of long-context prefill and the associated KV cache, all utilizing HBM4 memory for faster processing and increased bandwidth. Since the previous design did not require HBM4 integration, NVIDIA has redesigned the package, likely using TSMC's CoWoS-S or CoWoS-L for packaging.

The Framework Laptop 12 was an almost instant hit when it launched, thanks to its whimsical, compact design and durable chassis, but there was always a bit of a perception amongst hardware enthusiasts that the small laptop wasn't a serious computer, thanks to its outdated 13th-Gen Intel CPUs. Now, Framework has just announced that the convertible laptop has received an update to Intel's latest Core Series 3 CPUs.

Following the update, the Framework 12 is available for pre-order with the Intel Core 3 304, Core 5 320, and Core 7 350 CPUs, delivering Intel's latest six-core (2× P-Cores, 4× E-Cores) CPUs. Unfortunately, there are still no versions with the Intel Core Ultra X7 358H or any with a proper Intel Arc B390 iGPU. That said, the Framework Laptop 12 also now ships with two Thunderbolt 4-capable USB Type-C ports in the two rear-most expansion bays, theoretically allowing for eGPU expansion or other Thunderbolt peripherals.

Chinese memory giant CXMT has reportedly achieved a significant milestone in DDR5 production, with manufacturing yields now comfortably above 90%. This is comparable to what major manufacturers like Samsung, SK hynix, and Micron are achieving. According to reports, CXMT uses a 17 nm node for DDR5 chip production and has reached a 90% yield for these chips. This is only 2-3% lower than Samsung's yield on its 10 nm class node for DDR5 production, indicating that CXMT's production quality is matching that of the world's largest memory maker. This is a remarkable achievement for a company that excels in cleanroom construction time, taking only 12 months to build its fab's cleanroom compared to the two years others require. As the company plans to end the year with 350,000 DRAM wafers per month, high yields mean there is little waste or leftover.

Chinese MyDrivers also notes that PC OEMs have mixed reactions to adopting CXMT DRAM in their products. For instance, Dell has completely banned the use of CXMT in its machines, while HP only allows CXMT DRAM in its Chinese PCs. Major Taiwanese brands such as ASUS and Acer have increased adoption in some models, as they have more flexibility in supply chain procurement. However, both companies use CXMT memory for PCs sold in mainland China and some emerging markets, meaning the West is still far from receiving PCs with CXMT memory modules. PC OEMs are also cautious about procuring CXMT memory due to concerns about disrupting their relationships with the big three memory makers.

We have seen numerous reports about Chinese memory giant CXMT and its production. The company has now reportedly entered risk production of the latest LPDDR6 memory, reaching parity with Samsung, SK hynix, and Micron in both manufacturing and technological evolution. According to multiple reports, CXMT's LPDDR6 memory will operate at speeds of up to 12.8 Gbps, with the base speed of LPDDR6 being 10.7 Gbps. LPDDR6 picks up where LPDDR5X left off, which could reach speeds of up to 10.7 Gbps when highly overclocked. However, the newest standard is pushing for much higher speeds, with the first generation of LPDDR6 from CXMT's fabs reaching 12.8 Gbps. Additionally, CXMT is producing these chips with a 16 Gb capacity, equivalent to 2 GB, and the memory IC is placed inside a package-on-package (POP) casing, featuring about 1,295 pins in a ball grid array.

For now, this production will be only a trial run before high-volume manufacturing (HVM), which will likely take place weeks or even months after the initial risk production. Currently, the company is shipping its LPDDR6 to customers for validation, suggesting that we will likely see full risk production soon. As CXMT has manufacturing, construction, and other factors far ahead of industry standards, it is possible that HVM will occur in just a few weeks.

Renesas Electronics Corporation (TSE: 6723), a premier supplier of advanced semiconductor solutions, today announced its third-generation (Gen 3) DDR5 Multiplexed Rank Dual In-Line Memory Module (MRDIMM) chipset solutions, delivering DDR5 server-class MRDIMM speeds up to 16,000 mega transfers per second (MT/s). The new solutions are designed to address the growing demand for higher memory bandwidth driven by artificial intelligence (AI) data centers, cloud infrastructure and accelerated compute workloads.

Renesas will showcase memory interface components for Gen 3 MRDIMMs at FMS 2026 (Future of Memory and Storage Conference), Santa Clara, California, August 4-6, 2026. (Booth #807) They include third-generation Multiplexed Registering Clock Driver (MRCD, RRG5013) and Multiplexed Data Buffer (MDB, RRG5103) devices that support next-generation MRDIMM performance.

Major PC OEMs are beginning to secure deals with Chinese memory maker CXMT, as the supply chain has been depleted of available DRAM from the big three memory makers—SK hynix, Samsung, and Micron. According to a report from DigiTimes, CXMT's memory capacity has been booked through the end of 2027, leaving little room for new customers to source DRAM from China's largest memory producer. The report notes that large PC vendors such as Dell, HP, and Apple are the only ones that managed to receive allocations, while smaller brands with less presence in the desktop and laptop market have been left with nearly no allocation.

An important aspect of this situation is the negotiation leverage that larger PC vendors have. As they command more volume than smaller companies, Dell, HP, and Apple are willing to sign bigger contracts with CXMT to secure a larger DRAM supply, even at prices higher than historical levels. For CXMT, this is a positive indication that capacity expansion is well-received and a signal to continue building new fabs. DigiTimes also notes that the United States government has not yet made a firm decision on whether to keep CXMT on the barred entity list, so PC OEMs are taking advantage of the opportunity to stockpile any available DRAM. Future policy changes might impact this, but for now, long-term supply agreements are being made to secure any DDR5/LPDDR5X memory possible.

Chinese memory giant CXMT has been expanding its capacity so rapidly that the company is now able to compete with one of the largest memory manufacturers in the world—Micron. According to analysis from Citrini Research and their proprietary data, CXMT is projected to produce about 350,000 DRAM wafers per month by the end of 2026, assuming all current projects are completed on schedule. This is an impressive figure, especially considering that Micron, one of the "big three" DRAM manufacturers, is currently estimated to produce about 375,000 DRAM wafers per month by the end of 2026, assuming their capacity expansions are fulfilled. Micron has been a memory supplier for decades, while CXMT is a relatively new player that emerged to meet China's domestic needs.

A particularly surprising discovery is that CXMT is constructing its cleanrooms—where the actual silicon is produced—in just around 12 months, whereas the rest of the DRAM manufacturing industry typically takes 21-24 months, nearly a full year longer. The biggest challenge for Chinese memory makers like CXMT is acquiring advanced lithography solutions from around the world, as the United States government has imposed sanctions on the company and placed it on a blacklist, preventing access to modern EUV scanners. Instead, CXMT relies on older DUV lithography tools and uses multi-patterning techniques to work around the lack of EUV technology. Despite this, CXMT is matching Micron's output this year with nearly the same number of DRAM wafers, which is impressive for such a new market entrant.

TrendForce's latest memory pricing survey reveals that the DRAM market will remain extremely tight in the third quarter of 2026. However, weaker demand from consumer applications and the impact of a higher comparison base are expected to moderate contract price increases to 13-18% QoQ.

Demand for NAND Flash will continue to be driven primarily by AI inference and large-scale data center deployments. Yet, with contract prices already at record highs and consumer demand slowing, price tolerance among consumer customers has reached its limit. As a result, NAND Flash contract prices are projected to increase by 10-15% QoQ—a noticeably slower pace than in previous quarters.

Worldwide 300 mm fab equipment investment in the memory sector is projected to surpass $50 billion for the first time in 2026, rising 29% to $52 billion before increasing another 11% to $57 billion in 2027, SEMI reported today in its latest 300 mm Fab Outlook. The growth reflects continued AI-driven demand for advanced memory, supported by rising investment in AI infrastructure, data centers, and next-generation computing systems.

Looking further out, global 300 mm fab equipment spending in the memory sector is expected to grow at a 19% CAGR from 2024 to 2029. Worldwide 300 mm memory capacity is also projected to increase, reaching 4.1 million wafers per month in 2026 and 4.2 million wafers per month in 2027.

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.

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.

TrendForce's latest research reveals that structural tightening in mature-node DRAM supply is forcing consumer DRAM buyers to adopt legacy memory products to secure larger supply allocations. This has triggered a new wave of demand for procuring older-generation DRAM components, driving continued price momentum across legacy products such as DDR2 and DDR3.

Following strong gains in 1Q26, TrendForce estimates DDR2 contract prices will rise by approximately 55-60% in 2Q26, followed by a further 35-40% increase in 3Q26.

Chinese companies like Gloway and KingBank, which provide DDR5 memory kits for consumer platforms, have started integrating domestically produced CXMT DDR5 memory into their RAM kits, effectively ending their reliance on DRAM from Micron, SK hynix, and Samsung after years of external dependence. It has been reported that Gloway and KingBank have begun incorporating CXMT's 24 Gb DDR5 modules, which are being shipped as part of 48 GB RAM kits. With 3 GB per module, this results in eight modules per DDR5 DIMM for a total of 24 GB of capacity. In the regular dual DIMM configuration, Gloway and KingBank are offering consumers 48 GB of capacity.

Gloway has prepared the Longwuyi Yi Special Edition kit, which operates at 6,000 MT/s with latencies of CL36-38-38-80. This kit runs at a voltage of 1.2 V and is specialized for AMD CPUs. Reportedly, it features custom heat spreaders to maintain cooling and control heat dissipation, with 5 W/mK thermal pads included. Meanwhile, KingBank DIMMs use 2 mm heat spreaders and have a custom thermal interface material applied over the PMIC, along with dual-side RGB LED lighting that can be controlled via software.

We previously covered the announcement of the GPD Box, a new Panther Lake mini PC that features an MCIO 8i port and dual USB4 v2.0 ports for expanded connectivity with eGPUs and other high-bandwidth applications. Now, the GPD Box has finally launched on Indiegogo with a starting price of HK$11,377 ($1,452.25 converted). The base configuration is powered by an Intel Core Ultra 7 356H with 32 GB LPDDR5-8533 RAM and a 1 TB M.2 2280 NVMe SSD. The next level up upgrades the CPU to the Core Ultra X7 358H for HK$12,020 ($1,534 converted), and the bundles with the Box mini PC and the G2 MCIO eGPU Dock come in at HK$14,394 ($1,837 converted) for the Core Ultra 7 356H version and HK$15,036 ($1,919 converted) for the X7 358H variant. The biggest difference between the Core Ultra X7 358H and the Ultra 7 356H is the significantly more powerful Arc B390 iGPU in the former—importantly, there is no MCIO connection on the Core Ultra X7 358H version of the Box, leaving it with only USB4 for eGPU connections.

Despite its compact chassis, at just 175 × 134 × 39.5 mm, the GPD Box has decent I/O capabilities, with the highlight of the show being the MCIO 8i port that allows for eGPU expansion with minimal bandwidth loss compared to USB4 and Thunderbolt connections. According to GPD, the maximum bidirectional bandwidth of the GPD Box's MCIO port is 512 Gbps, which effectively matches a native PCIe 5.0 x8 cable connection. Notably, while the G2 eGPU dock does include a PSU with an 800 W ATX 3.1 PSU with two GPU power cables, you will need to bring your own GPU if you buy the bundle with the dock. The G2 dock can also be purchased separately via Indiegogo at a starting price of $385. The GPD Box itself also features dual 2.5 GbE RJ45 ports with link aggregation support. For wireless support, there is Bluetooth 5.3 and Wi-Fi 6E, while the physical I/O includes 2× USB4 V2 Type-C ports, 2× USB 3.2 Gen 2 Type-A, and a 3.5 mm audio combo jack on the front. The rear I/O has a DP 2.1 UHBR20 port, 1× HDMI 2.1 FRL ports, 2× RJ45 ports with 2.5 Gbps speeds, 2× USB 3.2 Gen 2 Type-A ports, and the MCIO 8i port.

Apacer has introduced GraTherX, an industrial-grade memory cooling technology aimed at addressing thermal challenges in high-speed DDR5 modules, particularly in fanless and space-constrained systems.

As demand for DDR5 continues to grow alongside AI applications, increasing thermal density and power consumption have raised concerns over system stability. According to the company, testing under high-load conditions shows that DDR5 modules equipped with GraTherX can achieve temperature reductions of up to 23.4°C, compared with the typical 3-5°C improvement seen in conventional cooling solutions. The design also improves heat dissipation across both sides of the module, helping maintain stable operation over extended periods.

Framework, the maker of various modular laptops, has updated its blog post regarding the pricing and availability of consumer memory and storage configurations for its products. As the company collaborates with its supply chain to equip its laptops and mini-PCs with necessary components, it possesses inside knowledge of pricing trends that ultimately affect consumer purchases. According to the update, Framework notes that the price of DRAM has stabilized for now, and the latest DDR5 memory has remained relatively unchanged in recent weeks. This is noteworthy as memory is one of the most expensive PC components these days. This means that suppliers are offering stable prices without requiring long-term contracts, although high demand is keeping prices elevated compared to the same period last year.

However, the situation is different for NAND Flash, as Framework notes that SSD prices are facing significant challenges. The company mentions that its inventory of SSDs purchased in 2025 still exists, and they have been selling these SSDs at prices well below the current market value. Essentially, Framework honored the pricing it set when acquiring these SSDs, providing customers with some relief when purchasing storage expansion. However, this will not remain the case for much longer, as Framework expects the old inventory to be depleted as soon as next month. The company will then have to sell SSDs at current acquisition prices, which are several times more expensive than those acquired last year. Framework previously sourced SSDs from Western Digital but is now introducing new modules from Sandisk, which now owns Western Digital SSDs, as well as from ADATA, Phison, and possibly other suppliers.

The Chinese DRAM market has just gained another player with POWEV entering the DDR5 market. POWEV is a sub-brand of the Chinese company Jiahe Jinwei, which announced its entry into the DDR5 market almost half a decade ago but is delivering now when it is needed most. POWEV has introduced DDR5 DRAM in form factors like UDIMM, SODIMM, and RDIMM, capable of up to 64 GB capacity and available in either 4,800 MT/s or 5,600 MT/s speeds. At these speeds, modules have different latencies, so it is best to refer to the specification table below. Firstly, POWEV separates its DDR5 products into "pure domestic" memory modules and industrial memory modules. This likely refers to some modules being for PC DIY and OEM applications, while others are for servers and industrial applications.

Both categories come with non-ECC unbuffered memory modules, while RDIMMs carry registered ECC on their own DIMMs. Every module operates at a voltage of 1.2 V, with a specific pin count depending on the form factor. Interestingly, it is unclear what technology POWEV uses to manufacture these DDR5 modules and create these DIMMs, as the underlying technology is undefined. South Korean memory makers such as SK hynix and Samsung usually reveal which iteration of their 10 nm manufacturing technology is used for chip manufacturing, but that is not the case here. Jiahe Jinwei has entered mass production of DDR5 memory, but no underlying technology has been disclosed. For many, this doesn't matter as long as the performance meets JEDEC's DDR5 specifications.

PC motherboard sales are on track for some of the biggest corrections in recent times as manufacturers struggle with weak demand, according to a DigiTimes report. What began as AI data center expansion quickly started affecting consumer PC DIY endeavors, as severe silicon shortages across the industry drove DRAM and CPU demand so high that prices have increased significantly for DDR4 and DDR5 memory kits, while regular CPUs have also seen a large price increase. In response, PC motherboard makers are caught in the middle of this shortage, seeing their motherboard unit sales revised down significantly. The report notes that all Taiwanese motherboard makers have significantly lowered their 2026 shipment targets, with some experiencing more than a 25% decrease in projected unit sales.

Interestingly, it's not only CPU and memory shortages driving this lowered demand; there are indications that consumers have slowed down their NVIDIA GPU upgrade cycles, which is impacting new motherboard sales. Particularly with the "Blackwell" GPU generation, consumers began purchasing PCIe 5.0 motherboards to achieve the greatest performance increase. However, as these GPUs became rarer and more expensive due to the global DRAM shortage, consumers have become reluctant to upgrade. ASUS is projected to sell about 10 million motherboards in 2026, while MSI and GIGABYTE are now projecting sales of less than 10 million units each. This represents about a 25% yearly decrease from 2025 sales. The worst position is estimated for ASRock, which is expected to see a 30% decrease according to the report.

In the latest Q1 2026 quarterly figures, ASML announced plans to ship over 60 EUV units this year, including both High-NA and Low-NA EUV lithography scanners. This is a significant increase from the 48 units shipped in 2025, but it doesn't break any records compared to the company's expectations for 2027. ASML predicts shipping about 80 EUV systems in 2027, driven by massive demand from memory makers. As memory has become the most sought-after commodity in AI data centers, there is simply not enough supply to meet demand. In Q1 2026, ASML reported that nearly half—45% to be precise—of its revenue came from South Korea.

While ASML hasn't provided specific data per manufacturer, we know that South Korean semiconductor makers primarily focus on memory, storage, and to a lesser extent, logic production. More than half of the quarterly shipments, at 51%, are dedicated to memory production. Companies like SK hynix and Samsung are purchasing many EUV systems to support their growing memory and storage businesses, which include the latest technologies like GDDR6, HBM3, HBM4, and DDR4/DDR5, all consumed by AI data centers worldwide. SK hynix plans to install 20 Low-NA EUV units in the next two years, all designed for HBM memory and advanced storage solutions, making SK hynix one of ASML's largest customers.

During a media tour at Lexar HQ in China, Digital Foundry spoke with Lexar's Europe General Manager, Grace Su, who explained what gamers are buying and where they are willing to compromise. Interestingly, Lexar found that, despite the pricing of DRAM and NAND Flash, users could purchase lower-capacity RAM kits without issues, while SSDs with capacities lower than 512 GB weren't selling well. Although this is not a direct comparison, when building a new PC system, you need both a RAM kit and an SSD. Based on their budget, enthusiasts often have to decide between a larger SSD or a bigger RAM kit. However, Lexar reports that users are so committed to having larger SSD space that they are even willing to revert to HDDs rather than buy an SSD smaller than 512 GB. This is understandable, considering that modern games can take up hundreds of gigabytes of space.

Currently, high RAM prices are driving PC gamers to invest considerably more in their gaming rigs to play smoothly. As 16 GB is somewhat considered the minimum for a Windows 11 system, some games and tasks are so demanding that PC DIY enthusiasts need to invest in 32 GB kits at prices that were unimaginable just a year ago. For example, we reported on the DDR4 spot market pricing, which saw the cost of a single 16 Gb module increase by about 2,200% over the past year, with only a minor 5% correction in March. We can only hope for more corrections soon so that PC gamers can finally see signs of relief, but with AI demand projected to continue rising, that seems far from reality.

Valve's Linux developer, Natalie Vock, has introduced a new method for handling memory management on GPUs with 8 GB or less VRAM. Since a large part of the gaming community uses systems with limited resources, improving performance and VRAM allocation is an excellent way to maximize frames per second. According to Vock's solution, GPUs with 8 GB of VRAM or less on Linux will now benefit from better background VRAM management, allowing more free space for games. Several Linux kernel patches have been submitted to guide how Linux manages resources when a game is loaded. Previously, when a game used too much VRAM on a low-VRAM GPU, the Linux Graphics Translation Table (GTT) would move data to system RAM to prevent the game from crashing. Since Linux doesn't prioritize programs, it could sometimes move the actual game from VRAM to system RAM to make space for something like a browser or another third-party app.

In Vock's testing before these new patches, running Cyberpunk 2077 resulted in the system using only 6 GB of the 8 GB of available GPU VRAM, with about 1.37 GB spilling over to GTT and being allocated in system RAM. This could lead to game stuttering and frame-pacing issues. Now, Vock has developed a real-time dmemcg-booster, which stands for Device Memory Control Groups, to inform Linux when a program running on the GPU VRAM needs to stay there without being moved to GTT and then to system RAM. With the new patch, running Cyberpunk 2077 now uses almost the entire available system VRAM, reaching 7.4 GB, while GTT allocation has been reduced to 650 MB, cutting usage by more than half. In the picture below, you can check the "GameThread" behavior before and after the patch is applied.

DDR4 pricing is finally showing signs of easing, as DDR4 16 Gb spot prices experienced a 5% correction in March after a more than 2,200% increase over the past year. The massive demand for AI data center expansion absorbed all available DDR production, making DDR4 and DDR5 pricing widely discussed topics among enthusiast communities. This has impacted PC DIY enthusiasts, laptop makers, PC OEMs, and everyone in between. Over the past year, the rapid price increase has effectively priced out many consumers from upgrading or installing new systems with larger RAM capacity. However, there are signs that the situation might improve as spot pricing has started to decrease.

First, readers need to understand the difference between spot and contract pricing in DDR memory. Contract pricing is what memory makers like Samsung, SK hynix, and Micron agree on with large buyers, such as PC OEMs like Apple, HP, or Dell, or even server makers who purchase DRAM in large quantities. On the other hand, spot pricing is an "on-the-spot" deal that provides immediate pricing without any contract. This spot pricing is typically reserved for smaller DRAM volumes and is usually bought by DRAM distributors and supply chain logistics. In the spot pricing market, DDR4 has fallen about 5% in March compared to February pricing, and the same applies to DDR5 memory spot pricing.

Major suppliers are continuing to phase out production of mature products below DDR4, according to TrendForce's latest research on the memory industry. As supply tightens structurally, DRAM prices have already posted significant cumulative increases in recent months.

TrendForce forecasts that consumer DRAM contract prices will continue to rise by 45-50% QoQ in 2Q26 after taking into account ongoing supply reductions, order transfers, and the slower pace of capacity expansion among Taiwanese suppliers.