➀ G.Skill and ASRock collaborate to redefine mini-PC performance with high-speed DDR5 CSO-DIMM memory modules; ➁ The new modules showcased in the ASRock DeskMini B860 system offer impressive speeds and capacities; ➂ G.Skill achieves DDR5-8133 speed with CL46-52-52-130 timings on a 48 GB kit.
Recent #memory news in the semiconductor industry
➀ V-color has introduced the world's first overclockable 256GB RDIMM DDR5 memory modules; ➁ Offering speeds up to 6000 MT/s, they support configurations up to 2TB; ➂ Optimized for GIGABYTE AI Top Series, AMD Ryzen Threadripper Pro, and Intel Xeon platforms.
➀ 35ELEMENTS致力于通过开发立方氮化镓半导体材料来帮助实现碳中和；➁ 公司计划在两年内在兼容CMOS的Si(100)衬底上扩大其材料规模，并寻求与代工厂的合作；➂ 35ELEMENTS拥有立方氮化镓材料的独家专利，并计划制造高效的光发射器，如创新型绿色发光二极管。
➀ Patriot celebrates its 40th anniversary with the Viper Xtreme 5 40 Years RGB memory at CES 2025, offering speeds up to 8000 MT/s and 48GB kits; ➁ The memory features RGB sync with major apps and a special edition heatshield; ➂ Other announcements include the Viper Xtreme 5 CKD Non-RGB DRAM and the iLuxe Stick C for Apple devices.
➀ G.Skill introduces new low-latency DDR5 memory kits; ➁ Available in 32GB (2x16GB) and 64GB (2x32GB) configurations with CL26 timings; ➂ World's first high-capacity, low-latency DDR5-6000 CL28 kits for content creators and professionals.
➀ Nvidia's upcoming GeForce RTX 5090 Laptop GPU is rumored to feature 24GB of GDDR7 memory; ➁ The GPU is expected to be based on the GB203 GPU with a 256-bit memory bus; ➂ The move to GDDR7 memory promises a substantial increase in bandwidth for the RTX 50 series.
➀ Samsung's Galaxy S25 smartphones will feature Qualcomm's Snapdragon 8 Elite processor, replacing its own Exynos 2400; ➁ The devices are rumored to use Micron's LPDDR5 RAM; ➂ Recent benchmarks show impressive performance, indicating a full shift to Snapdragon chips for international models.
➀ The article discusses the current state of server technology in 2025, including CPU core counts, networking speeds, SSD and HDD capacities, server memory, and AI-specific GPUs. It mentions the latest offerings from AMD, Intel, NVIDIA, Broadcom, and others. ➁ It highlights the trends in server technology, such as the increasing number of CPU cores, the rise of 400Gbps networking, and the growth in SSD capacities. ➂ The article also touches on the future of server technology, including the expected advancements in CPU performance, networking, and storage.
➀ China-made DDR5 memory chips are nearly 40% larger than Samsung's; ➁ CXMT's 16 Gb DDR5 memory IC has a larger die size due to less advanced chipmaking technology; ➂ The cost of CXMT's DDR5 chips is likely higher than Micron, Samsung, and SK Hynix due to the larger die size.
➀ DRAM prices are expected to decline by 8-13% in the first quarter of 2025, impacting PC, server, and GPU VRAM markets; ➁ The decline is driven by weak consumer demand and oversupply of DDR4 memory modules; ➂ Server DRAM prices are expected to drop by 5-10%, while GPU VRAM prices may fall by 5-10%.
➀ NVIDIA's next-gen GeForce RTX 5090 is rumored to cost $2600 with 32GB GDDR7 memory; ➁ Expected to offer 50-70% more performance at 4K than RTX 4090; ➂ RTX 5090 rumored to have 21,760 CUDA cores and 2.0TB/sec memory bandwidth.
➀ Researchers at the Advanced Institute for Materials Research (WPI-AIMR) have developed a cobalt-manganese-iron alloy thin film with high perpendicular magnetic anisotropy (PMA), crucial for MRAM devices; ➁ This alloy demonstrates a strong large PMA for the first time, addressing the challenges of current alloys; ➂ The research contributes to the development of innovative spintronics-based memory technologies for a sustainable future.
➀ Lexar has launched the world's fastest CFexpress 4.0 Type A Card Gold Series; ➁ The card offers read speeds of up to 1800 MB/s and write speeds of up to 1650 MB/s; ➂ It supports 8K RAW video capture and is designed for professional creators.
➀ SK hynix is preparing for mass production of its 16-Hi HBM3E memory; ➁ The company is integrating new equipment and optimizing facilities; ➂ Production tests have begun, with supply expected in the first half of 2025.
➀ A leak reveals NVIDIA's GeForce RTX 5090 PCB featuring the GB202 GPU and Samsung's 28Gbps GDDR7 memory modules; ➁ The card is expected to have 32GB of GDDR7 memory on a 512-bit interface, offering 1792GB/sec bandwidth; ➂ The PCB includes 28 VRMs and a single 16-pin power connector.
➀ Intel's shift to a more transparent strategy in technology announcements; ➁ The competition between Intel's PowerVia and TSMC's Super Power Rail; ➂ Andy Grove's philosophy of maintaining a 'healthy amount of paranoia'.
➀ The article commemorates the passing of Gordon Mah Ung, a respected technology journalist; ➁ The author reflects on the impact Gordon had on the technology industry and the personal regret of not expressing gratitude; ➂ The author mentions the importance of expressing gratitude and the family's philanthropic focus on pancreatic cancer research.
➀ SRAM has been used in high-performance computing architectures as embedded cache for decades, but its bit density expansion has slowed down and is affected by standby power issues.
➁ Spin Orbit Torque (SOT) MRAM offers advantages like low standby power, GHz-level switching speed, negligible leakage, almost infinite endurance, high reliability, and scalability, making it a promising alternative to SRAM.
➂ SOT-MRAM uses magnetic tunnel junctions (MTJ) as its basic building block, where the magnetic orientation can be vertical or in-plane, affecting the read and write operations.
➃ Recent advancements in SOT-MRAM technology, including improvements by imec, have demonstrated high switching speeds and durability, addressing key challenges in cache applications.
➄ Imec has developed innovative solutions for field-free switching, low dynamic power consumption, and scaling SOT-MRAM devices to the limit, making it a milestone for high-density SRAM applications.
➅ Further optimization of performance and reliability parameters, such as retention rate and write error rate, is being explored to bring SOT-MRAM closer to real-world specifications.
➆ Imec has also presented a novel composite free layer for MTJ, which improves the reliability of SOT-MRAM devices and reduces sensitivity to external magnetic disturbances.
➀ SRAM has been used in high-performance computing architectures as embedded cache for decades, but its bit density expansion has slowed down and is affected by standby power issues.
➁ Spin Orbit Torque (SOT) MRAM offers advantages like low standby power, GHz-level switching speed, negligible leakage, almost infinite endurance, high reliability, and scalability, making it a promising alternative to SRAM.
➂ SOT-MRAM uses magnetic tunnel junctions (MTJ) as its basic building block, where the magnetic orientation can be vertical or in-plane, affecting the read and write operations.
➃ Recent advancements in SOT-MRAM technology, including improvements by imec, have demonstrated high switching speeds and durability, addressing key challenges in cache applications.
➄ Imec has developed innovative solutions for field-free switching, low dynamic power consumption, and scaling SOT-MRAM devices to the limit, making it a milestone for high-density SRAM applications.
➅ Further optimization of performance and reliability parameters, such as retention rate and write error rate, is being explored to bring SOT-MRAM closer to real-world specifications.
➆ Imec has also presented a novel composite free layer for MTJ, which improves the reliability of SOT-MRAM devices and reduces sensitivity to external magnetic disturbances.
➀ Compute Express Link (CXL) technology is expected to move from a niche to mainstream use in 2025; ➁ CXL's support for memory expansion is a significant driver, with various server and memory solutions now available; ➂ CXL 2.0 and future generations of PCIe/CXL will enable more advanced use cases like switching and dynamic memory allocation.