普通人也眼热。二手网站充斥着刚从笔记本电脑上拆下来的"热乎"内存条，虽然只有一条，但不甘心的卖家坚定不移地相信：自己正站在风口。 金钱来得太快就像龙卷风。据说，内存条的龙卷风，一刮就要刮到2027年。 然而，12月3日，全球第三大DRAM供应商美光宣布退出消费级存储业务。哪怕供应再紧张，价格再飙升，美光也不做了。到口的肥肉，美光为什么不吃? 这股龙卷风，到底有什么玄机? 2025年以来，内存价格在6月一小涨，10月一大涨，涨幅基本超过100%，年底的价格可能更加疯狂。 内存条成了"电子茅台"，价格涨得比金条还快，力压世面所有理财产品。 64G的DDR5内存套装，价格超过4000元，比游戏主机PS5还贵出1000块。华强北商家已经不给内存条贴价格了，因为一天一个价，乃至上午一个价、下午一 个价。 威刚董事长公开表示，内存全部缺货，30多年没见过这场面，天天给客户道歉。小米董事长雷军在社交媒体上说"最近内存涨价太多"，以解释红米手机"微 微降价"背后的诚意。 连存储巨头三星都起了内讧。12月2日，三星电子半导体部门拒绝跟移动体验事业部签署一年以上的长期合同，只答应"3个月一签"。即使后者的高层亲自下 场协商，最后 ...

Core Viewpoint - Nvidia's CEO Jensen Huang announced the company's intention to exclude the Chinese market from future financial forecasts, indicating a potential complete withdrawal from China, marking a significant shift in the U.S. chip industry's strategy towards China [2][3]. Group 1: Nvidia's Challenges in China - Huang expressed disappointment with the Chinese market, citing restrictions faced during his visits and the possibility of Chinese internet companies halting purchases of Nvidia products [3][5]. - The company has faced difficulties with its specially developed H20 chips for China, which have been criticized for poor performance and low cost-effectiveness, leading to potential order cancellations [7][11]. - Nvidia's efforts to develop tailored chips to circumvent U.S. export controls have not yielded the expected results, resulting in significant inventory losses and a lack of interest from Chinese clients [11][13]. Group 2: U.S. Government Policies and Market Dynamics - U.S. government policies have created a challenging environment for Nvidia, with restrictions on advanced chip sales to China undermining its competitive edge in the market [13][15]. - Huang's comments reflect a broader concern that U.S. policies are detrimental to American companies' positions in the global AI landscape, potentially leading to a loss of competitive advantage [9][24]. - The ongoing geopolitical tensions have prompted a shift in the Chinese market, with local companies increasingly opting for domestic chip suppliers, reducing Nvidia's market share [19][21]. Group 3: Impact on Global Semiconductor Industry - The conflict between the U.S. and China is reshaping the global semiconductor supply chain, forcing companies to reconsider their market strategies [21][22]. - The global semiconductor industry is projected to shrink by approximately 12% in 2024 due to U.S.-China tensions, with further declines expected in 2025 [22]. - The situation highlights the end of an era of technological hegemony, as U.S. attempts to suppress Chinese technological advancement may inadvertently accelerate China's innovation efforts [24].

Group 1 - NVIDIA Corporation's newest AI chip for the Chinese market, the RTX6000D, is experiencing poor sales, with major tech companies in China opting not to place orders [1][2] - The RTX6000D is designed for AI inference tasks but is considered too expensive relative to its performance, which is reportedly inferior to the RTX5090 [2] - The RTX5090, although banned for use in China, is available through grey market channels at less than half the price of the RTX6000D [2] Group 2 - Chinese tech companies are awaiting clarity on the ability to order NVIDIA's H20 chip, which was recently allowed for sale, but shipments have not yet resumed [3] - There is hope among tech firms that NVIDIA's B30A chip, which is more powerful than the H20, will receive approval from U.S. authorities [3] - Analysts had previously expected strong demand for the RTX6000D, with forecasts of 1.5 million units to be produced in the second half of 2025 by JPMorgan and 2 million by Morgan Stanley [4] Group 3 - NVIDIA Corporation specializes in AI hardware and software, as well as high-performance computing solutions [4]

Core Viewpoint - The demand for NVIDIA's latest AI chip RTX6000D, tailored for the Chinese market, is lukewarm, with some major tech companies opting not to place orders [1][3]. Group 1: Product Performance and Market Demand - RTX6000D is primarily used for AI inference tasks, but its price is perceived as not commensurate with its functionality [3]. - Sample tests indicate that RTX6000D's performance lags behind the RTX5090, which is banned for use in China and is priced at less than half of RTX6000D's approximately 50,000 RMB [3]. - Despite optimistic forecasts from sell-side analysts, such as JPMorgan predicting a production of about 1.5 million units in the second half of the year and Morgan Stanley estimating a capacity of 2 million units, the actual demand remains low [3]. Group 2: Supply Chain and Regulatory Environment - Chinese tech giants like Alibaba, Tencent, and ByteDance are awaiting clarity on whether their orders for NVIDIA's H20 chip will be processed, following NVIDIA's re-approval for H20 chip sales in July [3]. - These companies are also hoping for approval from Washington for NVIDIA's more powerful B30A chip [3]. - The RTX6000D, H20, and B30A chips are downgraded versions for sale outside China, developed to comply with U.S. export restrictions aimed at curbing China's technological advancement [3]. Group 3: Recent Developments - NVIDIA has reportedly begun shipping the RTX6000D this week [4]. - A spokesperson for NVIDIA stated that the market is highly competitive, and the company is doing its best to provide the best products [5].

Core Viewpoint - Nvidia's newly launched AI chip RTX6000D faces weak market demand in China, with major tech companies reportedly not placing orders due to its high price and performance issues compared to the banned RTX5090 [2][3]. Group 1: Market Demand and Performance - The RTX6000D is primarily designed for AI inference tasks but is considered expensive relative to its performance [2]. - Sample tests indicate that the RTX6000D's performance is inferior to that of the RTX5090, which is available through gray market channels at less than half the price of the RTX6000D, approximately 50,000 RMB (around 7,000 USD) [2]. - Despite the weak demand, analysts from JPMorgan and Morgan Stanley had optimistic forecasts, predicting production of 1.5 million and 2 million units of RTX6000D, respectively, in the second half of the year [3]. Group 2: Regulatory and Trade Context - The ability to obtain advanced AI chips has become a focal point in the US-China trade tensions, with China's market regulator investigating Nvidia for potential antitrust violations [5]. - Chinese regulatory authorities have also questioned companies like Tencent and ByteDance regarding their procurement of the H20 chip, expressing concerns about information risks [5]. - Nvidia emphasizes that its products do not have any "backdoor risks" that would allow remote access or control [5]. Group 3: Chip Specifications and Future Prospects - The RTX6000D is based on Nvidia's latest Blackwell architecture, featuring a bandwidth of 1398 GB/s, slightly below the 1.4 TB/s threshold set by export restrictions [6]. - The H20 chip, priced between 10,000 to 12,000 USD, utilizes an older Hopper architecture but offers a higher bandwidth of 4 TB/s; however, its shipment has not yet commenced due to regulatory issues [6]. - The upcoming B30A chip, also based on the Blackwell architecture, is expected to deliver performance up to six times that of the H20 at a price only double that of the H20, pending approval from Washington [6].

Core Insights - Nvidia's RTX6000D AI chip is experiencing weak demand in the Chinese market, with major tech firms refraining from placing orders due to its high cost and performance issues compared to banned alternatives [1][2][5] - Chinese tech giants are awaiting clarity on orders for Nvidia's H20 chip, which has regained permission for sale, while also hoping for approval of the more powerful B30A chip [3][4] - The demand for RTX6000D contrasts with optimistic production forecasts from analysts, with expectations of 1.5 million to 2 million units in the pipeline for the second half of the year [5] Industry Context - The availability of advanced AI chips in China is a significant point of contention in U.S.-China trade relations, with the U.S. easing some export restrictions while China pushes for domestic chip adoption [7] - Nvidia faces additional challenges as Beijing has accused the company of violating anti-monopoly laws, adding uncertainty to its operations in China [8]

Core Viewpoint - The article discusses the evolution and current state of high refresh rate displays in gaming devices, highlighting the disparity between technological advancements and actual user experience and application support [1][5][25]. Group 1: High Refresh Rate Technology - High-end gaming devices are expected to have powerful CPUs, GPUs, large memory, and fast storage as core components [1]. - The introduction of high refresh rate displays, such as 750Hz for desktop PCs and 480Hz for laptops, is seen as a significant technological advancement [7][9]. - The smartphone market is also witnessing advancements, with brands like Redmi and OPPO considering 165Hz displays for their devices [11]. Group 2: User Experience and Market Demand - The initial benefits of high refresh rates were widely recognized, leading to a consensus in the consumer electronics industry [5]. - However, the current demand for refresh rates above 120Hz appears to be lukewarm among users, as many high-end laptops with 480Hz displays do not achieve the necessary frame rates in most games [15]. - The high pricing of top-tier GPUs raises concerns about the sustainability of a market focused solely on ultra-high refresh rates [15]. Group 3: Software and Ecosystem Support - The effectiveness of high refresh rate displays relies not only on the hardware but also on software adaptation, which is currently lacking for many applications [13][19]. - Historical examples show that previous high refresh rate initiatives in smartphones did not lead to widespread software support, limiting their practical use [19][21]. - The potential for 165Hz displays in smartphones is contingent on the development of compatible software, which remains uncertain [17][23]. Group 4: Future Outlook - The future application of ultra-high refresh rate displays may be confined to a narrow range of esports titles, limiting their impact on the broader industry [25]. - Without significant advancements in software adaptation and a clear value proposition beyond gaming, the appeal of these high refresh rate technologies may diminish [25].

Group 1 - The next-generation FSD chip AI5/HW5 from Tesla has entered mass production, with TSMC as the primary foundry [1] - The AI5/HW5 chip boasts a computing performance of 2,000 to 2,500 TOPS, which is five times that of the current HW4 chip at approximately 500 TOPS [1] - The AI5/HW5 chip will utilize a 3nm N3P process, with Samsung as a backup foundry, and is expected to be used in mass production vehicles by 2026 [1] Group 2 - Tesla plans to enhance the FSD hardware suite with upgraded lenses, including weather-resistant lenses from Samsung that can melt ice and snow in one minute [2] - A pilot program for Robotaxi was launched in Austin, with the first 12 Model Y vehicles equipped with HW4 hardware for testing autonomous driving [2]

Core Viewpoint - The article discusses the current market situation of the NVIDIA RTX 5090 graphics card, focusing on its price decline, rental market, computing power, power consumption, performance, heat issues, and networking capabilities. Pricing - The initial market price of the RTX 5090 was over 40,000 yuan, but it has dropped to just over 20,000 yuan within four months, with some brands listed as low as 23,000 yuan. This price decline is attributed to overheating issues, rumors of performance bottlenecks, initial high pricing by manufacturers, and competition from the previous generation RTX 4090 [2]. Rental Market - The high initial price of the RTX 5090 led to slow development in the rental market. As of May, with prices dropping, some data centers began offering RTX 5090 models for rent. The investment payback period for an 8-card machine is approximately four years, which may deter AI companies given the rapidly changing demand for computing power [3]. Computing Power - The RTX 5090 exhibits impressive computing power, particularly in AI training and inference scenarios, with a single card delivering 419 TFLOPS and an 8-card machine achieving around 3.4 PFLOPS. A cluster of 300 RTX 5090 cards can reach PFLOPS-level computing power. The rental price for a single card is about 10,000 yuan per year, while an 8-card machine costs around 300,000 yuan [4]. Power Consumption - The RTX 5090 has a rated power consumption of 575W, with peak consumption reaching up to 900W. An 8-card machine consumes approximately 6kW, leading to monthly electricity costs of about 3,600 yuan based on a rate of 0.6 yuan per kWh. High power consumption increases operational costs and necessitates robust cooling and power supply systems [6]. Performance - In AI inference scenarios, the RTX 5090 supports low-precision calculations (FP8 and FP4), significantly enhancing efficiency. It shows about a 50% improvement in inference speed compared to the previous generation RTX 4090. In gaming, it outperforms the 4090 at 4K resolution, but optimal performance requires targeted optimization [7]. Heat Issues - The RTX 5090 faces heat issues primarily related to the chip and power connectors, particularly the 12V-2x6 connector. Although such problems are infrequent, they require attention. Solutions include limiting peak power through driver or BIOS settings and enhancing cooling systems [8][11]. Networking Capabilities - Initial concerns about potential "lock card" issues in multi-card networking have been largely alleviated. Testing has shown stable performance in NVLink and PCIe networking, making the RTX 5090 suitable for building high-performance AI clusters [10][12].

Core Viewpoint - The article discusses the current market situation of the NVIDIA RTX 5090 graphics card, focusing on its price, rental market, computing power, power consumption, performance, heat generation, and networking capabilities since its release in January 2025 [1]. Pricing - The initial expected price of the RTX 5090 was over 40,000 yuan, but it has dropped to just over 20,000 yuan within four months, with some brands listed as low as 23,000 yuan on platforms like JD.com. This price decline is attributed to concerns over chip overheating, rumors of performance bottlenecks in multi-card setups, initial high pricing by manufacturers, and the competitive appeal of the previous generation RTX 4090 [2]. Rental Market - The high initial price of the RTX 5090 (over 30,000 yuan) led to slow development in the rental market. It wasn't until May, when prices fell, that some data centers began to offer RTX 5090 models for rent. Currently, the investment payback period for an 8-card machine is approximately four years, which may be too long for AI companies given the rapidly changing demand for computing power [3][6]. Computing Power - The RTX 5090 excels in computing power, particularly in AI training and inference scenarios, with a single card achieving 419 TFLOPS and an 8-card machine reaching about 3.4 PFLOPS. A cluster of 300 RTX 5090 cards can form a computing cluster capable of trillions of floating-point operations, making it advantageous for large language model training and high-performance computing tasks [4]. Power Consumption - The RTX 5090 has a rated power consumption of 575W, with peak consumption reaching up to 900W. An 8-card machine consumes approximately 6kW, leading to monthly electricity costs of around 3,600 yuan based on a rate of 0.6 yuan per kWh. This high power consumption increases operational costs and necessitates robust cooling and power supply systems [7]. Performance - In AI inference scenarios, the RTX 5090 supports low-precision calculations (FP8 and FP4), significantly enhancing efficiency. It shows about a 50% faster inference speed compared to the previous generation RTX 4090. In gaming, it outperforms the 4090 at 4K resolution, but optimal performance requires targeted optimization, especially in low-precision inference [8]. Heat Generation - The RTX 5090 faces heat issues primarily related to the chip and power connectors, particularly the 12V-2x6 connectors. Although such overheating incidents are rare, they require attention. Solutions include limiting peak power through driver or BIOS settings, using liquid cooling or turbo fans, and employing original power cables to avoid compatibility issues [9][10]. Networking - Initial concerns about potential "lock card" issues or performance bottlenecks in multi-card setups have not been substantiated in practical tests. Actual tests showed no such problems, and many companies using the RTX 5090 reported stable performance in NVLink and PCIe networking, making it suitable for building high-performance AI clusters [11].