PyTorch dataloader has 2 terrible default settings.Fixing them gave me ~5x speedup.When you train a PyTorch model on a GPU:- .to(device) transfers the data to the GPU.- Everything after this executes on the GPU.This means when the GPU is working, the CPU is idle, and when the CPU is working, the GPU is idle.Memory pinning optimizes this as follows:- When the model is trained on the 1st mini-batch, the CPU can transfer the 2nd mini-batch to the GPU.- This ensures that the GPU does not have to wait for the ne ...

Core Viewpoint - MoE Technology is on track to become the first domestic GPU company to complete its IPO, which will enhance its technological leadership and promote the self-controllable process of the domestic GPU industry [2][7]. Company Overview - Founded in 2020, MoE Technology focuses on the research, design, and sales of full-function GPUs and related products, aiming to provide computing acceleration platforms for high-performance computing fields such as AI and digital twins [3]. - The company has a diverse product portfolio that supports digital transformation across various industries, covering everything from cloud to edge computing [3]. Financial Performance - In the first half of 2025, MoE Technology achieved a revenue of 702 million yuan, surpassing the total revenue of 608 million yuan from 2022 to 2024, with a compound annual growth rate exceeding 200% [3]. - The revenue growth is attributed to the continuous expansion of its product matrix and sustained market demand, with four generations of GPU architecture chips launched from 2021 to 2024 [3]. - The net losses of MoE Technology have gradually narrowed, with losses of 1.839 billion yuan in 2022, 1.673 billion yuan in 2023, and 1.491 billion yuan in 2024, reducing to 270 million yuan in the first half of this year [4]. Profitability and Margins - The gross margin has improved significantly from -70.08% in 2022 to 25.87% in 2023, and is projected to reach 70.71% in 2024, maintaining a high level of 69.14% in the first half of 2025 [5]. - This trend indicates an enhancement in market competitiveness and profitability quality, with a gradual decrease in expense ratios as the scale effect becomes evident [5]. Research and Development - MoE Technology has invested over 4.3 billion yuan in R&D, employing nearly 900 staff, with over 75% holding master's degrees or higher [7]. - The company has applied for over 1,000 patents, with 514 granted, leading the domestic GPU industry in terms of authorized patents, showcasing strong independent innovation capabilities [7]. Future Outlook - MoE Technology anticipates significant performance growth by 2027, with expectations of achieving consolidated profitability [6][7]. - The funds raised from the IPO will primarily be used for the development of next-generation AI chips, graphics chips, and AI SoC chips, further enhancing its technological edge and promoting the self-controllable process of the domestic GPU industry [7].

Investment Strategy & Value Proposition - Magnetar focuses on being value-added to portfolio companies by helping them finance GPU and compute needs for data centers [3] - The firm prioritizes understanding the asset value and customer contracts to unlock value for companies [4] - Magnetar emphasizes a diversified portfolio approach, prioritizing downside protection by assessing assets and contracts [10] - The company aims to lower the cost of capital for its partners by engaging with management teams [9][11] Datacenter Infrastructure & Market Dynamics - Demand for GPUs is exceptionally high, with companies needing more [7] - Infrastructure investment is driven by skill, scale, and scarcity [5] - Datacenter projects are experiencing significant scale, moving from 100 GPU clusters to potentially millions of GPUs [5] - Large-scale neo cloud infrastructure projects are attracting legacy finance and strategic partners due to increasing capital requirements [8] Customer & Asset Mix - The customer mix and asset mix are crucial for lowering the cost of financing [13] - A portfolio with only non-investment grade customers is more difficult to finance [14] - Balancing non-investment grade (high-growth) and investment grade customers is important for portfolio financing decisions [14]

Intel uh coming off its uh best day in a long long time. The stock surged 22% following Nvidia's announcement that it will invest $5 billion in Intel. Joining us now is Ray Wong, Constellation Research founder and chairman.What I'm looking over in Becky's chair. It's a lot of room there. Normally you're in here with us.Uh Ray, I put on some extra makeup because you usually take a a picture and tweet it out. What What happened. Why aren't you here.Hey, I'd love to be in New York. Happy Friday. I I ended up i ...

DDN Exoscaler产品特性与优势 - DDN Exoscaler是一个并行文件系统，专为大规模数据处理而设计，旨在加速GPU流量，提高GPU的生产力，适用于生命科学、金融等多种行业 [1] - 该技术通过消除IO等待时间，使GPU能够持续获取数据，从而加速模型训练、推理以及提高token生成速度 [1] - DDN的解决方案旨在以最小的物理空间、功耗和网络占用提供最大的性能 [1] - DDN提供多种闪存配置（TLC、QLC）和混合系统（HDD），以满足不同客户在成本、性能和容量方面的需求，并支持将这些不同介质类型挂载到同一挂载点 [1][2] - DDN Exoscaler的客户端具备智能性，能够感知数据位置，从而优化数据访问路径，提高效率 [2] - DDN提供数据缩减系统，通过客户端压缩机制，在不影响存储性能的前提下实现数据压缩，数据缩减率通常在2到4倍之间，对于文本和日志数据最高可达50倍 [2] - DDN提供在线升级功能，允许在系统运行过程中进行升级，这对需要保持服务持续性的客户至关重要 [1][2] - DDN提供EMF分析工具，用于全面测试网络，帮助客户快速发现和解决网络问题，确保系统稳定运行 [2] - DDN Exoscaler支持多种协议访问，包括S3、NFS、SMB以及原生并行文件系统，并兼容Prometheus和Grafana等开源监控工具 [2] - DDN的监控系统能够显示哪些用户、客户端或作业正在对文件系统造成压力，帮助云服务提供商确保公平的数据访问 [2] DDN AI400X3产品 - DDN推出AI400X3，专为Nvidia Blackwell架构设计，旨在满足GPU技术快速发展带来的数据存储和访问需求 [1] - AI400X3在2U空间内提供150 GB/s的网络吞吐量，并提供95 GB/s的checkpoint速度 [1][3]

Key Industry Takeaways from Nvidia’s $5 Billion Investment in Intel1. Partnership Could Define and Accelerate the AI PC LandscapeFor Nvidia, developing its own Windows-on-ARM processors carries high uncertainty; for Intel, establishing a competitive edge in GPUs is difficult. Teaming up (CPU + GPU) could create powerful synergies and advantages across the PC ecosystem.2. Significant Synergistic Potential in x86 / Mid & Low-Range / Inference AI ServersA key trend ahead is enterprises building x86-based / mid ...

Collaboration Synergies - Nvidia and Intel collaboration aims to define AI PC and accelerate its development, leveraging combined CPU and GPU strengths [1] - Potential high synergy in x86/mid-to-low end/inference AI servers, combining Intel's x86 server enterprise client and channel resources with Nvidia's AI chip technology [2] Competitive Landscape Impact - Investment may shift market share among competitors like AMD in PC, GPU, and x86 server chips, and Broadcom in networking chips [2] TSMC's Perspective - TSMC's advanced process technology leadership is expected to remain until at least 2030, unaffected by Nvidia-Intel collaboration [1] - TSMC's AI chip orders are unaffected as AI chips require the most advanced processes [2] - Overall risk to TSMC is considered controllable, considering continued Nvidia and Intel orders, and lower contribution from networking products using less advanced processes [2]

Core Viewpoint - Chongda Technology (002815) has announced its collaboration with domestic GPU manufacturers, including Baidu, to deliver PCB products for the server sector in bulk [1] Company Summary - Chongda Technology is actively engaging in partnerships with domestic GPU manufacturers to enhance its product offerings in the server market [1]

Core Points - Intel China Chairman Wang Rui will retire this month as part of a planned management transition following the appointment of Wang Zhichong in February [1][3] - Wang Rui was the first female chairman of Intel China, joining the company in 1994 and holding various leadership roles [1][3] - Under Wang Rui's leadership, Intel China redefined its strategy, upgraded its organizational structure, and enhanced collaboration with global operations [3] - Despite these efforts, Intel faces challenges in the Chinese market due to deteriorating global business conditions and competition from rivals like NVIDIA and AMD [3][4] Company Developments - Intel's recent second-quarter earnings report showed revenue of $12.9 billion, flat year-over-year, but a net loss of $2.9 billion, marking the sixth consecutive quarter of losses [3] - The company is undergoing significant personnel changes globally, including the appointment of Chen Lifeng as the new CEO, who aims to reform the company and correct past mistakes [4][5] - Intel plans to reduce its workforce by approximately 15%, bringing the total number of employees down to around 75,000 by the end of the year [5] Management Transition - Wang Rui and Wang Zhichong have been working closely to ensure a smooth transition in the management of Intel China [7] - Wang Zhichong previously served as Vice President of Intel's Marketing Group and was appointed as Vice Chairman of Intel China in February [7] - The company has not confirmed whether Wang Zhichong will take over as Chairman of Intel China [7]

Market Overview - The A-share market experienced a slight decline with the Shanghai Composite Index closing at 3856.45 points, down 0.1% [1][2] - The total trading volume in the Shanghai and Shenzhen markets reached 1.48 trillion yuan, with 2881 stocks rising and 2376 stocks falling [2] Sector Performance - Technology sectors such as computing chips, semiconductors, and humanoid robots led the market gains, with the Yushubot concept index rising by 3.62% [5][6] - The rare earth sector faced a downturn, with notable declines in stocks like Northern Rare Earth, which fell by 4.8% [10][11] Notable Stocks - Hanwei Technology saw a significant increase, hitting the daily limit with a 20% rise, while Jiangsu Leili and Jingxing Paper also reached their daily limits with increases of 10.07% and 10.04% respectively [6][5] - Longxin Zhongke surged over 14%, and Haiguang Information rose by 8%, both reaching historical highs [7][8] Policy and Economic Context - Recent policies have been favorable for the technology sector, with the Ministry of Industry and Information Technology emphasizing the acceleration of new-generation information technology development [6] - The AI industry is projected to exceed 160 billion yuan by 2027, indicating strong growth potential in this sector [6] Gold Market - Gold prices reached a new high, with spot gold surpassing $3680 per ounce and COMEX futures hitting $3728.4 per ounce [12][13] - Domestic gold jewelry prices also increased, with brands like Chow Tai Fook and Chow Sang Sang pricing gold jewelry close to 1100 yuan per gram [13]