大算力芯片

Search documents
特斯拉放弃Dojo对理想的潜在启发
理想TOP2· 2025-08-25 08:18
群友2: 阿里和字节有芯片,目前可能不如寒武纪。 群友1: 芯片最大的挑战不是芯片本身, 是能不能造出一个生态,类似于大家都以为英伟达是个纯芯片公司, 但是忽略了他的生态, 类似CUDA 和 TensorFlow 这类东西可以兼容各种平台, 其实英伟达是个算力模拟公司。寒武纪基本没有软件生态能力, 就硬件来说,不和英伟达比 ,和华为海思就 差距很大。 这个只能算是远期的畅想(本质是不严谨的),走到那步还有很长很长的路,咱们也不可能帮李想去做决策和判断。 TOP2:这块能细化展开一下吗? 群友1: 其实特斯拉给了另外一个很好的思路, 是芯片。 大算力芯片, AI6 基本可以替代Dojo了, 也就是说理想如果神经网络芯片+软件能力很强, 基 本就可以锚定寒武纪,做超算。 但本质还是得从业务和商业角度出发。 群友1: 抱歉,我对半导体研究不多,没法给出一个详细有力的论述。 但特斯拉最近把Dojo叫停了 (本身用于云端计算的大算力芯片,本来想一半用英 伟达一半用Dojo),其实不是放弃,大概率是Musk 发现这个2nm 的 AI6芯片既能上车还能组大集群做云端算力了。现在国家明显就是缺大算力芯片(看 最近的寒武纪就知 ...
沃格光电(603773.SH):与行业内多家客户合作开发的玻璃基大算力芯片先进封装在大型服务器的产业化应用
Ge Long Hui· 2025-08-14 08:00
Group 1 - The company, Woge Optoelectronics (603773.SH), is collaborating with multiple clients in the industry to develop advanced packaging for glass-based high-performance chips [1] - The project related to the industrial application of these chips in large servers is progressing smoothly [1] - The company indicates that while the project is on track, it will require some time before entering full industrial application [1]
何小鹏开启算力角逐:从“造车”到“造脑”的AI汽车突围战
2 1 Shi Ji Jing Ji Bao Dao· 2025-06-14 02:28
Core Viewpoint - Xiaopeng Motors has launched its self-developed Turing chip in its new model, the Xiaopeng G7, which is positioned to enhance the company's capabilities in the competitive smart driving market [1][2]. Group 1: Product Launch and Specifications - The Xiaopeng G7 is available in two versions, Max and Ultra, with a pre-sale price of 235,800 yuan. The G7 Max features dual NVIDIA Orin-X chips with a computing power of approximately 500 Tops, while the G7 Ultra is equipped with three Turing chips [1]. - The effective computing power of one Turing chip is equivalent to three NVIDIA Orin-X chips or two Tesla FSD chips, with ongoing optimizations expected to enhance this capability further [1][2]. - The G7's total effective computing power exceeds 2,200 Tops, significantly surpassing industry standards and meeting the requirements for L3-level intelligent driving assistance [1]. Group 2: Chip Development and Strategy - Xiaopeng invested over 10 billion yuan and spent five years developing the Turing chip, which began with team formation in late 2020 and achieved successful functionality verification within 40 days last year [2][3]. - The company aims to improve the efficiency of computing power usage, distinguishing its self-developed chip from generic industry chips that may include unnecessary computational capabilities [3]. - The Turing chip is designed to support both intelligent driving and smart cockpit functionalities, marking a significant evolution in automotive AI capabilities [4][5]. Group 3: Market Positioning and Competition - The G7 is strategically positioned in the competitive 250,000 yuan SUV market, traditionally dominated by Tesla's Model Y, which sold over 100,000 units in China in the first five months of this year [8]. - Xiaopeng's G7 has already received over 10,000 orders within 46 minutes of its pre-sale launch, indicating strong market interest [8]. - The company acknowledges the challenges of competing in a market that prioritizes practicality and space over technological performance, suggesting that sustained growth will depend on market acceptance [9]. Group 4: Research and Development Investment - Xiaopeng plans to allocate approximately 4.5 billion yuan of its R&D budget this year towards AI development, with total R&D expenditures nearing 10 billion yuan [10]. - The company is confident in achieving profitability by the fourth quarter, supported by its high R&D investment and the successful launch of the G7 [10]. - Xiaopeng aims to establish a competitive edge in the AI automotive sector, emphasizing the importance of innovation and a unique development path [10].