专题：联想创新科技大会 新浪科技讯 北京时间2026年1月7日，AMD董事会主席兼CEO苏姿丰在联想Tech World上表示，联想将 成为首批采用AMD Helios机架级AI架构的系统供应商之一。 专题：联想创新科技大会 新浪科技讯 北京时间2026年1月7日，AMD董事会主席兼CEO苏姿丰在联想Tech World上表示，联想将 成为首批采用AMD Helios机架级AI架构的系统供应商之一。 苏姿丰称，全球企业都在思考如何让AI更贴近自身数据，同时保持灵活性和随时间推演进化的能力。 AMD与联想将合作推出搭载AMD EPYC处理器的AI推理服务器ThinkSystem SR675i，解决这一需求。 联想董事长兼CEO杨元庆称，这款推理服务器将提供强大的本地AI推理能力，助力客户获取实时洞 察、加速决策流程，并确保企业数据安全。 苏姿丰再次强调，AI将无处不在，广泛地存在于数据中心、工厂、医院、个人电脑以及边缘设备之 中。企业需要能够在云端、本地和边缘自由部署 AI，同时基于自有数据，按照自身业务需求灵活选择 CPU、GPU 与软件的最佳组合。 苏姿丰称，全球企业都在思考如何让AI更贴近自身数据，同时保 ...

责任编辑：刘万里 SF014 新浪科技讯 1月4日晚间消息，近日，联想集团推出首款基于AMD EPYC处理器的2U单路（2U1S）标准 机架服务器：联想问天 WR5215 G5。该产品集成了第五代AMD EPYC处理器，CPU核心数较上一代大 幅提升50%，并配备高达3TB的TruDDR5内存及14个PCIe扩展插槽，能为AI训练、虚拟化与科学计算等 高负载场景提供强劲算力。 新浪科技讯 1月4日晚间消息，近日，联想集团推出首款基于AMD EPYC处理器的2U单路（2U1S）标准 机架服务器：联想问天 WR5215 G5。该产品集成了第五代AMD EPYC处理器，CPU核心数较上一代大 幅提升50%，并配备高达3TB的TruDDR5内存及14个PCIe扩展插槽，能为AI训练、虚拟化与科学计算等 高负载场景提供强劲算力。 据了解，通过深度软硬一体化优化，联想问天 WR5215 G5服务器不仅在AI工作负载性能上实现25%的 显著提升，更在能效与成本层面实现双重突破：其单路设计可比传统双路方案节省高达25%的整机功 耗，配合联想问天海神液冷技术实现关键部件全覆盖，散热能力提升100%。 同时，该机型100%兼容国 ...

Core Insights - AMD has released a document detailing the architecture of its Zen 6 CPUs, indicating a complete redesign rather than an evolution from Zen 5, with a focus on throughput and performance [1][3] - The Zen 6 architecture supports up to 256 cores and utilizes TSMC's 2nm process technology, featuring an eight-slot dispatch engine and simultaneous multithreading (SMT) [1][2] Performance Characteristics - The design allows two hardware threads to dynamically compete for shared dispatch resources, which may result in lower single-thread performance compared to Apple's CPUs under certain conditions, but it is expected to unleash high performance in specific scenarios [2] - Zen 6 enhances visibility into vector and floating-point execution, emphasizing its capability for intensive mathematical workloads, although it does not directly confirm performance superiority in AVX-512 [2][6] Architectural Focus - Zen 6 is characterized as AMD's first microarchitecture designed from the ground up specifically for data center applications, with uncertain retention of features in consumer products [3] - The architecture supports full-width AVX-512 execution for various data formats, including FP64, FP32, FP16, and BF16, and maintains high throughput that necessitates merged performance counters for accurate measurement [6]

Core Insights - Intel has removed the next-generation 8-channel "Diamond Rapids" processors from its roadmap, focusing instead on 16-channel memory configurations for future server processors [2][13] - The transition to 16-channel memory is expected to be completed by the second half of 2026, aligning with the needs of future AI cluster builds [2][3] - Intel's Xeon 6700P series remains popular due to its cost-effectiveness and lower configuration costs compared to AMD EPYC processors [11][12] Summary by Sections Product Roadmap Changes - Intel's new leadership in the data center division has led to a significant change in the roadmap, with the 8-channel "Diamond Rapids" being removed [2][13] - The focus will now be on 16-channel processors, which will provide advantages for various customer applications [13] Memory Configuration and Performance - The shift from 12-channel to 16-channel memory is seen as a necessary evolution, with 16-channel configurations expected to offer similar memory capacity as 8-channel designs [12][14] - The 12-channel memory design previously offered a 50% theoretical bandwidth increase over 8-channel designs, but 8-channel platforms allow for more DIMM slots, enhancing memory capacity [5][7] Competitive Landscape - Intel's Xeon 6700 series is favored for its cost-effectiveness, allowing for configurations that do not require high core counts, thus appealing to a broader range of users [11][12] - The upcoming Granite Rapids-WS series is expected to compete aggressively with AMD's Threadripper 9000WX series, with specifications that may lead to a shift in market share [16][17] Future Developments - Intel is preparing to launch Granite Rapids-WS processors, which are anticipated to have up to 128 cores, enhancing its competitive position in the workstation market [16][17] - The performance of Granite Rapids-WS is expected to surpass that of AMD's EPYC processors, indicating a potential shift in the competitive dynamics of the server market [17][19]

Core Insights - AMD's AI business revenue is expected to reach several billion dollars by 2027, driven by increasing demand for AI-related products and services [4] Financial Performance - For Q3 2025, AMD reported revenue of $9.2 billion, a 36% year-over-year increase, marking a record high for quarterly revenue [1] - The company's net profit for the same quarter was $1.2 billion, up 61% year-over-year, with a gross margin of 52% [1] - The data center segment generated $4.3 billion in revenue, a 22% increase year-over-year, primarily due to demand for the fifth-generation AMD EPYC processors and Instinct MI350 series GPUs [3] - The client and gaming segment saw revenue of $4 billion, a significant 73% year-over-year increase [3] - The embedded segment reported revenue of $857 million, down 8% year-over-year [3] Future Outlook - AMD's revenue guidance for Q4 is set at a midpoint of $9.6 billion, with a fluctuation range of $300 million, exceeding previous market expectations [3] - The company anticipates a non-GAAP gross margin of approximately 54.5% for Q4 [3] - AMD plans to deploy the first 1 GW of AMD Instinct MI450 GPUs in collaboration with OpenAI in the second half of 2026 [3][4] - CEO Lisa Su indicated that many customers are planning significant CPU expansions to support AI workloads, with substantial CPU spending expected in 2026 [3] Market Context - AMD's stock price has increased by 107.01% from the beginning of the year to November 4, 2023, reaching a historical high of $264.33 per share at the end of October [4] - Despite exceeding revenue guidance, AMD's stock fell by 3.7% on November 4, amid broader market declines, with the Nasdaq down 2.04% [4]

Core Viewpoint - TSMC's advancement in wafer fabrication technology is leading to increased costs, impacting clients like Apple, Qualcomm, and MediaTek, although the premium for the new 2nm process may be less severe than initially expected, ranging from 10% to 20% compared to the 3nm process [1][2]. Pricing Dynamics - The anticipated price for TSMC's 2nm wafers is projected to be around $30,000 each, with the potential for a 50% price increase next year [5]. - TSMC's current 3nm process is expected to see price hikes, with the second-generation N3E reaching approximately $25,000 and the third-generation N3P around $27,000 [2][5]. Client Adaptation - Major clients like Qualcomm and MediaTek are adapting to the price increases, with Qualcomm planning to transition to the 2nm process for its Snapdragon 8 Elite Gen 6 SoC by 2026, and MediaTek already having successfully taped out its first 2nm SoC [3][5]. - Despite the cost pressures, clients are still willing to invest in TSMC's latest technology, indicating a strong demand for advanced semiconductor processes [3]. Market Demand and Competition - TSMC is reportedly experiencing high demand for its 2nm process, with 15 major companies preparing to adopt it, including AMD and MediaTek, and expectations that Apple will also become a client [5][6]. - The semiconductor industry is facing inflationary pressures, with rising prices for memory and storage components contributing to overall cost increases [5]. Production Capacity - TSMC is expanding its production capabilities by constructing multiple 2nm fabs in Taiwan and a third fab in Arizona, aiming to meet the growing market demand [6].

Group 1 - AMD's CEO Lisa Su will be a keynote speaker at CES 2026, discussing the company's vision for future AI solutions across various sectors [1][2] - The keynote will cover AMD's comprehensive product portfolio, including CPU, GPU, adaptive computing, and AI software, aimed at empowering customers and partners to tackle global challenges [2][3] - The event is scheduled for January 5, 2026, at 6:30 PM in the Palazzo Ballroom at the Venetian Hotel [3] Group 2 - AMD's EPYC processors are highlighted for their leading position in data centers, while AMD Instinct GPUs are noted for their exceptional acceleration capabilities [3] - The advancements of Ryzen CPUs and Radeon graphics cards in the AI PC and gaming sectors will also be discussed during the keynote [3]

Core Insights - Intel's CFO David Zinsner announced that the next-generation Nova Lake processors will be officially launched in 2026, promising to significantly outperform the current Arrow Lake and address strategic shortcomings in the high-end desktop CPU market [1][3] Group 1: Product Development - Intel is experiencing a decline in market share and sales in the high-performance desktop segment due to a lack of competitive products [3] - The Arrow Lake processors, despite utilizing TSMC's N3B process and featuring a 5.4GHz boost clock with a 36MB cache, have performance limitations due to architectural flaws [3] - Nova Lake is expected to provide a more comprehensive SKU lineup, covering all price segments from mainstream to high-end, thus filling the market gap left by Arrow Lake's limited product range [3] Group 2: Server Business Transition - In addition to client-side operations, Intel is accelerating its transformation in the server sector [3] - The Xeon Diamond Rapids processors have shown superior performance in certain scenarios compared to AMD EPYC, but overall performance has not met expectations [3] - Future developments will shift towards supporting simultaneous multithreading (SMT) technology, moving away from the previously planned non-SMT approach, with the next-generation Coral Rapids expected to launch between 2027 and 2028 [3]

Core Viewpoint - AMD reported strong revenue growth in Q2 2025, driven by demand for computing and AI products, despite facing challenges related to export controls on its MI308 data center GPUs [2][11]. Financial Performance - Q2 2025 revenue reached $7.685 billion, a 32% increase year-over-year from $5.835 billion in Q2 2024, and a 3% increase from $7.438 billion in Q1 2025 [4][6]. - Gross profit for Q2 2025 was $3.059 billion, up 7% from $2.864 billion in Q2 2024, but down 18% from $3.736 billion in Q1 2025 [4]. - The gross margin was 40%, down 9 percentage points year-over-year and down 10 percentage points quarter-over-quarter [4]. - Operating expenses increased by 23% year-over-year to $3.193 billion [4]. - Net income for Q2 2025 was $872 million, a 229% increase from $265 million in Q2 2024, and a 23% increase from $709 million in Q1 2025 [4][6]. Segment Performance - Data center revenue was $3.2 billion, a 14% increase year-over-year, driven by strong demand for AMD EPYC processors [11]. - Client and gaming segment revenue reached $3.6 billion, a 69% increase year-over-year, with client revenue hitting a record high of $2.5 billion [11]. - Embedded segment revenue was $824 million, down 4% year-over-year due to fluctuations in end-market demand [11]. Future Outlook - AMD expects Q3 2025 revenue to be approximately $8.7 billion, a year-over-year increase of about 28% and a quarter-over-quarter increase of about 13% [11]. - The company anticipates a Non-GAAP gross margin of around 54% for Q3 2025 [11]. - AMD's CEO expressed optimism about expanding AI business to hundreds of billions in revenue, highlighting strong interest in the next-generation MI400 series [13][14].

Core Insights - The partnership between Ruijie Networks and AMD aims to address the increasing global demand for computing power and the acceleration of digital transformation in China [1][8] - The collaboration focuses on a dual-driven model of "computing architecture innovation and scenario innovation" to provide efficient solutions for key areas such as intelligent computing centers and cloud desktops [1][5] Group 1: Market Context - According to a report by IDC and Inspur, China's intelligent computing power is expected to reach 1037.3 EFLOPS by 2025, representing a 43% year-on-year growth, with an even higher growth rate of 74.1% anticipated for 2024 [1] - The increasing demand for computing power is driven by applications in generative AI and autonomous driving, making the integration of network infrastructure and computing capabilities a core issue for industrial intelligence [1] Group 2: Technical Collaboration - The fourth-generation AMD EPYC processors utilize a 5nm process with up to 96 cores and support for PCIe 5.0 and CXL 1.1 memory expansion, achieving a 2.7 times improvement in energy efficiency compared to previous generations [2] - The collaboration includes the development of a dual-A cloud desktop solution that reduces total cost of ownership (TCO) by 30% and an AI training cluster based on a Rack Scale architecture that supports 64 interconnected GPUs [5] Group 3: Product Features - Ruijie Networks has introduced intelligent gateways in the data center network field, integrating AMD processors to achieve deep collaboration between computing and networking [4] - The RG-CS7010 cloud server, equipped with dual EPYC processors, supports 128 physical cores and has improved virtualization performance by 200% compared to previous models, making it suitable for high-frequency interaction scenarios in education and healthcare [4] Group 4: Strategic Goals - The partnership aims to build a self-controlled computing ecosystem, leveraging Ruijie Networks' self-developed Rack Scale architecture compatible with domestic GPUs to create a differentiated competitive edge in AI training clusters [6] - The collaboration aligns with national strategies such as "East Data West Computing" and aims to meet the goal of achieving a Power Usage Effectiveness (PUE) of less than 1.25 for new data centers by 2025 [7] Group 5: Future Outlook - The companies plan to launch the next generation of EPYC processors based on the sixth-generation Zen6 architecture and Ruijie's next-generation intelligent switches by 2025, which will enhance computing density and network bandwidth [7] - This partnership represents a significant shift in industry logic, aiming to overcome traditional IT architecture performance bottlenecks through deep collaboration between computing and networking [7][8]