Tag: Granite rapids

Intel recently released its Q2 FY2025 financial results, reporting revenue of $129 billion, flat compared to the same period last year. However, due to restructuring and asset impairment, net loss widened to $29 billion, including $19 billion in restructuring charges, $800 million in equipment impairment, and $200 million in one-time costs. Intel CEO Li-Wu Chen confirmed that a further 15% reduction in workforce will be implemented by the end of this year, bringing the global workforce down to 75,000 to optimize overall production capacity. Chen also stated that the company is accelerating organizational simplification and operational efficiency improvements, focusing more on core products and AI applications, while enhancing financial discipline in its foundry business. Although the transformation will take time, Intel remains confident in its future competitiveness and profitability. CFO David Zinsner pointed out that the reduction in operating expenses and asset disposals are showing initial results. Intel aims to reduce non-GAAP operating expenses to $17 billion by FY2025 and further to $16 billion by FY2026. According to its financial report, Intel's operating cash flow in the second quarter of fiscal year 2025 was $21 billion. Client Computing (CCG) revenue reached $79 billion, a 3% decrease year-over-year; Data Center and AI (DCAI) revenue was $39 billion, a 4% increase year-over-year; Intel Foundry revenue reached $44 billion, a 3% increase year-over-year; and other business segments saw a 20% increase, reaching $11 billion in revenue. Intel also announced the suspension of its expansion plans in Germany and Poland and the consolidation of its Costa Rican assembly and testing operations to Vietnam and Malaysia. Furthermore, construction of its new Ohio plant will be slowed to adjust capital expenditures in response to market demand. Regarding products, Intel launched three sixth-generation Xeon Scalable processors for server demand, supporting AI workloads and touting compatibility with the NVIDIA DGX B300 acceleration platform. The next-generation laptop processor, codenamed "Panther Lake," is expected to begin shipping by the end of the year, and Intel's 18A process technology has already entered production in Arizona. In addition, Intel explained that it sold 57.5 million shares of Mobileye stock to revitalize assets and strengthen its financial position, raising $920 million from the sale. Intel also emphasized that it remains a major shareholder of Mobileye. Intel forecasts revenue for the third quarter of fiscal year 2025 to be between $126 billion and $136 billion, with gross margins estimated at 34.1% (GAAP) and 36.0% (non-GAAP), and a loss per share of $0.24. Three Strategies to Create a "New Intel" In a letter to internal employees, Chen Liwu explained that the company will streamline its organization and reshape its organizational culture to create a "new Intel." Chen emphasized that these measures aim to create a faster, more agile Intel with engineering innovation at its core, eliminating cumbersome bureaucratic procedures, reducing costs to free up resources, and investing in growth momentum. Intel also confirmed that it will fully implement a return-to-the-office policy starting this September. Meanwhile, Chen Liwu also explained that the company will focus on three key strategies for restructuring: • Establishing a financially disciplined wafer foundry business: In the past, excessive capital investment without fully understanding market demand led to a scattered factory layout and idle capacity. Going forward, capital expenditures will be strictly controlled, expansion plans in Germany and Poland will be discontinued, and assembly and testing operations in Costa Rica will be consolidated to Vietnam and Malaysia. The construction of the new factory in Ohio, USA, will also be delayed, adjusted according to actual customer demand. Regarding process technology, Intel will fully promote the mass production of the Intel 18A process and use this technology to support customers including the US government. The subsequent Intel 14A will be jointly designed and put into production based on specific customer needs, emphasizing that "every investment must have a reasonable return." • ...

At its Hot Chips 2024 event, Intel showcased its optical computing interconnect chiplet design capable of handling 4TB of data transfer per second. This enables XPU-to-XPU connectivity, strengthening the heterogeneous integration of CPUs with different computing components and driving higher computing performance. Developed by Intel's Integrated Photonics Solutions (IPS) division, the design utilizes optical interconnects to allow the CPU to integrate different computing components, providing higher bandwidth and lower power consumption to meet ever-increasing computing demands. The current design can construct 64 bidirectional channels over approximately 100 meters of fiber optic cable, supporting a data transfer rate of 32Gbps. The ability to transfer approximately 4TB of data per second accelerates computational efficiency, particularly in artificial intelligence. This design is expected to be used in future processor products, while also driving data center and high-performance computing capabilities, and enabling consistent memory expansion and resource allocation across computing components. In addition to announcing the optical computing interconnect chiplet, Intel further revealed details of its sixth-generation Xeon Scalable server processors, codenamed "Granite Rapids," expected to launch in the first half of 2025. These include a system-on-a-chip (SoC) combining chiplets with Intel's 4-processor-built I/O controller, resulting in improved transistor density, performance, and energy efficiency. It supports up to 32 lanes of PCIe 5.0 and 16 lanes of CXL 2.0, as well as dual-port 100G Ethernet. Memory will be available in 4-channel and 8-channel configurations, packaged in BGA. Furthermore, it enhances edge computing applications by increasing workload temperature range and industrial-grade reliability. Intel also detailed the upcoming Lunar Lake client processors, announcing further information on September 3rd. Details of the new Intel Gaudi 3 AI accelerator will be announced along with the sixth-generation Xeon Scalable server processors this September.

Following the previous unveiling of its upcoming Xeon lineup, including the "Sierra Forest" and "Granite Rapids" series utilizing different combinations of P-core and E-core, as well as the "Emerald Rapids" series, which will become the 5th generation Xeon Scalable server processor, Intel further revealed at this year's Intel Innovation Day that the next-generation E-core Xeon product, codenamed Clearwater Forest, expected to launch in 2025, will be built using Intel's 18A process. The next-generation Xeon Scalable server processors will be released on December 14th this year. The "Emerald Rapids," as the 5th generation Intel Xeon Scalable server processor, will provide higher computing performance using the same amount of power. The "Sierra Forest," featuring a 288-core E-core design, is expected to launch in the first half of 2024, primarily increasing rack density by 2.5 times and performance per watt by 2.4 times compared to the fourth generation. As for the Xeon Scalable server processor, codenamed "Granite Rapids" and equipped with P-core performance cores, it will be launched after "Sierra Forest," offering 2-3 times the AI ​​computing performance of the fourth-generation products. Furthermore, Alibaba's Chief Technology Officer, Zhou Jingren, confirmed that their recently launched large-scale natural language model, "Tongyi Qianwen," and automatically generated AI services both operate on fourth-generation Xeon Scalable server processors, improving average response time by 3 times. Meanwhile, Intel CEO Pat Gelsinger anticipates that future computing development will be inextricably linked to AI, and Intel will continue to drive AI applications through its products. Furthermore, the OpenVINO software architecture can be used to create hybrid AI computing structures, making AI more ubiquitous.

At its Hot Chips conference, Intel announced its upcoming Xeon lineup, including "Sierra Forest" and "Granite Rapids," which utilize different core combinations of P-core and E-core. It also announced "Emerald Rapids," the 5th generation Xeon Scalable server processor, and new features for its Agilex 9 Direct RF series FPGA applications. The 5th generation Xeon Scalable server processor, codenamed "Emerald Rapids," is expected to launch in the fourth quarter of this year. Building on the success of the 4th generation Xeon Scalable server processor (codenamed "Sapphire Rapids"), which has shipped over 100 million units, Intel aims to accelerate the development of its data center product portfolio. The 5th generation Xeon Scalable server processor, codenamed "Emerald Rapids," adopts a P-core core combination design and supports new DDR memory specifications and high-bandwidth transmission standards. It also features hardware-managed data transfer between DDR5 and CXL memory via Intel Flat Memory Mode technology, while allowing software-defined control over the total memory usage. In terms of channels, it supports up to 136 PCIe 5.0 or CXL 2.0 lanes, and also supports 6 I/O ports linked via UPI. The product codenamed "Sierra Forest" is Intel's first Xeon Scalable server processor built entirely with E-Core energy-efficient cores, and also the first product to use Intel's 3-processor. It boasts a maximum core count of 144 and has set a record for multi-system boot performance. Unsurprisingly, it will be targeted at cloud virtualization, multitasking applications, and is expected to launch in the first half of 2024. According to Intel, "Sierra Forest" will increase server rack density by up to 2.5 times and performance per watt by up to 2.4 times, with a minimum TDP of 200W. It supports enhanced security, virtualization, and modern instruction sets such as AVX. It also features basic memory RAS capabilities and comes standard with machine checks...

Intel recently announced its new Agilex 7 FPGA architecture processor designed for data centers. Featuring an R-Tile chiplet design and combining PCIe 5.0 and CXL connectivity, it is already in mass production. The PCIe 5.0 connectivity provides more than double the data transfer bandwidth, while the CXL connectivity increases data throughput by four times. Combined with Intel's fourth-generation Xeon Scalable server processors, it can handle greater data processing volumes and efficiency. The FPGA architecture allows for adjustments based on workload and computational needs, adapting to real-world usage patterns and increasing deployment flexibility. It also enables businesses to more precisely control computing costs. Furthermore, at this year's ISC International Supercomputing Conference, Intel emphasized the higher computing performance its products deliver in combined supercomputer and artificial intelligence applications. Its GPU Max series 1550 accelerator for data centers is even touted as being 30% faster than the NVIDIA H100 GPU accelerator. Intel also previewed its next Xeon Scalable server processor, codenamed "Granite Rapids," and its next data center GPU accelerator, codenamed "Falcon Shores," to support various supercomputers and artificial intelligence, as well as the increasingly prevalent automated AI workloads. Previously, "Granite Rapids" was said to continue the design of "Sierra Frost," shortening the product launch time. It will utilize what is touted as the industry's fastest memory channel, employing a new DIMM slot called MCR (Multiplexer Array) to achieve data transfer rates of up to 8800MHz in DDR5 memory modules, offering up to 80% more peak bandwidth than most memory modules used in the industry. As for the "Falcon Shores" architecture, it integrates x86 processors and Xe GPUs into a single socket, achieving 5 times the computing performance-to-power ratio, 5 times the computing density, and correspondingly 5 times the memory throughput and bandwidth. Meanwhile, the "Falcon Shores" architecture will utilize Å (Ångström) level manufacturing processes, potentially employing Intel's 20A process, which is scheduled for production in the first half of 2024, or even the later 18A process, to support higher computing performance. During this ISC International Supercomputing Conference, Intel reiterated its collaboration with HPE and Argonne National Laboratory to create a more complete supercomputer, Aurora, aiming to drive the development of artificial intelligence and large-scale natural language model applications. Intel also emphasized leveraging tools such as oneAPI to make it easier for developers to access supercomputing and AI workloads.

Intel has announced its roadmap for upcoming Xeon Scalable server processors, confirming that the 5th generation Xeon Scalable server processor, codenamed "Emerald Rapids" and featuring a new generation of P Core performance cores, will be available in the fourth quarter of 2023. It also plans to launch a product codenamed "Sierra Forest" in the first half of 2024, featuring an all-E Core energy-efficient core design, followed by processors codenamed "Granite Rapids" and "Clearwater Forest". ▲Intel announces the 5th generation Xeon Scalable server processor codenamed "Emerald Rapids" and featuring a new generation of P Core performance cores. Intel has already launched the 4th generation Xeon Scalable server processor codenamed "Sapphire Rapids," and emphasizes that its competitor AMD's 4th generation EPYC server processor, with its multi-core design, can improve workload performance by an average of 4x in deep learning applications through its AMX accelerated architecture. In the recently announced future product roadmap, the product codenamed "Emerald Rapids" is confirmed to be the 5th generation Xeon Scalable server processor, expected to launch in the fourth quarter of 2023. Test samples have already been provided to select customers, and it will utilize the new generation P Core performance cores. ▲The 5th generation Xeon Scalable server processor is expected to launch in the fourth quarter of 2023. As for the product codenamed "Sierra Forest," it will be Intel's first Xeon Scalable server processor built entirely with E Core energy-efficient cores, and also the first product to use Intel's 3rd generation manufacturing process. It will have a maximum core count of 144 and has set a record for multi-system boot performance. It will undoubtedly be targeted at cloud virtualization, multitasking applications, and is expected to launch in the first half of 2024. Additionally, the product codenamed "Granite Rapids" is planned to be part of the "Sierra Forest" series...

In addition to announcing that Arc-branded graphics cards will be available in laptop versions in the first quarter and server versions are planned for the third quarter, Intel also announced its upcoming processor lineup for both the consumer and server markets. Among them, the 13th generation Core processors, codenamed "Raptor Lake," are expected to ship in the second half of this year. Compared to the 12th generation Core processors codenamed "Alder Lake," they will deliver double-digit performance improvements, while also offering ample overclocking headroom. They will feature a hybrid architecture supporting up to 24 cores and 32 threads, and will be built using Intel's 7-processor, making them compatible with existing 12th generation Core processor sockets. The next Core series processors, codenamed "Meteor Lake" and "Arrow Lake," will be built using Intel's 4 and 20A process technologies, respectively. "Arrow Lake" is expected to be manufactured using TSMC's foundry resources. Both will integrate artificial intelligence and GPU design, significantly improving XPU computing performance. "Meteor Lake" is expected to ship in 2023, while "Arrow Lake" is targeted for release in 2024. The Core series processors, codenamed "Lunar Lake," will be manufactured using both internal Intel processes and external manufacturing resources, and will likely also collaborate with TSMC. Furthermore, Intel reiterated that its Custom Compute Group will tailor products for blockchain, edge supercomputing, automotive advanced infotainment, immersive displays, and other emerging workloads. Products designed for blockchain will be unveiled at the upcoming 2022 International Solid-State Circuits Conference. In the server market, in addition to the upcoming "Sapphire Rapids" processors built with Intel's 7-processor technology, expected to deliver a 30x performance boost, Intel has also revealed plans for "Emerald Rapids," a compatible Sapphire Rapids socket design also built with Intel's 7-processor, to be released in 2023. This "Emerald Rapids" processor will offer performance improvements along with expanded memory and security advantages. Furthermore, the performance core (P-core) and efficiency core (E-core) designs currently used in "Alder Lake" will also be applied to future Xeon server processors, such as the "Sierra" processors expected to launch in 2024, built with Intel's 3-processor design...