Tag: Apple M5

  • Apple M6, M7 SoCs and Powerful AI Chip with 6x M3 Ultra Cores

    Apple M6, M7 SoCs and Powerful AI Chip with 6x M3 Ultra Cores

    Key Takeaways

    1. Apple is expected to launch the iPhone 17 series, new Apple Watch models, and a MacBook Pro with the M5 SoC in Fall 2025.
    2. Apple is developing new SoCs, including the M6 “Komodo” and M7 “Borneo,” which may debut in 2026 and 2027, respectively.
    3. A new AI server chip, codenamed “Baltra,” is in development and could feature 64 to 192 CPU cores and 160 to 480 GPU cores, expected to release in 2027.
    4. The current Apple M3 Ultra chip boasts significant performance improvements over the M2 Ultra, with up to 32 CPU cores and 80 GPU cores.
    5. Apple is also working on another SoC named “Sotra” for future Macs, though details about its specifications and target devices remain unclear.


    Like in the past few years, Apple is likely to unveil many new products in the Fall of 2025. We might witness the launch of the new iPhone 17 series, various accessories like updated Apple Watch models, and a new MacBook Pro before 2025 wraps up. This upcoming MacBook Pro could feature the new Apple M5 SoC.

    New Developments in SoCs

    Bloomberg has reported that besides the M5, Apple is also developing several other SoCs, including a new AI server chip. According to the report, Apple is working on the M6 “Komodo” and the M7 “Borneo” SoCs. If they stick to their usual launch timeline for SoCs, we could see the Apple M6 and M7 debut in 2026 and 2027, respectively.

    Exciting AI Chip News

    In a more exciting twist, the report mentions that Cupertino is creating a new AI server chip with the codename “Baltra.” Supposedly, Apple’s new AI chip is in collaboration with Broadcom and is expected to hit the market in 2027. This chip may have an enormous core count, as Apple appears to be testing configurations with 2x, 4x, and 6x the CPU and GPU cores compared to the M3 Ultra.

    For some context, the Apple M3 Ultra features up to 32 CPU cores, 80 GPU cores, and a 32-core Neural Engine for AI functions. Apple has stated that the M3 Ultra is about 1.5x quicker than the M2 Ultra when it comes to CPU performance and 2x faster for GPU tasks.

    Speculations on Core Counts

    Thus, Apple’s new AI server chip could potentially feature anywhere from 64 to 192 CPU cores and 160 to 480 GPU cores. These figures are substantial, so we should reserve judgment until more information is available.

    In addition to the M6, M7, and the “Baltra” AI server chip, it seems Apple is also working on an SoC codenamed “Sotra” for the Macs. No technical specs have been shared about this SoC, nor do we know whether it is aimed at future desktop Macs or MacBooks. If we were to make an educated guess, a high-end desktop chip appears to be the most logical choice, given that Apple’s current M-series SoCs serve the laptop market quite effectively.

    Source:
    Link

  • Apple M5 Starts Mass Production at TSMC’s N3P Node

    Apple M5 Starts Mass Production at TSMC’s N3P Node

    Apple’s new chip for laptops and tablets, called the M5, has begun mass production. This means we can expect it to be available around mid to late 2025, probably with either an iPad Pro or a MacBook Pro. Unlike earlier reports suggested, it won’t utilize TSMC’s N2 node but will rely on the N3-based N3P instead. This choice should provide a 5% boost in performance and enhance power efficiency by up to 10%.

    Improvements in Design

    Even though it won’t feature TSMC’s cutting-edge N2 gate-all-around technology (Nanosheets), the Apple M5 will include several significant upgrades. Among these is the SoIC-mH (System-on-Integrated-Chips-holding-Horizontal) packaging, enabling a more compact design. Additionally, the new substrate design will permit more chips to be stacked vertically, making the architecture more efficient. The report also hints at the introduction of an Apple M5 Ultra chip.

    Future Production Plans

    Industry analyst Ming-Chi Kuo hinted at the M5 Ultra’s production starting in 2026, signaling a comeback of the Ultra-branded chips from Apple after a two-generation gap. Kuo also mentioned that the Apple M5 will uniquely separate the unified CPU and GPU, which could be a strategic move to enhance AI performance—an area that has gained significant attention recently.

    Source:
    Link

  • TSMC’s 2nm Process Costs Over $30,000 Per Wafer

    TSMC’s 2nm Process Costs Over $30,000 Per Wafer

    The next major development for TSMC is its future N2 manufacturing process. As of early October, this technology is still under development, but an article from Commercial Times reveals potential costs associated with producing chips using this advanced method.

    Cost of N2 Technology

    The information suggests that creating a single 300-mm wafer with N2 technology will exceed $30,000. This is a significant amount, especially compared to past processes.

    In comparison, wafers made with N3 technology (commonly known as the 3nm process) are priced at around $18,500. Older technologies like N4 and N5 are even cheaper, costing about $15,000 each. This indicates that the expense of utilizing the upcoming N2 technology could be nearly double that of the N4 and N5 processes.

    Benefits of N2 Technology

    The increased costs might be largely justified by the various advantages the new technology promises to deliver. As reported by Tom’s Hardware, the 2nm process is expected to be highly efficient, boasting over 25% lower energy usage compared to N3E.

    It is important to mention that these figures are not officially provided by TSMC, so they should be viewed with caution. Additionally, pricing may fluctuate based on several factors, including order size and customer needs.

    Potential Early Adopters

    Apple is likely to be one of the first companies to adopt N2 technology. This includes the M5 series chips for devices like iPads, Mac minis, iMacs, and MacBooks, as well as the next A-series chips for iPhones. Other companies such as Qualcomm, Intel, AMD, Mediatek, and Nvidia may join later on in the process.

    Commercial Times (in Chinese) via Tom’s Hardware