L
lulala
Guest
ST announced last week the world’s most powerful implementation of the ARM® Cortex®-M7 processor for the embedded market. Come and see it in action on our booth Hall 5, Stand 207.
It is more than twice as fast as the STM32F7 series, the previous STM32 flagship series, meaning that its core frequency of 400 MHz has enabled ST to become the first ever to reach 2010 points in CoreMark with a Cortex-M MCU. This is possible because ST is the first to have shrunk its M7 implementation from a 90 nm process node to 40 nm. Although some manufacturers have started or are about to start mass producing SoCs in 10 nm technologies, it is important to understand that these components only have digital circuits, unlike ST’s embedded MCU, which includes digital circuitry as well as Flash memory, and analog components which require much more complex processes.
Some reasons other than the technology shrink that enable the STM32H7 to set new performance records are detailed below.
Three Domains, Memory-Packed
Optimized Memory and FPU
Another great feature of the STM32H7 series is the ability to use ECC SRAM and Flash. The speed increase compared to the STM32F7 series is so high that ST now has the computational resources to add error correction and still break performance records. By providing ECC, ST not only ensures data integrity, but also improves data retention in the Flash.The inclusion of a double precision (FP64) floating point unit may not always be obvious, but some of the products that will benefit the most from the STM32H7 series need to perform DSP-type computations. For instance, an embedded system that monitors a power grid and will need to compute fast Fourier Transform algorithms, or a connected device that will run a precise GPS system will rely heavily on double precision computations.The STM32H7 series also builds on the previous generation by adding 10 more communication peripherals, making a total of 35, it still offers cryptographic and hashing hardware acceleration, and remains pin to pin as well as software compatible with the STM32F7 series.
Power Saving Features
Despite all this performance the STM32H7’s dynamic power consumption is 50% lower at only 250uA/MHz and it is possible to put D1 and D2 in a very low-powered standby mode (7µA) while D3 continues to capture data in its SRAM without needing to wake up the other domains, therefore greatly saving energy. There’s also a complex and elaborate clock-control scheme to ensure that different parts of the architecture run at varying speeds in order to further improve the MCU’s efficiency.The record-breaking STM32H7 series is sampling today to specific partners, and will be in mass production in Q2 2017. At this time, ST will have updated the mbed development platform to ensure developers can take full advantage of this groundbreaking architecture.More information about the STM32H7 series may be found on ST’s blog post or on it’s website. More information about ST’s Electronica presence, including the presentation program, can be found on the dedicated event pages.Meet also with the ST teams on the electronica Fast Forward startup platform.
It is more than twice as fast as the STM32F7 series, the previous STM32 flagship series, meaning that its core frequency of 400 MHz has enabled ST to become the first ever to reach 2010 points in CoreMark with a Cortex-M MCU. This is possible because ST is the first to have shrunk its M7 implementation from a 90 nm process node to 40 nm. Although some manufacturers have started or are about to start mass producing SoCs in 10 nm technologies, it is important to understand that these components only have digital circuits, unlike ST’s embedded MCU, which includes digital circuitry as well as Flash memory, and analog components which require much more complex processes.
Some reasons other than the technology shrink that enable the STM32H7 to set new performance records are detailed below.
Three Domains, Memory-Packed
Optimized Memory and FPU
Another great feature of the STM32H7 series is the ability to use ECC SRAM and Flash. The speed increase compared to the STM32F7 series is so high that ST now has the computational resources to add error correction and still break performance records. By providing ECC, ST not only ensures data integrity, but also improves data retention in the Flash.The inclusion of a double precision (FP64) floating point unit may not always be obvious, but some of the products that will benefit the most from the STM32H7 series need to perform DSP-type computations. For instance, an embedded system that monitors a power grid and will need to compute fast Fourier Transform algorithms, or a connected device that will run a precise GPS system will rely heavily on double precision computations.The STM32H7 series also builds on the previous generation by adding 10 more communication peripherals, making a total of 35, it still offers cryptographic and hashing hardware acceleration, and remains pin to pin as well as software compatible with the STM32F7 series.
Power Saving Features
Despite all this performance the STM32H7’s dynamic power consumption is 50% lower at only 250uA/MHz and it is possible to put D1 and D2 in a very low-powered standby mode (7µA) while D3 continues to capture data in its SRAM without needing to wake up the other domains, therefore greatly saving energy. There’s also a complex and elaborate clock-control scheme to ensure that different parts of the architecture run at varying speeds in order to further improve the MCU’s efficiency.The record-breaking STM32H7 series is sampling today to specific partners, and will be in mass production in Q2 2017. At this time, ST will have updated the mbed development platform to ensure developers can take full advantage of this groundbreaking architecture.More information about the STM32H7 series may be found on ST’s blog post or on it’s website. More information about ST’s Electronica presence, including the presentation program, can be found on the dedicated event pages.Meet also with the ST teams on the electronica Fast Forward startup platform.
Last edited by a moderator: