No sector will be more deeply affected by the current tensions between the USA and China than the semiconductor industry. Even if the Trump administration comes to a settlement over trade and cybersecurity issues tomorrow, and removes the latest restrictions on Huawei, China will still be reconfirmed in its determination to build its own chip industry and become technologically self-sufficient. The world’s second biggest economy cannot be subject to the kind of abrupt, and very public, actions that almost crippled ZTE last year, and have now been imposed on Huawei.
This will hit western semiconductor players in two ways. If sanctions persist, they will lose some significant customers. Even if they do not, they will lose Chinese business if ZTE, Xiaomi and others decide to protect themselves by migrating to local suppliers. And if normal trade relations resume in future, they will face powerful new competitors as the Chinese and Taiwanese chip companies gather strength, and target international markets.
None of this is simple, however black and white a situation US officials may be portraying since Huawei was placed on the US entity list (companies on the list are barred from procuring from US firms, and effectively, vendors from other countries also face US trade sanctions if they supply the banned firms). ARM quickly broke ties with Huawei, despite the Chinese vendor recently giving a huge boost to ARM’s server market ambitions by launching its own infrastructure chips. But it will be interesting to see whether Huawei is still able to deal with ARM China, a local joint venture set up specifically to give Chinese firms independent access to ARM’s IP.
Likewise, Intel and Qualcomm may publicly be pulling back from Huawei, but in the wider Chinese context, both companies have extensive investments, made over the past decade to ensure they were part of China’s bid to build a dominant homegrown semiconductor business. Intel has a stake in the group which owns several important Chinese companies, including modem maker Spreadtrum; Qualcomm has invested in start-ups and foundries.
While relationships like these are not unravelled overnight, despite the current tensions, it is likely that the USA and China will become the centers of increasingly separate technology islands – which could affect the move towards establishing truly global platforms in areas like 5G and cloud.
This is also likely to drive further consolidation as western semiconductor companies look for the scale to withstand reduced addressable markets and new competition. The industry has been in the grip of waves of mergers and acquisitions for several years now, with deals including Intel’s purchase of Altera and Marvell’s of Cavium.
Both these were geared to helping the acquirer build a full portfolio to address the key infrastructure markets of the early 2020s, 5G networks and cloud platforms from edge to data center.
5G will be rich in semiconductors, with complex technologies like millimeter wave antenna arrays and virtualized base stations. This will create opportunities for companies like ST Micro, which said at a recent investor event that it expected 5G to be a billion-dollar business, partly offsetting the difficult conditions in some of its core markets (see below).
For Intel, Altera brought FPGAs (field programmable gate arrays) to enable its buyer to surround its processors with accelerators to offload specialized tasks – essential for demanding use cases like virtualized RAN and artificial intelligence (AI). Cavium added a strong suite of cloud and network infrastructure processors to Marvell’s diverse range. Marvell has made several smaller purchases recently, to fill gaps in that platform, and has, in turn, sold some non-core assets – notably its WiFi and Bluetooth unit to NXP, earlier this month.
As well as 5G and cloud, many of the deals have focused on the automotive sector and some wider IoT markets, and NXP has been driven further towards those businesses since its plan to be acquired by Qualcomm failed. The latest proposed acquisition is Infineon’s of Cypress, which had itself acquired Broadcom’s IoT unit in 2016 (see below).
Acquisitions, new directions and, of course, a growing Chinese industry, will reshape the semiconductor landscape over the next few years, with a profound effect on how the platform and supply chain for 5G evolves.
ARM boosts high end cores by 40% with Intel and AI in its sights
ARM has been extending its licensable cores upwards into servers, and downwards into a wide range of IoT controllers and processors. But its main revenue generator is still the smartphone market, even though that is facing a slowdown in growth. ARM has a near-monopoly of smartphone processor IP, and it is unclear how possible it would be for China to develop an alternative, should the Huawei sanctions spread to other companies – and whether the existence of ARM China will be enough to keep the IP cores fully available to handset makers in that country.
Regardless of geopolitics, ARM has to keep pushing the boundaries of smartphone processor performance to keep its primary business robust. This month, it showed off three next generation cores for high end handsets, which are expected to be in commercial devices next year.
These are upgrades to current cores rather than a significant architecture change – they use the same 7-nanometer node and 3 GHz peak frequencies as the current generation, but some significant tweaks to the design can deliver performance improvements of 20% to 40%, ARM said.
There are three cores in the new x77 range – the Cortex-A77 CPU design, the Mali-G77 graphics core and the Mali-D77 display processor. They are upgrades to last year’s x76 family, which ARM said had sufficient performance to power laptops, coming within 10% of the capabilities of Intel x86 in that market.
The Cortex-A77 promises up to 20% more instructions per clock than the current A76, within the same thermal/power envelope and a similar or smaller area. ARM claims improvements including a 50% increase in instruction dispatch and integer-execution bandwidth, a 33% larger branch-target buffer, and a 25% increase in the out-of-order window size.
The Mali-G77 graphics core boasts about 40% overall performance boost thanks to its enhanced Valhall architecture, which features a simplified instruction set, quad texture mapper, and a 16-channel data path (compared to eight channels on the current G76). This should deliver about 30% faster game play with similar gains in battery life. The G77 can be configured with seven to 16 shader cores and two to four slices of L2 cache, totaling 512KBytes to 4MBytes.
The Mali-D77 display processor offloads tasks that could otherwise consume more than 15% of the cycles of the graphics core and is particularly optimized for VR acceleration. This element is targeting virtual reality headsets, to enable future models to support 4K resolutions at 90 frames per second (fps) or 3K at 120 fps.
However, the x77 does not represent the same leap forward as its predecessor, which as ARM’s first really to offer comparable performance to Intel Skylake. The A76 was more than an update for the high end Cortex-A line – it involved a full redesign aimed at mobile devices with PC-class performance. ARM said it delivered 90% of the performance of the Skylake mobile CPU, based on the Specint2006 benchmark, but with 25% of the footprint and half the power consumption. This compared to a norm of an ARM core delivering about 75% of the performance of an Intel contemporary.
That represented a far bigger leap than ARM usually delivers between generation of cores in the same family – 35% in performance terms compared to the 10-nm Cortex-A72 core, with a 40% reduction in power. ARM typically offers improvements of between 15% and 25% with its annual upgrades.
While it is important to keep pushing the CPU cores to support more demanding 5G tasks, the growth for ARM and its customers will be in accelerators to surround that CPU and handle even more intensive processes. As in the server market, there is a debate in the smartphone processor space over how many advanced tasks the CPU can handle as its performance grows, and how many need to be offloaded to the specialist chips, thus increasing the cost and complexity of the handset system-on-chip (SoC).
One of the main areas where accelerators are starting to be designed into high end, and even midrange, models is for AI tasks such as machine learning and vision processing. This demonstrates the challenge for ARM (one also experienced by Intel in infrastructure – can it retain control of the full handset IP platform by making its machine learning cores as dominant as its CPUs? It has had a good measure of success in the GPU (graphical processor unit) area – despite a slow start for its Mali GPU cores, it gained significant ground in recent years, and its main rival, Imagination Technologies, was dealt a serious blow when key customer Apple took its GPU design inhouse (though Imagination is still targeting the 5G GPU market, particularly looking to establish its ray tracing capabilities as a standard for advanced graphics).
Now ARM wants to chase an element of the smartphone platform which promises more growth and margin than the CPU – the ML processor, designed to enable mobile devices to do far more of the intensive processing required for AI-related tasks locally, rather than having to go back to the cloud. That would reduce the strain on connectivity, improving performance and responsiveness, and also help handset-focused chip suppliers to have greater control over the mobile experience, and the revenues it can generate.
But while most smartphone chipmakers have opted for ARM CPU cores – the largest ones able to customize them extensively through an architectural licence – it is less sure they will turn to a third party for their ML designs. Huawei, Apple and Samsung are all working on specialized deep learning accelerators, but it is not clear whether they will go for building or licensing for at-scale commercial offerings. Qualcomm, the smartphone SoC leader, has so far used a combination of GPU, DSP and other cores to support ML. It would be the key catch for ARM’s own ML cores, but Qualcomm has been developing more and more technology, such as its GPUs, without recourse to licensable cores. ARM said recently that several companies have licensed the ML cores which it unveiled last year as part of a major AI strategy, but it would not name them, and those cores are not yet in commercial devices.
In February 2018, ARM kicked off Project Trillium, to develop devices that will add advanced AI capabilities to mobile devices, particularly focused on ML and on object detection (OD). It will not have this field to itself. Google has its tensor processing unit (TPU), though on the handset side it is using Qualcomm technology; Intel is developing chips based on its Nervana acquisition; and CEVA and Ambarella are also active. A host of smaller start-ups will hope to pile on the pressure – including Cerebras, DeePhi, Graphcore, Horizon Robotics and Mythic, while even old CPU core rival MIPS has been acquired by ML-focused Wave.
The first ARM ML core was based on two elements – the Fixed-Function Engine (FFE) and the Programmable Layer Engine (PLE). ARM says that additional PLEs can be added to the design, to support non-convolutional layers – meaning that the chip can be customized to suit the need of the developer.
The ML Processor can provide 4.6TOPS, at an efficiency of 3TOPS per Watt (TOPS/W), which is a big step forward from using systems based on non-specialized chips such as CPUs, GPUs or DSPs, and improves cost and power efficiency.
Meanwhile, ARM OD has been built to recognize features and movements of human bodies, with technology based on the firm’s 2016 acquisition of Apical. ARM is keen on the idea of using both the ML and OD in the same device, using the OD chip to parse the incoming images (from a camera), which can then flag areas of interest that the ML processor can focus on.
The ML chips are not only geared to increasing ARM’s reach in the smartphone, but to pushing its designs into other devices such as smart cameras, VR headsets, drones, wearables, robotics and medical systems. Thus it is very important to the company that it sees these cores appearing in large-scale commercial devices next year.
And it will need powerful partners – one of them being Nvidia, which formed a deep learning alliances with ARM last year. The two plan to combine Nvidia’s open source Deep Learning Accelerator software and framework (NVDLA) with ARM’s recently announce Project Trillium – to create a chip design that can be used for ML and AI applications at the network edge.
This was the first example of ARM incorporating third party designs under its Trillium umbrella.
“Accelerating AI at the edge is critical in enabling ARM’s vision of connecting a trillion IoT devices,” said Rene Haas, and president of the IP Group, at the time of the announcement. “Today we are one step closer to that vision by incorporating NVDLA into the ARM Project Trillium platform, as our entire ecosystem will immediately benefit from the expertise and capabilities our two companies bring in AI and IoT.”
Nvidia uses ARM CPUs inside its SoCs, but has (until now) been pitching its offerings at larger devices – such as cars, drones, and cloud-computing arrays. With the ARM partnership, Nvidia is hoping that its presence in the low-power ARM ecosystem might help boost its adoption among IoT device developers – who might then use Nvidia hardware in future designs.
Infineon in huge deal for Cypress as auto and wireless drive sector’s M&A
Most of the discussion about 5G chips centers on processors and accelerators, but the intersection of 5G connectivity with the Internet of Things (IoT) and artificial intelligence (AI) – as well as expansion into far more spectrum bands and new antenna and power architectures – mean 5G will require a huge variety of chips. That will present opportunities for companies focused on RF, analog and other chips to increase their strategic position in 5G devices and networks.
Two big European chip groups, Infineon Technologies of Germany and French/Italian company ST Micro, are in this position. Both have significant activities in RF front ends, automotive, IoT, electro-mobility and other technologies which contribute directly or indirectly to 5G platforms. But companies of this kind have different decisions to make about their role in the 5G ecosystem, compared to the processor giants.
In the infrastructure processor area, providers are seeking to put together end-to-end platforms for 5G, to maximize their addressable market and enhance their strategic importance to network manufacturers. In the device SoC and modem space, the overall objective is to achieve economies of scale, as merchant chip suppliers face a situation where major handset makers are moving back to inhouse semiconductors.
The providers of RF, analog and 5G-surrounding chips also have the challenge of responding to the rise of inhouse developments by former customers – Qualcomm doing its own RF front end and DSP is an example. But they also have the opportunity of demand for a far wider variety of semiconductors as wireless connectivity – from Bluetooth to 5G – moves into many new devices and use cases. The downside of that is potential fragmentation of effort, and large firms need to prioritize the areas where they think they can achieve the best differentiation and market share, to create a rational portfolio, while offloading assets where they have less leadership, or which fit less well with other product lines.
Infineon has been going through this process as it formulates its strategy for 5G, wireless, automotive and IoT in the 2020s, building on its core strengths in security controllers, power chips and the auto and industrial sectors.
For instance, a year ago it sold its RF power business to Cree Wolfspeed for €345m, a deal which enhanced the latter’s ability to target 5G and achieve leadership in RF technology and packaging. This was not the optimum outcome for Infineon, which had attempted to buy Wolfspeed RF & Power from Cree in 2016 for $850m, but was prevented by the US Committee on Foreign Investment in the United States (CFIUS) on national security grounds. Without that deal, it made less sense for Infineon to stay in the business.
For Infineon, as the German firm’s CEO Reinhard Ploss said, the 2018 sale contributed to a process of identifying core growth opportunities, which he listed as “electro-mobility, autonomous driving, renewables and technologies for a connected world”. He said: “We will be able to focus our resources more effectively on Infineon’s strategic growth areas and will retain a strong technology portfolio for the wireless market.”
As part of that process, Infineon has also been on the acquisition trail. Last week, it said it would pay €9bn ($10.1bn) for Cypress Semiconductor. Based on its pro forma revenue reports of €10bn in fiscal 2018, the acquisition will make Infineon the eighth largest chipmaker in the world and number one in automotive.
The deal will improve its scale and competitiveness in that important automotive sector, especially connected and autonomous vehicles. Here, it competes with, among others, NXP (which scaled up its auto and IoT businesses when it acquired Freescale in late 2015). In particular, Infineon gains Cypress’s WiFi and Bluetooth connectivity chips for automotive and IoT, at a time when these are hot assets – NXP recently bought Marvell’s WiFi and Bluetooth business, while ON Semiconductor acquired WiFi specialist Quantenna earlier this year.
More consolidation is sure to follow as these large players, in Europe and Asia, respond to the deepening intersection between wireless connectivity, automotive and IoT technologies.
The Cypress acquisition is the largest in Infineon’s history and Ploss referred to it as “a landmark step in Infineon’s strategic development. With this transaction, we will be able to offer our customers the most comprehensive portfolio for linking the real with the digital world. This will open additional growth potential in the automotive, industrial and Internet of Things sectors”.
Cypress said it received several approaches and accepted Infineon’s cash offer of $23.85 per share because it represented “significant value creation”. It will certainly help the firm respond to a projected revenue decline of more than 10% in 2019, to about $2.2bn, as part of a larger group.
“For the past three years, our Cypress 3.0 strategy has delivered tremendous results and restructured the entire organization to focus on markets that matter,” said Steve Albrecht, chairman of the board of directors at Cypress, in a statement. “This transaction will create product opportunities that are increasingly important in the competitive automotive, industrial, and consumer markets.”
Cypress CEO Hassane El-Khoury said the deal would “capitalize on the multibillion dollar opportunities from the massive rise in connectivity and computing requirements of the next technology waves”. It will enhance the impact of Cypress’s connectivity portfolio by bringing it together with Infineon’s strengths in security and industrial systems. Infineon also gains a stronger presence in the USA and Japan, at a time of uncertainty with regards to ability to target Chinese markets, and it adds to its R&D resource in Silicon Valley.
Antonio Garzon, a senior analyst with IHS Markit’s automotive electronics and semiconductor group, told EE Times: “Half of Cypress’s $800m automotive business comes from infotainment microcontroller and connectivity solutions, but ADAS, where Infineon was already growing with radar and sensor fusion applications, will be also reinforced by the growing demand of NOR flash memory for autonomous driving.” Infineon will gain strategic assets in growth areas such as automotive memory while improving its position in the cockpit and infotainment elements of a vehicle.
If it gains all the necessary approvals, the deal should close around the end of this calendar year or in early 2020, about the same time as NXP expects to close its $1.76bn purchase of the Marvell assets. NXP’s messaging about that deal was similar to Infineon’s. NXP CEO Richard Clemmer said in a statement that combining the Marvell wireless assets with NXP’s embedded processing would “offer our customer base the broadest portfolio of edge solutions, which includes tailored security and a full suite of wireless connectivity spanning WiFi, Bluetooth, Bluetooth Low Energy, Zigbee, Thread and NFC”.
The Marvell WiFi and Bluetooth business employs about 550 people worldwide and it generated about $300m in revenue in Marvell’s fiscal 2019. NXP expects revenue associated with the acquired assets to double by 2022.
Marvell CEO Matt Murphy said the sale “yields a premium valuation and substantially higher economic return for Marvell shareholders while accelerating our transformation into a leading infrastructure supplier spanning 5G, data center, enterprise and automotive Ethernet applications”.
Millimeter wave may turn STMicro into a mobile chip company again
STMicroelectronics is not the first chip company that comes to mind when thinking about 5G, but it has put the mobile platform at the heart of its growth strategy, laying out a carefully targeted strategy to turn 5G into one of its projected billion-dollar businesses by the end of 2021.
At its recent 2019 Capital Markets Day, the Swiss-based French/Italian company said 5G was an important element of its goal to reach $12bn in revenue by the second half of 2021 – from $9.66bn in fiscal 2018, which was up almost 16% on the previous year. The company is targeting a 2019 revenue of $9.45bn to $9.85bn.
According to CEO Jean-Marc Chery, ST has 10 ‘macro programs’ which will help achieve this ambitious 2021 growth target, although acquisitions are widely expected to be necessary too, as well as a rising reliance on third party foundries to supplement ST’s own manufacturing base and enable it to expand into new products and sectors more quickly. “We will also acquire from time-to-time to strengthen the ST value proposition,” said Chery.
The strategy for 2021 is to break most of the business into 10 macro programs that would each generate $1bn in sustainable revenue, and which would have greater autonomy and agility than the current, more monolithic structure. The proof point for this new approach has been the imaging business, which was identified in 2017 as a strong opportunity, and which now generates $1bn in revenue. Chery said: “Two to three years ago, we identified wide bandgap materials as a unique opportunity for us to develop silicon carbide and gallium nitride for motion control. I have reviewed with the team other billion-dollar opportunities.”
The next major focus is 5G, both as a specific opportunity in areas like RF front ends, particularly for millimeter wave bands, and as a driver for growth in other sectors such as industrial networks and IoT. ST lists its current main areas of strength as automotive (the products split between car electrification and car digitalization), and industrial automation. Automotive and industrial end markets represent about 30% apiece of ST’s revenues, and these proportions are expected to remain stable for the next two years, but after that to be increased as a result of 5G investments.
Chery said: “For automotive and industrial, a key technology will clearly be 5G cellular data connectivity, rather than voice”, and the firm plans to become stronger in embedded processing to extend its reach in these core markets against rivals like NXP.
In the direct 5G market it is ambitiously targeting 50% market share (though with no timeline) in its element of the base station and network infrastructure segment, mainly by targeting millimeter wave components. ST is also moving into chips for handsets, though Marco Monti, president of the auto and discrete group, told the event: “For mobile handsets, we’d prefer not to give you a size, it’s a market that’s new for us. We’ll give a target when we start production end of this year.”
One development targeting the device side of 5G is gallium nitride (GaN) on-silicon RF power amplifiers for handsets and indeed, GaN-on-silicon is highly strategic to many of ST’s mobile initiatives, especially as it targets millimeter wave spectrum. It is the focus of a recently announced alliance with a US-based expert in this field, Macom (see below). GaN devices are generally faster and more efficient than silicon and silicon is reaching its limits, but while GaN has been widely used in optoelectronics products such as LEDs, it has been less commonly used in transistors. However, implementing GaN over silicon has started to be mainstream in power and RF applications, replacing LDMOS in the latter, though GaN still remains far away from processors.
The opportunity in direct 5G networks and handsets lies in the larger number of chips that will be required than in 4G, some of them important for ST, such as RF front end modules addressing new bands like mmWave. ST’s current revenue from front end modules in 4G is around $100m but it expects its mmWave technology to increase this substantially. Expertise in mmWave is highly strategic – ST believes this remains relatively rare and it can be applied to multiple segments, including mobile infrastructure and devices, small cells and repeaters (very important in mmWave 5G), and also in its satellite communication solutions, targeting mmWave phased array beamforming for tracking low earth orbit (LEO) satellites.
At this year’s Mobile World Congress, ST was focusing on 5G infrastructure and mmWave. Claude Dardanne, president of the microcontroller & digital IC group, said experience from the set-top box and satellite communications businesses will put ST in a good position in mmWave. He provided the following proof points:
- ST’s wide range of RF solutions including BiCMOS, CMOS, FD-SOI and RF-SOI
- Existing mmWave RF solutions offering RF transceivers up to 60 GHz
- ST pioneered beamforming for the satellite communication market, including RF front end, transceiver, and beamforming products
ST also still has some of the wireless products it co-developed under its SG-Ericsson joint venture, even though that was dissolved in 2013, and these can be repurposed for higher frequencies.
The mmWave market, especially for small cell backhaul and fixed wireless access, is still immature and dominated by designs from start-ups like BluWireless and Peraso. Analog Devices is also a player, but most giants are at the early stage, especially when it comes to applying mmWave to devices, so the field is wide open.
Partners and acquisitions will be necessary to enable the kind of expansion ST is eyeing in 5G-related markets. One of its collaborations is with Macom, with which it is co-developing GaN-on-silicon technology, initially for base stations. Among Macom’s areas of activity are transceivers and amplifiers for mmWave and optical systems.
The alliance with Macom will see ST expanding its 150-millimeter GaN-on-silicon production capacity in Sicily, Italy. The partners argue that GaN-on-silicon will be more central to 5G than the GaN-on-SiC promoted by competitors, offering better cost and scale. Macom CEO John Croteau said it aims to have “capabilities to service up to 85% of the global 5G network build-out”. ST brings the US firm its foundry services, and ST has licensed Macom’s IP on GaN-on-silicon manufacturing, while the two firms are also working together on new chip designs.
This should enable ST to introduce the GaN-on-silicon technology to segments that are not addressed by Macom, such as automotive and handsets. Driving these new materials into handsets is an opportunity for the mid-2020s, but ST argues that GaN-on-silicon will be more suited to mass market devices like handsets, than traditional gallium arsenide or Gan-on-SiC, because it has major cost advantages.
Despite this ambitious approach, it does rely on acquiring high market share in several areas which are likely to be highly competitive. ST is carefully targeting aspects of 5G that build on its existing expertise, including mmWave, but it will be in a crowded market and, having sidelined the mobile market in recent years, it is less entrenched in the ecosystem than some of its rivals, and RF specialists like Skyworks. A critical weakness is lack of experience in the very tight integration required for handsets and small cells, in particular. Qualcomm’s move into RF showed the advantage of controlling most of the key elements of the handset and integrating them to reduce cost and power but ST, despite its successful STM32 cores, lacks that integration expertise or the ability to combined the power and transceiver chips with basebands.
So the prospects for ST depend partly on how far the industry remains willing to buy components from multiple specialist suppliers and integrate them, rather than obtaining a single solution. Texas Instruments, once a giant of mobile infrastructure and device basebands and DSPs, is in a similar position. It exited the handset processor and baseband segments when its close alliance with Nokia broke down, and it has focused in recent years on analog businesses, and its strong markets are not in auto, IoT, sensors and mixed-signal. Like ST, it may become a bigger force in the mobile industry again as 5G impacts on its core sectors, and as 5G requires a wider variety of chips, but it has the same dependence on the 5G industry evolving towards a best-of-breed approach.
Jim McGregor, principal analyst at semiconductor specialists Tirias Research, told EETimes: “ST is a broad company in many markets with a variety of discrete analog and digital components. So I would envision ST being part of the solution for power, RF, and maybe even control through their MCUs[microcontroller units] or ASICs. They have long had components in these areas, so extending into the mmWave arena is not a surprise.”
Another challenge for ST is that it will need to scale up its design engineering and product marketing activities to support its new targeted products – or form further alliances like the one with Macom. In recent years, the firm has dramatically reduced its staff numbers and in areas like handsets, which have been deemed non-core, it has focused more heavily on developing intellectual property and patents than full products.
ST announces latest microcontrollers:
STMicroelectronics has launched its latest microcontrollers, targeting wireless and IoT applications. The STM32H7 dual-core microcontrollers are based on one 480 MHz ARM Cortex-M7 core, the top end of the Cortex-M family, and one 240 MHz Cortex-M4 core to handle lower-power tasks. ST says the MCUs hit the highest speeds in the sector at 1327 DMIPS, and to maximize energy efficiency, each core operates in its own power domain and can be turned off individually when not required.
The two cores can be used flexibly – typically, legacy code might run on the M4 core with new elements such as a more sophisticated user interface on the higher performance core. Or intensive tasks like neural networks, audio codecs or DSP filtering could be offloaded.
The key area where ST believes the new chips will succeed is in applications that combine artificial intelligence with real time control, which might be in smart home, factory or medical applications.
It says it has invested in significant new security features and provides 2MB of Flash and 1MB of SRAM on-chip plus fast access to external memory. Other features include error code correction (ECC) for all Flash and RAM memory to increase safety; external ambient temperature range up to 125°C for use in severe environments; and an Ethernet controller.
The STM32H7 dual-core microcontrollers are entering production and samples are available now.