Category Archives: Device Architecture

Kingston Digital, Inc., the Flash memory affiliate of Kingston Technology Company, Inc., a developer of memory products and technology solutions, today announced A1000 PCIe NVMe SSD. The M.2 drive is Kingston’s first entry-level consumer-grade PCIe NVMe SSD utilizing 3D NAND. A1000 delivers twice the performance of SATA at near SATA pricing.

The single-sided M.2 2280 (22mm x 80mm) form factor makes A1000 ideal for notebooks and systems with limited space. The PCIe NVMe drive features a Gen 3.0 x2 interface, 4-channel Phison 5008 controller, and 3D NAND Flash. It delivers 2x the performance of SATA SSDs with read/write speeds1 up to 1500MB/s and 1000MB/s giving it exceptional responsiveness and ultra-low latency.

“Kingston is excited to release its newest SSD for the entry-level PCIe NVMe market. Designed with 3D NAND Flash memory, A1000 is more reliable and durable than a hard drive, and doubles the performance of a SATA SSD. Now we can give consumers the benefit of PCIe performance at about the same price as SATA,” said Ariel Perez, SSD business manager, Kingston. “Consumers can replace a hard drive or slower SSD with A1000 and have the storage needed for applications, videos, photos and more.”

A1000 is available in 240GB, 480GB and 960GB2 capacities and is backed by a limited five-year warranty, free technical support and legendary Kingston reliability.

Silicon solar cells dominate the global photovoltaic market today with a share of 90 percent. With ever new technological developments, research and industry are nearing the theoretical efficiency limit for semiconductor silicon. At the same time, they are forging new paths to develop a new generation of even more efficient solar cells.

The Fraunhofer researchers achieved the high conversion efficiency of the silicon-based multi-junction solar cell with extremely thin 0.002 mm semiconductor layers of III-V compound semiconductors, bonding them to a silicon solar cell. To compare, the thickness of these layers is less than one twentieth the thickness of a human hair. The visible sunlight is absorbed in a gallium-indium-phosphide (GaInP) top cell, the near infrared light in gallium-arsenide (GaAs) and the longer wavelengths in the silicon subcell. In this way, the efficiency of silicon solar cells can be significantly increased.

Silicon-based multi-junction solar cell consisting of III-V semiconductors and silicon. The record cell converts 33.3 percent of the incident sunlight into electricity.  © Fraunhofer ISE/Photo: Dirk Mahler

Silicon-based multi-junction solar cell consisting of III-V semiconductors and silicon. The record cell converts 33.3 percent of the incident sunlight into electricity.
© Fraunhofer ISE/Photo: Dirk Mahler

“Photovoltaics is a key pillar for the energy transformation,” says Dr. Andreas Bett, Institute Director of Fraunhofer ISE. “Meanwhile, the costs have decreased to such an extent that photovoltaics has become an economically viable competitor to conventional energy sources. This development, however, is not over yet. The new result shows how material consumption can be reduced through higher efficiencies, so that not only the costs of photovoltaics can be further optimized but also its manufacture can be carried out in a resource-friendly manner.

Already in November 2016, the solar researchers in Freiburg together with their industry partner EVG demonstrated an efficiency of 30.2 percent, increasing it to 31.3 percent in March 2017. Now they have succeeded once again in greatly improving the light absorption and the charge separation in silicon, thus achieving a new record of 33.3 percent efficiency. The technology also convinced the jury of the GreenTec Awards 2018 and has been nominated among the top three in the category “Energy.”

The Technology

For this achievement, the researchers used a well-known process from the microelectronics industry called “direct wafer bonding” to transfer III-V semiconductor layers, of only 1.9 micrometers thick, to silicon. The surfaces were deoxidized in a EVG580® ComBond® chamber under high vacuum with a ion beam and subsequently bonded together under pressure. The atoms on the surface of the III-V subcell form bonds with the silicon atoms, creating a monolithic device. The complexity of its inner structure is not evident from its outer appearance: the cell has a simple front and rear contact just as a conventional silicon solar cell and therefore can be integrated into photovoltaic modules in the same manner.

EVG ComBond automated high-vacuum wafer bonding platform  (Photo courtesy of EV Group).

EVG ComBond automated high-vacuum wafer bonding platform
(Photo courtesy of EV Group).

The III-V / Si multi-junction solar cell consists of a sequence of subcells stacked on top of each other. So-called “tunnel diodes” internally connect the three subcells made of gallium-indium-phosphide (GaInP), gallium-arsenide (GaAs) and silicon (Si), which span the absorption range of the sun’s spectrum. The GaInP top cell absorbs radiation between 300 and 670 nm. The middle GaAs subcell absorbs radiation between 500 and 890 nm and the bottom Si subcell between 650 and 1180 nm, respectively. The III-V layers are first epitaxially deposited on a GaAs substrate and then bonded to a silicon solar cell structure. Here a tunnel oxide passivated contact (TOPCon) is applied to the front and back surfaces of the silicon. Subsequently the GaAs substrate is removed, a nanostructured backside contact is implemented to prolong the path length of light. A front side contact grid and antireflection coating are also applied.

On the way to the industrial manufacturing of III-V / Si multi-junction solar cells, the costs of the III-V epitaxy and the connecting technology with silicon must be reduced. There are still great challenges to overcome in this area, which the Fraunhofer ISE researchers intend to solve through future investigations. Fraunhofer ISE’s new Center for High Efficiency Solar Cells, presently being constructed in Freiburg, will provide them with the perfect setting for developing next-generation III-V and silicon solar cell technologies. The ultimate objective is to make high efficiency solar PV modules with efficiencies of over 30 percent possible in the future.

Project Financing

Dr. Roman Cariou, the young scientist and first author, was supported through the European Union with a Marie Curie Stipendium (HISTORIC, 655272). The work was also supported by the European Union within the NanoTandem project (641023) as well as by the German Federal Ministry for Economic Affairs and Energy BMWi in the PoTaSi project (FKZ 0324247).

Correction: A previous version of this article incorrectly state “imec” in the headline, instead of Fraunhofer ISE. Solid State Technology regrets the error.

The Semiconductor Industry Association (SIA), representing U.S. leadership in semiconductor manufacturing, design, and research, today announced worldwide sales of semiconductors reached $36.8 billion for the month of February 2018, an increase of 21.0 percent compared to the February 2017 total of $30.4 billion. Global sales in February were 2.2 percent lower than the January 2018 total of $37.6 billion, reflecting typical seasonal market trends. All monthly sales numbers are compiled by the World Semiconductor Trade Statistics (WSTS) organization and represent a three-month moving average.

“The global semiconductor market continued to demonstrate substantial and consistent growth in February, notching its 19th consecutive month of year-to-year sales increases and growing by double-digit percentages across all major regional markets,” said John Neuffer, president and CEO, Semiconductor Industry Association. “The Americas stood out once again, with sales increasing nearly 40 percent compared to last year, and sales were up year-to-year across all major semiconductor product categories.”

Year-to-year sales increased significantly across all regions: the Americas (37.7 percent), Europe (21.7 percent), China (16.4 percent), Asia Pacific/All Other (16.2 percent), and Japan (15.5 percent). Month-to-month sales increased slightly in Europe (0.9 percent), but fell somewhat in Japan (-0.9 percent), Asia Pacific/All Other (-1.5 percent), China (-2.6 percent), and the Americas (-4.3 percent).

For comprehensive monthly semiconductor sales data and detailed WSTS Forecasts, consider purchasing the WSTS Subscription Package. For detailed data on the global and U.S. semiconductor industry and market, consider purchasing the 2017 SIA Databook.

Feb 2018

Billions

Month-to-Month Sales                              

Market

Last Month

Current Month

% Change

Americas

8.63

8.26

-4.3%

Europe

3.40

3.43

0.9%

Japan

3.21

3.18

-0.9%

China

12.01

11.70

-2.6%

Asia Pacific/All Other

10.35

10.19

-1.5%

Total

37.60

36.75

-2.2%

Year-to-Year Sales                         

Market

Last Year

Current Month

% Change

Americas

6.00

8.26

37.7%

Europe

2.82

3.43

21.7%

Japan

2.75

3.18

15.5%

China

10.05

11.70

16.4%

Asia Pacific/All Other

8.77

10.19

16.2%

Total

30.38

36.75

21.0%

Three-Month-Moving Average Sales

Market

Sep/Oct/Nov

Dec/Jan/Feb

% Change

Americas

8.77

8.26

-5.8%

Europe

3.42

3.43

0.1%

Japan

3.21

3.18

-1.0%

China

11.90

11.70

-1.7%

Asia Pacific/All Other

10.39

10.19

-1.9%

Total

37.69

36.75

-2.5%

 

Nobuaki Kurumatani today took office as the first Chairman and CEO of Toshiba Corporation (TOKYO:6502) to be appointed from outside the company in over 50 years.

Commenting on his appointment as Representative Executive Officer and Chairman and CEO, Mr. Kurumatani said, “I am honored to be appointed CEO, and very much aware of the responsibilities I take on. Toshiba is not just any company. Its corporate DNA has realized countless Japan- and world-first technologies and products, made Toshiba a source of pride in Japan for nearly 145 years, and also made us a global leader.

“I believe that helping Toshiba back on its feet is my true calling. I am here at Toshiba to support change and transformation, and I see my role as to build on the company’s resilience and to lead its recovery. To secure growth, we must radically improve our earning power and reinforce our finances. We must move out of our comfort zone and promote fundamental reforms.”

Mr. Kurumatani most recently served as President of CVC Asia Pacific Japan (CVC). Before joining CVC in May 2017, he was Deputy President and a Director of Sumitomo Mitsui Financial Group, one of the largest financial institutions in Japan, where his career was devoted to corporate planning, public relations and internal auditing. He is a graduate of the University of Tokyo, where he studied Economics.

Satoshi Tsunakawa has taken on a new role in Toshiba as Representative Executive Officer and President, and Chief Operations Officer (COO). From today on, Mr. Kurumatani and Mr. Tsunakawa will together execute the management of Toshiba Group.

 SiFive, a provider of commercial RISC-V processor IP, today announced it raised $50.6 million in a Series C round led by existing investors Sutter Hill Ventures, Spark Capital and Osage University Partners alongside new investor Chengwei Capital, and strategic investors including Huami, SK Telecom and Western Digital and other companies that are among the most respected and iconic companies in the industry. This Series C round brings the total investment in SiFive to $64.1 million. Additionally, the company also announced it has signed a multi-year license to its Freedom Platform with Western Digital, which has pledged to produce 1 billion RISC-V cores.

This investment will enable SiFive to continue to innovate and provide leadership in bringing highly disruptive RISC-V technologies to the marketplace. “Over the past two years, SiFive has been at the forefront of the RISC-V ecosystem,” said Stefan Dyckerhoff, managing director at Sutter Hill Ventures and member of the SiFive board of directors. “Sutter Hill Ventures is confident that SiFive will continue to provide innovative solutions that will fundamentally change the semiconductor industry.”

Said Martin Fink, chief technology officer, Western Digital: “RISC-V delivers a platform for innovation unshackled from the proprietary interface of the past. This freedom allows us to bring compute closer to data to optimize special purpose compute capabilities targeted at Big Data and Fast Data applications. The next generation of applications like Machine Learning, AI, and Analytics require this ability to focus on a specific task. Western Digital is focused on the next generation of innovation to enable this new class of applications to deliver the possibilities of data.”

This Series C financing comes amid continued milestones for SiFive since its last round of funding in May 2017. Since then, SiFive has expanded its executive team with seasoned industry veterans including CEO Naveed Sherwani. The company also moved to a new, larger headquarters in Silicon Valley, a move that was prompted by a projected 3X growth in headcount.

“We are honored by the continued partnership with our investors and energized by new engagements with longtime industry leaders,” said Naveed Sherwani, CEO of SiFive. “This funding from our investors and licensing agreements with strategic partners establishes a strong financial foundation which will help us to continue our trailblazing path of engineering innovations and extend our market leadership around the world.”

SiFive’s mission is to democratize access to custom silicon through its IPs and platforms, globally. Since becoming available, HiFive1 and HiFive Unleashed software development boards have been deployed in more than 50 countries. Additionally, the company has engaged with multiple customers across its IP and SoC products, shipped the industry’s first RISC-V SoC in 2016 and the industry’s first RISC-V IP with support for Linux in October 2017.

Plastics are excellent insulators, meaning they can efficiently trap heat – a quality that can be an advantage in something like a coffee cup sleeve. But this insulating property is less desirable in products such as plastic casings for laptops and mobile phones, which can overheat, in part because the coverings trap the heat that the devices produce.

Now a team of engineers at MIT has developed a polymer thermal conductor — a plastic material that, however counterintuitively, works as a heat conductor, dissipating heat rather than insulating it. The new polymers, which are lightweight and flexible, can conduct 10 times as much heat as most commercially used polymers.

Researchers at MIT have designed a new way to engineer a polymer structure at the molecular level, via chemical vapor deposition. This allows for rigid, ordered chains, versus the messy, 'spaghetti-like strands' that normally make up a polymer. This chain-like structure enables heat transport both along and across chains. Credit: MIT News Office / Chelsea Turner

Researchers at MIT have designed a new way to engineer a polymer structure at the molecular level, via chemical vapor deposition. This allows for rigid, ordered chains, versus the messy, ‘spaghetti-like strands’ that normally make up a polymer. This chain-like structure enables heat transport both along and across chains. Credit: MIT News Office / Chelsea Turner

“Traditional polymers are both electrically and thermally insulating. The discovery and development of electrically conductive polymers has led to novel electronic applications such as flexible displays and wearable biosensors,” says Yanfei Xu, a postdoc in MIT’s Department of Mechanical Engineering. “Our polymer can thermally conduct and remove heat much more efficiently. We believe polymers could be made into next-generation heat conductors for advanced thermal management applications, such as a self-cooling alternative to existing electronics casings.”

Xu and a team of postdocs, graduate students, and faculty, have published their results today in Science Advances. The team includes Xiaoxue Wang, who contributed equally to the research with Xu, along with Jiawei Zhou, Bai Song, Elizabeth Lee, and Samuel Huberman; Zhang Jiang, physicist at Argonne National Laboratory; Karen Gleason, associate provost of MIT and the Alexander I. Michael Kasser Professor of Chemical Engineering; and Gang Chen, head of MIT’s Department of Mechanical Engineering and the Carl Richard Soderberg Professor of Power Engineering.

Stretching spaghetti

If you were to zoom in on the microstructure of an average polymer, it wouldn’t be difficult to see why the material traps heat so easily. At the microscopic level, polymers are made from long chains of monomers, or molecular units, linked end to end. These chains are often tangled in a spaghetti-like ball. Heat carriers have a hard time moving through this disorderly mess and tend to get trapped within the polymeric snarls and knots.

And yet, researchers have attempted to turn these natural thermal insulators into conductors. For electronics, polymers would offer a unique combination of properties, as they are lightweight, flexible, and chemically inert. Polymers are also electrically insulating, meaning they do not conduct electricity, and can therefore be used to prevent devices such as laptops and mobile phones from short-circuiting in their users’ hands.

Several groups have engineered polymer conductors in recent years, including Chen’s group, which in 2010 invented a method to create “ultradrawn nanofibers” from a standard sample of polyethylene. The technique stretched the messy, disordered polymers into ultrathin, ordered chains — much like untangling a string of holiday lights. Chen found that the resulting chains enabled heat to skip easily along and through the material, and that the polymer conducted 300 times as much heat compared with ordinary plastics.

But the insulator-turned-conductor could only dissipate heat in one direction, along the length of each polymer chain. Heat couldn’t travel between polymer chains, due to weak Van der Waals forces — a phenomenon that essentially attracts two or more molecules close to each other. Xu wondered whether a polymer material could be made to scatter heat away, in all directions.

Xu conceived of the current study as an attempt to engineer polymers with high thermal conductivity, by simultaneously engineering intramolecular and intermolecular forces — a method that she hoped would enable efficient heat transport along and between polymer chains.

The team ultimately produced a heat-conducting polymer known as polythiophene, a type of conjugated polymer that is commonly used in many electronic devices.

Hints of heat in all directions

Xu, Chen, and members of Chen’s lab teamed up with Gleason and her lab members to develop a new way to engineer a polymer conductor using oxidative chemical vapor deposition (oCVD), whereby two vapors are directed into a chamber and onto a substrate, where they interact and form a film. “Our reaction was able to create rigid chains of polymers, rather than the twisted, spaghetti-like strands in normal polymers.” Xu says.

In this case, Wang flowed the oxidant into a chamber, along with a vapor of monomers – individual molecular units that, when oxidized, form into the chains known as polymers.

“We grew the polymers on silicon/glass substrates, onto which the oxidant and monomers are adsorbed and reacted, leveraging the unique self-templated growth mechanism of CVD technology,” Wang says.

Wang produced relatively large-scale samples, each measuring 2 square centimeters – about the size of a thumbprint.

“Because this sample is used so ubiquitously, as in solar cells, organic field-effect transistors, and organic light-emitting diodes, if this material can be made to be thermally conductive, it can dissipate heat in all organic electronics,” Xu says.

The team measured each sample’s thermal conductivity using time-domain thermal reflectance — a technique in which they shoot a laser onto the material to heat up its surface and then monitor the drop in its surface temperature by measuring the material’s reflectance as the heat spreads into the material.

“The temporal profile of the decay of surface temperature is related to the speed of heat spreading, from which we were able to compute the thermal conductivity,” Zhou says.

On average, the polymer samples were able to conduct heat at about 2 watts per meter per kelvin – about 10 times faster than what conventional polymers can achieve. At Argonne National Laboratory, Jiang and Xu found that polymer samples appeared nearly isotropic, or uniform. This suggests that the material’s properties, such as its thermal conductivity, should also be nearly uniform. Following this reasoning, the team predicted that the material should conduct heat equally well in all directions, increasing its heat-dissipating potential.

Going forward, the team will continue exploring the fundamental physics behind polymer conductivity, as well as ways to enable the material to be used in electronics and other products, such as casings for batteries, and films for printed circuit boards.

“We can directly and conformally coat this material onto silicon wafers and different electronic devices” Xu says. “If we can understand how thermal transport [works] in these disordered structures, maybe we can also push for higher thermal conductivity. Then we can help to resolve this widespread overheating problem, and provide better thermal management.”

The semiconductor industry closed out 2017 in blockbuster fashion, posting the highest year-over-year growth in 14 years. Global semiconductor revenue grew 21.7 percent, reaching $429.1 billion in 2017, according to IHS Markit (Nasdaq: INFO).

Recording year-over-year growth of 53.6 percent, and its highest semiconductor revenue ever, Samsung replaced Intel as the new market leader of the semiconductor industry in 2017. Intel was followed by SK Hynix, in third position.

“2017 was quite a memorable year,” said Shaun Teevens, semiconductor supply chain analyst, IHS Markit. “Alongside record industry growth, Intel, which had led the market for 25 years, was supplanted by Samsung as the leading semiconductor supplier in the world.”

Among the top 20 semiconductor suppliers, SK Hynix and Micron enjoyed the largest year-over-year revenue growth, growing 81.2 percent and 79.7 percent, respectively. “A very favorable memory market with strong demand and high prices was mainly responsible for the strong growth of these companies,” Teevens said.

Qualcomm remained the top fabless company in 2017, followed by nVidia, which moved into the second position, after growing 42.3 percent over the previous year. Among the top 20 fabless companies, MLS enjoyed the highest market share gain, moving from number 20 to number 15 in the IHS Markit revenue ranking.

Figure 1

Figure 1

Memory was the strongest industry category

Memory integrated circuits proved to be the strongest industry category, growing 60.8 percent in 2017 compared to the previous year. Within the category, DRAM grew 76.7 percent and NAND grew 46.6 percent — the highest growth rate for both memory subcategories in 10 years. Much of the revenue increase was based on higher prices and increased demand for memory chips, relative to tight supply.

“The technology transition from planar 2D NAND to 3D NAND drove the market into an unbalanced supply-demand environment in 2017, driving prices higher throughout the year,” said Craig Stice, senior director, memory and storage, IHS Markit. “Entering 2018, the 3D NAND transition is now almost three-quarters of the total bit percent of production, and it is projected to provide supply relief for the strong demand coming from the SSD and mobile markets. Prices are expected to begin to decline aggressively, but 2018 could still be a record revenue year for the NAND market.”

Excluding memory, the remainder of the semiconductor industry grew 9.9 percent last year, largely due to solid unit-sales growth and strong demand across all applications, regions and technologies. Notably, semiconductors used for data processing applications expanded 33.4 percent by year-end. Intel remained the market leader in this category, with sales almost two times larger than second-ranked Samsung.

 

Mobile Semiconductor today announced the introduction of its three new 40nm ULP memory compilers which are available immediately.  The 40nm ULP compilers allow the engineer to create memory designs that maximize battery life while occupying the smallest amount of expensive silicon real estate. Mobile Semiconductor’s silicon-proven embedded memory technology offers these compilers on Taiwan Semiconductor’s 40nm ULP process.

These solutions are available in a range of formats that include:

  • Single Port, High Speed, Ultra Low Power
  • Single Port, Low Voltage (dual supply), Ultra Low Power
  • Single Port, High Speed, Ultra Low Power with a Reduced Mask Set

Founder and CEO Cameron Fisher states, “Mobile Semiconductor is one of the leaders in providing low power solutions.  The new 40nm ULP with flash allows the engineers to build products that may, for example, need periodic security updates and/or benefit from field updates to improve functionality.  Having on-board flash also serves to reduce the chip count on a board which is a further saving.  We support the process version with or without embedded Flash in any variant.”

The ULP process can lower power consumption by up to 30% while at the same time cutting leakage current by as much as 70%.  Overall performance is improved at virtually no cost to the customer.  Further, the high density 0.242 um2bit cell allows for reduced geometries, further reducing costs.

“The 40nm process technology has been around for a few years but the addition of Flash makes it applicable to a wider range of devices”, Fisher continued, “and the price points we are offering our 40nm ULP compilers sets Mobile Semiconductor apart from other memory compiler solutions. We are confident that our new 40nm ULP compliers are the perfect choice for wide range of new designs in the high-performance battery powered products market space.”

Combined sales for optoelectronics, sensors and actuators, and discrete semiconductors (known collectively as O-S-D) increased 11% in 2017—more than 1.5 times the average annual growth rate in the past 20 years—to reach an eighth consecutive record-high level of $75.3 billion, according to IC Insights’ new 2018 O-S-D Report—A Market Analysis and Forecast for Optoelectronics, Sensors/Actuators, and Discretes. Total O-S-D sales growth is expected to ease back in 2018 but still rise by an above average rate of 8% in 2018 to $81.1 billion, based on the five-year forecast of the new 375-page annual report, which became available this week.

In 2017, optoelectronics sales recovered from a rare decline of 4% in 2016, rising 9% to $36.9 billion, while the sensors/actuators market segment registered its second year in a row of 16% growth with revenues climbing to $13.8 billion, and discretes strengthened significantly, increasing 12% to $24.6 billion.  The new O-S-D Report forecast shows optoelectronics sales growing 8% in 2018, sensors/actuators rising 10%, and discretes growing 5% this year (Figure 1).

Figure 1

Figure 1

Between 2017 and 2022, sales in optoelectronics are projected to increase by a compound annual growth rate (CAGR) of 7.3% to $52.4 billion, while sensors/actuators revenues are expected to expand by a CAGR of 8.9% to $21.2 billion, and the discretes segment is seen as rising by an annual rate of 3.1% to $28.7 billion in the final year of the report’s forecast.  In the five-year forecast period, O-S-D growth will continue to be driven by strong demand for laser transmitters in optical networks and CMOS image sensors in embedded cameras, image recognition, machine vision, and automotive applications as well as the proliferation of other sensors and actuators in intelligent control systems and connections to the Internet of Things (IoT).  Power discretes (transistors and other devices) are expected to get a steady lift from the growth in mobile and battery-operated systems as well as good-to-modest global economic growth in most of the forecast years through 2022, the report says.

Combined sales of O-S-D products accounted for about 17% of the world’s $444.7 billion in total semiconductor sales compared to less than 15% in 2007 and under 13% in 1997.  Since the mid-1990s, total O-S-D sales growth has outpaced the much larger IC market segment because of strong and relatively steady increases in optoelectronics and sensors. However, this trend was reversed recently mostly due to a 77% surge in sales of DRAMs and 54% jump in NAND flash memory in 2017.

The 2017 increase for total O-S-D sales was the highest growth rate in the market group since the 37% surge in the strong 2010 recovery year from the 2009 semiconductor downturn.  In addition, 2017 was the first year since 2011 when all three O-S-D market segments reached individual record-high sales, says IC Insights’ new report.  The 2018 O-S-D Report also shows that sales of sensor and actuator products made with microelectromechanical systems (MEMS) technology grew 18% in 2017 to a record-high $11.5 billion.

NVIDIA and Arm today announced that they are partnering to bring deep learning inferencing to the billions of mobile, consumer electronics and Internet of Things devices that will enter the global marketplace.

Under this partnership, NVIDIA and Arm will integrate the open-source NVIDIA Deep Learning Accelerator (NVDLA) architecture into Arm’s Project Trillium platform for machine learning. The collaboration will make it simple for IoT chip companies to integrate AI into their designs and help put intelligent, affordable products into the hands of billions of consumers worldwide.

“Inferencing will become a core capability of every IoT device in the future,” said Deepu Talla, vice president and general manager of Autonomous Machines at NVIDIA. “Our partnership with Arm will help drive this wave of adoption by making it easy for hundreds of chip companies to incorporate deep learning technology.”

“Accelerating AI at the edge is critical in enabling Arm’s vision of connecting a trillion IoT devices,” said Rene Haas, executive vice president, and president of the IP Group, at Arm. “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.”

Based on NVIDIA® Xavier™, an autonomous machine system on a chip, NVDLA is a free, open architecture to promote a standard way to design deep learning inference accelerators. NVDLA’s modular architecture is scalable, highly configurable and designed to simplify integration and portability.

NVDLA brings a host of benefits that speed the adoption of deep learning inference. It is supported by NVIDIA’s suite of powerful developer tools, including upcoming versions of TensorRT, a programmable deep learning accelerator. The open-source design allows for cutting-edge features to be added regularly, including contributions from the research community.

The integration of NVDLA with Project Trillium will give deep learning developers the highest levels of performance as they leverage Arm’s flexibility and scalability across the wide range of IoT devices.

“This is a win/win for IoT, mobile and embedded chip companies looking to design accelerated AI inferencing solutions,” said Karl Freund, lead analyst for deep learning at Moor Insights & Strategy. “NVIDIA is the clear leader in ML training and Arm is the leader in IoT end points, so it makes a lot of sense for them to partner on IP.”