"Micro-Digest": 4 news from the world of developments for IoT

Today we have prepared a digest, in which we collected several news from the sphere of "chip building" for IoT. We will tell you about new devices for data encryption, the smallest computer in the world from IBM and about NVIDIA solution, which simplifies the integration of deep learning systems into microprocessors.
"Micro-Digest": 4 news from the world of developments for IoT

/photo Santi CC

The smallest computer in the world from IBM

At the IBM Think 2018 conference in March, the company introduced the smallest computer in the world. Its dimensions are are 1x1 mm, which is even less than a grain of salt.
The computer has a processor with several hundreds of thousands of transistors, SOS, a solar power supply system and a communication module with LEDs and a photodetector. The microcomputer will be equal in power to the processor with the x86 architecture of the 90s.
IBM told , that the new microchip will find application in blocking technologies: it will serve as a source of data for the blocking applications. For example, for logistics companies such a solution will help to detect the actions of fraudsters in supply chains.
In this case, it becomes possible to track the origin of goods. Low cost of chip production (about 10 cents) will allow massively build chips, for example, in electronic equipment, so that buyers can trace where the goods came from and make sure of its quality. In IBM are called its chip "crypto-anchor" (crypto-anchor), which protects data from theft and change.
In addition, the microchip can perform simple tasks for AI systems, for example, classify the data provided.
Release dates have not yet been named, but it is known that the developers are already testing the first prototype. ZDNet claims that the microchip is appears. on the market in a year and a half. In TechCruch predict the emergence of new items for 5 years.

The generator of truly random numbers from SK Telecom

Scientists from the South Korean company SK Telecom have developed a microcomputer capable of generating truly random numbers. Similar generators are already were created. earlier and even used in the work of cryptographic systems. However, the Korean company became the first who embodied this idea in the chip the size of 3х5х1 mm (LxWxH).
A tiny random number generator will be used in IoT devices to ensure the protection of encrypted data during their transfer to other devices.
The device uses the phenomenon of quantum shot noise (quantum shot noise). LED chips emit photons, which "bounce" from the internal walls of the device. They are captured by the built-in CMOS matrix , and the pulses generated by it are already are transmitted. the randomness-extraction algorithm.
SK Telecom and Nokia for the first time have demonstrated this technology in action last year. During the experiment, the SK Telecom server generated encryption keys and transferred them to the Nokia 1830 fiber optic switch.
The exact cost of the device is not called, however Sean Kwak, head of the quantum technology laboratory SK Telecom, says , that it will be "a few dollars".

Energy efficient chip for IOT-cryptosystems from MIT

At the Massachusetts Institute of Technology (MIT) have developed energy efficient microchip, which consumes 400 times less energy than software implementations of public key encryption. The device works 500 times faster.
Like most public-key cryptosystems, the chip uses the methods. elliptical cryptography . Moreover, it is is capable of work with any elliptical curves, and its blocks "responsible" for modular arithmetic can handle numbers up to 256 bits (classical systems work with 16- or 32-bit values). The Datagram Transport Layer Security protocol ( DTLS ), Which is responsible for processing encrypted data, is "wired" into the chip, which reduces the amount of memory required for its operation.
About testing and specific plans for using the device in MIT have not yet been reported.
/photo Fritzchens Fritz PD

Deep training in IoT: joint project of NVIDIA and Arm

Within the framework of cooperation, declared president of NVIDIA Jensen Huang, two companies decided to integrate the open architecture of NVIDIA Deep Learning Accelerator (NVDLA) into the Arm Project Trillium platform for machine learning. The joint project is designed to facilitate and accelerate the implementation of deep learning systems in mobile and IoT-devices.
NVDLA This is an accelerator for deep learning systems, which has an open architecture and is based on the NVIDIA Xavier processor. NVDLA is based on powerful NVIDIA tools for developers (drivers, libraries, SDK), among which will soon appear new versions of the programmable accelerator deep learning TensorRT.
As for the Arm processor, it is specially " sharpened "For working with machine learning systems. It performs more than 4.5 trillion operations per second (on mobile platforms), and this number can increase by 2-4 times with its "overclocking".
The companies hope that together these solutions are will help chip makers and developers to simplify the integration of AI systems into processors for IOT devices and provide the market with affordable products that support machine learning.
P.S. Materials from the First blog about the corporate IaaS:
How to smoothly migrate to the cloud: 9 useful tips
Cloud IT infrastructure: features of international projects
The Cisco UCS B480 M5 server blade: unboxing
P.P.S. A selection of posts from our blog on Habré:
Deep training in the cloud: optical computers will replace GPU
Why did computer chips become older "faster" and what to do with it
"Supercomputer" digest: 4 news from the world of high-performance computing
+ 0 -

Add comment