Next-Generation Computers Tech Digest - May 2017

The Machine

HPE (Hewlett Packard Enterprises) has revealed a prototype of its next-generation computer named ‘The Machine’. The prototype doesn’t offer all of the advances that HPE has been promising for the computer. Despite this, The Machine’s abilities are still impressive. It features 1,280 microprocessor cores – each of which work can co-operate with others simultaneously – that can utilise 160 terabytes of memory – which HPE says is roughly 10,000 times more than the average laptop. Information is transferred to different components with the use of optical fibres. The machine is still lacking memristors, a promised element which attracted the project much media attention. 

Nvidia GPU

Nvidia has released its Tesla V100 datacentre GPU designed for artificial intelligence (AI), high-performance computing and graphics. The chip uses the Nvidia Volta GPU architecture, to provide the performance of 100 CPUs. Nvidia claims that the Volta is the first GPU to break the 100 teraflops barrier for deep learning offering 30 times higher inference than a CPU server and allows for deep learning systems to be trained within a few days. 

Apple acquires Lattice Data

According to TechCrunch, Apple has acquired Lattice Data, a company that uses AI-enabled inference engines to take dark data and turn it into structured data. Dark data is a term that describes the estimated 70 to 80 percent of digital data that is unstructured (such as tables, text, figures and images), making it hard to use for processing and analytics. Lattice uses machine learning to order and categorise this data to make it amenable to analysis. Apple has not commented on the acquisition or the reasons behind it.  Lattice’s website has been taken down since the news broke. 

Computer simulation of human heart

NVIDIA and Barcelona Supercomputing Center (BSC) have presented a real-time interactive visualisation of a heart, showing the potential of HPCs (high performance computers) to simulate the cardiovascular system of humans. The demonstration used the Ayla simulation suite developed at BSC to work through the algorithms and mathematics behind the cardiovascular system, and NVIDIA’s IndeX to provide the in-situ visualisation. The ability to simulate these complex dynamic natural systems could allow for better personalised drugs, and new healing therapies while decreasing time taken, money spent and the number of animals used for testing. 

Bistable transistor

University of Illinois researchers have created a transistor laser that has two stable energy states and supports quick transitioning between them, meaning it could be used to boost computer processor speeds by delivering optical and electrical functions within a single switch. The device has been shown to be operational at -50°C. The professor leading the research, Milton Feng, believes that having entirely optical state transmission is unlikely to happen, because you always need electrical current to generate light in the first place. The next step in the research is to test the device at room temperature. 

Memristor photo generator

University of Michigan researchers have developed a memristor – an electrical resistor with a memory, able to store and process data simultaneously – computer circuit prototype that could potentially process complex data such as images and video at speeds substantially faster than today’s most advanced systems, according to scientists. This circuit uses pattern recognition to bypass the energy-intensive methods used by conventional image processing circuits. The research demonstrates the use of sparse coding – a method that humans, as well as AI, use to quickly analyse input data based on salient, categorizable characteristics – on a 32-by-32 array of memristors to analyse and recreate several photos. The team hopes to scale up the system to make it practical for real-time analysis of video content from sensors and cameras. 

3D BiCS FLASH memory

At the Dell EMC World 2017 conference Toshiba announced its 3D BiCS FLASH memory technology consisting of 64 layers to allow for 64GB of data to be stored on a single chip. 3D BiCS FLASH memory vertically stacks memory on top of other memory increasing the amount of data that can be held compared to 2D NAND flash configurations. According to the company, the spaces between memory cells in the BiCS FLASH are also wider than those in 2D NANDs, making it possible to increase the processing speed by increasing the amount of data for a single shot programming sequence. Toshiba additionally says its 3D BiCS FLASH technology consumes less power than 2D NAND. These advantages appear promising for smartphones, tablets, memory cars and other battery powered devices. 

Fujitsu and 1QBit team up

Fujitsu, a Japanese electronics company, and 1QBit, a Canadian tech company, announced that they are collaborating on using quantum-inspired technology for AI. Software developed by 1QBit’s quantum computer machine learning software will be able to use a digital annealer created by Fujitsu and the University of Toronto. A digital annealer is a computing architecture that can solve combinatorial optimization problems using current semiconductor technology. The aim of the collaboration is to use Fujitsu’s computer architecture and hardware technology with 1QBit’s software to enable machine learning to solve complicated, large scale optimisation problems. It is hoped that these advances could help businesses solve computing challenges in many fields including finance, the life sciences, energy, retail and distribution. 

 

Add this: