Connect with us

Featured

IFA: Huawei slices 7nm chip

At this year’s IFA, Huawei announced its TSMC 7nm process technology that will enable the Kirin 980 to pack 6.9 billion transistors within a 1cm die size, 1.6 times of the previous generation.

Published

on

In his IFA 2018 keynote titled “The Ultimate Power of Mobile AI”, Huawei Consumer Business Group CEO Richard Yu has introduced the Kirin 980, the system on a chip (SoC) that he says will bring about the next evolution of mobile AI. It is the world’s first commercial SoC manufactured with Taiwan Semiconductor Manufacturer Company’s (TSMC) 7nm process.

“Last year, we showed the world the potential of On-Device AI with the Kirin 970, and this year, we’ve designed an all-round powerhouse that not only features outstanding AI capabilities, but also brings cutting-edge raw performance to consumers,” said Yu. “Equipped with all-new CPU, GPU and Dual NPU, the Kirin 980 is the ultimate engine to power next-generation productivity and entertainment applications.”

Huawei provided the following information:

The cutting edge TSMC 7nm process technology enables Kirin 980 to pack 6.9 billion transistors within a 1cm2 die size, 1.6 times of the previous generation. Compared to the 10nm process, the 7nm process delivers 20 percent improved SoC performance and 40 percent improved SoC power efficiency.

The Kirin 980 is also the first SoC to embed Cortex-A76 based cores, which are 75 percent more powerful and 58 percent more efficient compared to their previous generation. The Kirin CPU subsystem uses an intelligent Flex-scheduling mechanism that creates a 3-level energy efficiency architecture consisting of two super-big cores based on Cortex-A76, two big cores based on Cortex-A76, and four little cores Cortex-A55. Compared with the traditional big.LITTLE design, this solution designates the large high-performance cores to handle immediate, intensive workloads; the large, high-efficiency cores to provide sustained performance; and ultra-efficiency cores to tackle everyday, light activities with extreme power efficiency. Running at higher clock speeds compared to the prior generation, Kirin 980 enables quicker app launch times, better multi-tasking and a generally smoother user experience.

As graphics in mobile games have become more and more sophisticated in recent years, Huawei has integrated the Mali-G76 GPU into the Kirin 980 to deliver unparalleled gaming experiences. Debuting with the Kirin 980, Mali-G76 offers 46 percent greater graphics processing power at 178 percent improved power efficiency over the previous generation. Mali-G76 also features an innovative clock boosting technology that utilizes AI to intelligently identify gaming workloads and adjust resource allocation for optimal gaming performance.

Industry’s First “Dual-Brain” Power

The latest Kirin SoC represents a new era of On-Device AI. The Dual NPU Kirin 980 elevates the On-Device AI experience with greater processing power and intelligence. The synergy between the Dual NPU results in an image recognition capability that is greater than the sum of two—the Kirin 980 can recognize up to 4,500 images per minute, up 120 percent compared to Kirin 970, further demonstrating Huawei’s industry leadership in the On-Device AI space. Additionally, Kirin 980 supports common AI frameworks such as Caffee, Tensorflow and Tensorflow Lite, and provides a suite of tools that simplifies the difficulty of engineering On-Device AI, allowing developers to easily tap into the leading processing power of the Dual NPU.

Full-Featured ISP

In pursuit of the best smartphone photography experience, Huawei integrated its proprietary fourth-generation ISP into the SoC. In addition to a 46 percent increase in data throughput compared to its predecessor, the new ISP also provides better support for multi-camera configurations, as well as an all-new HDR color reproduction technology that can manipulate picture contrast to highlight objects on various parts of an image. In addition, Kirin 980 utilizes the Multi-pass noise reduction solution that accurately removes artifacts without scrubbing away image details, resulting in better quality on images taken in low-light scenarios. Another new feature of the ISP is improved motion tracking. When a user attempts to snap a photo of a moving person, the ISP can still recognize the subject with 97.4 percent accuracy, so any user can capture fleeting moments with ease.

The rise of video-centric social media platforms brought together with it a surge of demand for video capture features. Kirin 980 adopts a new pipeline dedicated to processing video captures, allowing the camera module to shoot videos with 33 percent shorter delay.

World-Class Connectivity

To deliver the best connectivity to users of Kirin 980-powered devices, Huawei integrated the world’s first modem supporting LTE Cat.21 with a peak download speed of 1.4Gbps. Additionally, the Kirin 980 supports carrier aggregation, even across frequency bands, so users are free to choose their mobile operators and still enjoy the same premium connectivity experience, regardless of where they are.

Inspired by Nature

At IFA 2018, Huawei also announced the launch of two new gradient variants for the HUAWEI P20 Series: Morpho Aurora and Pearl White , expanding its range of gradient colors to four. When creating these two unique hues, Huawei once again turned to nature for inspiration. The Morpho Aurora was inspired by the Morpho butterfly, while the Pearl White exudes the same elegance as the iridescent mother-of-pearl.

 

Featured

Data gives coaches new eyes in sports

Collecting and analysing data is entering a new era as it transforms both coaching and strategy across sports ranging from rugby to Formula 1, writes ARTHUR GOLDSTUCK

Published

on

Coaches and managers have always been among the stars of any sports. They become household names as much as the sports heroes that populate their teams. Now, thanks to the power of data collection and analysis, they are about to raise their game to unprecedented levels.

The evolution of data for fine-tuning sports performance has already been experienced in Formula 1 racing, baseball and American football. All are known for the massive amount of statistic they produce. Typically, however, these were jealously guarded by coaches trying to get an edge over their rivals. Thanks to the science of “big data”, that has changed dramatically.

“American baseball has the most sophisticated data science analytics of any sports in the world because baseball has this long history of stats,” said Ariel Kelman, vice president of worldwide marketing at Amazon Web Services (AWS), the cloud computing giant that is working closely with sports teams and leagues around the world. “It’s an incredibly opaque world. I’ve tried for many years to try and get the teams to talk about it, but it’s their secret sauce and some of these teams have eight, nine or ten data scientist.”

In an interview during the AWS Re:Invent conference in Las Vegas last week, Kelman said that this statistical advantage was not lost on other sports, where forward-thinking coaches fully understood the benefits. In particular, American football, through the National Football League there, was coming on board in a big way.

“The reason they were behind is they didn’t have the player tracking data until recently in in the NFL. They only had the player tracking data three years ago. Now the teams are really investing in it. We did an announcement with the Seattle Seahawks earlier this week; they chose us as their machine learning, data science and cloud provider to do this kind of analysis to help figure out their game strategy. 

“They are building models predicting the other teams and looking at players and also evaluating all their practices. They are setting up computer vision systems so that they can track the performance of the players during their practices and have that inform some of the game strategies. The teams then even talk about using it for player evaluation, for example trying to figure out how much should we pay this player.”

Illustrating the trend, during Re:Invent, Kelman hosted a panel discussion featuring Rob Smedley, a technicalconsultant to Formula 1, Cris Collinsworth, a former professional footballer in the NFL and now a renowned broadcaster, and Jason Healy, performance analytics managerat New Zealand Rugby.

Healey in particular represents the extent to which data analysis has crosses sporting codes. He has spent four yearswith All Blacks, after 10 years with the New Zealand Olympic Committee, helping athletes prepare for the OlympicGames. 

“The game of rugby is chaos,” he told the audience. “There’s a lot of a lot of things going on. There’s a lot of trauma and violence and it can be difficult to work out the load management of each player. So data collection is a big piece of the technical understanding of the game.

“A problem for us in rugby is the ability to recall what happened. We have to identify what’s situational and what’s systemic. The situational thing that happens, which is very unlikely to be replicated, gets a lot of attention in rugby. That’s the sensational big moment in the game that gets talked about. But it’s the systemic plays and the systemic actions of players that lies underneath the performance. That’s where the big data starts to really provide some powerful answers. 

“Coaches have to move away from those sensational andsituational moments. We’re trying to get them to learn what is happening at that systemic level, what is actually happening in the game. How do we adjust? How do we make our decisions? What technical and defensive strategies need to change according to the data?”

Healey said AWS was providing platforms for tracking players and analysing patterns, but the challenge was to bring people on this technology journey.

“We’re asking our coaching staff to change the way they have traditionally worked, by realising that this data does give insights into how they make their decisions.”

Kelman agreed this was an obstacle, not just in sport, but in all sectors.

“Across all of our customers, in all industries, one of the things that’s often underestimated the most is that getting the technology working is only the first step. You have to figure out how to integrate it with the processes that us humans, who dislike change, work with. The vast majority of it is about building knowledge. There’s ways to transfer that learning to performance.”

Of course, data analytics does not assure any side of victory, as the All Blacks discovered during the recent Rugby World Cup, when they were knocked out in the semi-finals, and South Africa went on to win. We asked Healey how the data-poor South Africans succeeded where the data-rich All Blacks couldn’t.

“You have to look at how analytics and insights and all thesetechnologies are available to all the coaches these days,” he said. The piece that often gets missed is the people piece. It’s the transformation of learning that goes into the player’sactual performance on the field. We’re providing them with a platform and the information, but the players have to make the decisions.. We can’t say that this particular piece of technology played a role in winning or losing. It’s simply just a tool.”

The same challenge faces motor racing, which generates massive amounts of data through numerous sensors and cameras mounted in vehicles. Rob Smedley, who spent 25 years working in engineering roles for Formula 1 teams, quipped that his sport had a  “big data” problem before the phrase was invented. 

“We’ve always been very obsessive about data. Take car telemetry, where we’ve got something like 200 to 300 sensors on the car itself. And that goes into something like two to three thousand data channels. So we’re taking about around 600 Gigabytes of data generated every single lap, per car. 

“On top of that, where we’ve also got all the time data and GPS data. The teams are using it for performance advantage. We’re into such marginal gains now because there are no bad teams in Formula 1 anymore. Data analytics provide those marginal gains.”

• Arthur Goldstuck is founder of World Wide Worx and editor-in-chief of Gadget.co.za. Follow him on Twitter and Instagram on @art2gee

Continue Reading

Featured

IoT faces 5-year gap

Published

on

In five years, the world will have more than 40 billion devices. Locally, IoT specialist,Eseye, says that South African CIOs are recognising IoT (Internet of Things) and M2M (Machine to Machine) technologies as strategic imperatives, but the journey is still in its infancy.

“As legacy systems start to reach end of life, digital shifts will become inevitable. This, coupled with an increasing demand for improved bottom line results from existing and new markets, makes IoT a more viable option over the next five years. This is particularly prevalent in manufacturing, especially where time to market and product diversification has become necessary for business survival,” says Jeremy Potgieter, Regional Director – Africa, Eseye.

He says that within this sector one thing matters – output: “Fulfilling the product to market lifecycle is what makes a manufacturer successful. Addressing this functionality and production optimisation through technology is becoming more critical as they focus on increasing output and reducing downtime. By monitoring machinery and components in the production line, any concerns that arise, which impacts both the manufacturer and consumers alike, will be more efficiently dealt with by using an IoT approach.”

Potgieter says that there is also the growing strategic approach to increase the bottom line through new markets. As manufacturers seek new revenue streams, Eseye is encouraging the use of rapid IoT enabled device product development : “By addressing the connectivity aspects required at deployment, manufacturers are immediately diversifying their portfolios. Eseye, as an enabler, assists by providing market ready SIMs, which can be embedded into IoT connected devices at OEM level, connecting them to a plethora of services (as designed for) upon entry to market, anywhere in the world.”

In addition, Potgieter says that organisations are increasingly looking towards IoT connectivity managed services to capitalise on specialist expertise and ensure the devices are proactively monitored and managed to ensure maximum uptime, while reducing data costs.

Impacting IoT adoption though, is undoubtedly the network infrastructure required. Potgieter says that this varies significantly and will depend on criteria such as sensor types and corresponding measurements, the overall communication protocols, data volume, response time, and analytics required: “While the majority of IoT implementations can be enabled using cloud-based IoT platform solutions, the infrastructure required still remains important. A cloud platform will simplify infrastructure design and enable easy scaling capability, while also reducing security and data analytics implementation issues.”

Continue Reading

Trending

Copyright © 2019 World Wide Worx