Tag: Tegra

As the Nintendo Switch 2 is set to launch on June 5th in Japan, the United States, and other regions, Nintendo released a special Creator's Voice video featuring NVIDIA CEO Jensen Huang explaining how the two companies collaborated on custom processors for the Nintendo Switch 2. Huang stated that NVIDIA and Nintendo have been collaborating for over a decade, sharing the same belief that technology should serve creativity, and that "fun" is a goal worth meticulously crafting. Huang also mentioned that former Nintendo president Satoru Iwata shared his dream game console: a console with the performance to run large-scale cinematic games, yet small enough to be portable. This seemingly impossible task ultimately led to the creation of the original Nintendo Switch. While Iwata couldn't actually participate in the Nintendo Switch's launch, the custom processor initially used in the Switch required over 500 engineers and a full year of development from NVIDIA. The entire system, from chip architecture and operating system to APIs and even the game engine, was completely rethought and redesigned to ensure the console could run high-performance games while also being portable. After the Nintendo Switch sold over 150 million units, attracting numerous gamers and families worldwide, assisting many independent game developers in unleashing their creativity, and redefining the game console, Nintendo and NVIDIA have once again collaborated to launch the Nintendo Switch 2, ushering in a new chapter. To achieve even greater goals, Jensen Huang stated that the custom processor used in the Nintendo Switch 2 is an unprecedented innovation for NVIDIA, incorporating three major technological breakthroughs: • The most advanced graphics processing capabilities in mobile device history • Full support for real-time ray tracing and high dynamic range (HDR) for brighter highlights and deeper shadows • A new architecture supporting backward compatibility Furthermore, the processor used in the Nintendo Switch...

During the recent CES 2025, many accessory vendors showcased peripherals labeled as compatible with the Nintendo Switch 2, and even exhibited what appeared to be a mock-up of the device. Earlier reports also revealed the anticipated design of the Nintendo Switch 2, including a larger 8.4-inch screen. The Joy-Con controllers will remain similar to the current models, using a magnetic attachment and allowing for quick release via buttons. Other design features include a power button, volume control buttons, a 3.5mm headphone jack, and retention of the current cartridge compatibility design, meaning the new console can still play previous Nintendo Switch games – a design previously confirmed by Nintendo. Regarding the processor, the Nintendo Switch 2 appears to still use a derivative of NVIDIA's Tegra processor, but this has not yet been confirmed. Nintendo clarified through Japanese media that the images and videos currently reported are not official and that further details will be revealed by Nintendo later. As for when Nintendo is expected to announce the Nintendo Switch 2, there has been no official statement yet. However, many sources believe that Nintendo will announce it as early as mid-January this year, with the earliest possible release date around March or April. Regarding preparations for the new console's release, Nintendo has stated that it has made certain preparations to avoid the shortages and scalping that occurred with the original Nintendo Switch. However, the specific methods used are still unconfirmed. Scoop alert🚨Here's a sneak peek at the Nintendo Switch 2 design! More details...

The endangered leopard cat, a native feline species in Taiwan, has received significant attention from relevant authorities in recent years. With the rise of urban and rural development, environmental and life threats to wildlife have increased, and the wild leopard cat population is now less than 500. Statistics show that between 2015 and 2018, at least one leopard cat roadkill incident occurred every month. In response, the Taiwanese startup DT42 (灼灼科技) collaborated with government and academic institutions using an AI image recognition system to protect local animals through more immediate and efficient methods, creating an environment where humans and animals can coexist harmoniously. The roadkill warning system, jointly developed by the Directorate General of Highways, the Endemic Species Research and Conservation Center, and the Department of Mechanical Engineering at National Chung Hsing University, was deployed in a demonstration area on Provincial Highway 3 in Zhuolan, Miaoli County, in May of this year. The system detects excessive vehicle speeds and simultaneously activates a roadkill warning system with built-in animal identification and audio-visual emission systems, prompting leopard cats currently on or preparing to cross the road to stay away, thereby achieving the goal of reducing roadkill incidents. This system is designed not only for the endangered leopard cat, but also to protect other animals that frequent this area, such as the white-nosed hearth dog and ferret badger. It may even reduce the risk of stray dogs and cats being struck by vehicles. Challenges – The training data for animal recognition is extensive and complex. The necessary technology and infrastructure for roadkill warning systems are not yet fully developed, making the initial data processing and setup extremely tedious and complex. The team initially used AWS cloud computing for animal recognition, but considering the bandwidth required for real-time recognition and data transmission, coupled with the overall slower-than-expected response time, it was necessary to further integrate analytical and computational capabilities into the terminal devices. DT42 Project Manager and co-founder Chung Wan-Chia stated, "Because vehicles travel at extremely high speeds, once the system detects a vehicle speed exceeding 40 kilometers per hour, it immediately activates the roadkill warning system. Animals' movements are unpredictable, requiring the generation of lights and high-frequency sounds to keep them away from the road and prevent collisions. If the roadkill warning system can activate in a shorter time, it can be truly effective in saving animal lives." Solution – Enhancing the learning efficiency of the recognition system through GPU acceleration, allowing the roadkill warning system to operate smoothly. To accelerate data analysis and processing, as well as the time and manpower spent on subsequent image data labeling and correction, the design team chose to adopt NVIDIA's Jetson TX2 platform designed for edge devices. This lightweight and power-efficient solution integrates CUDA and CuDNN computing architectures to improve training efficiency and, combined with TensorRT for inference acceleration, allows a learning model to be completed in just 3 hours and quickly deployed on edge computing devices, thereby enabling the roadkill warning system to operate smoothly. The Jetson TX2 platform allows terminal devices to complete preliminary analysis with higher efficiency, without relying entirely on cloud-based collaborative computing resources, and remains unaffected even when the device is offline. With the assistance of artificial intelligence technology, the system can reduce roadkill through image recognition combined with vehicle speed detection and other systems. GPU-accelerated image recognition can quickly determine whether animals are preparing to cross the road from image details, ensuring the roadkill warning system functions properly. The Endemic Species Conservation Research Center team has also installed fences in the demonstration area to prevent animals from crossing the road and cleared culverts for animals to walk through, guiding them to the other side of the road by other means, thereby reducing the roadkill rate. The actual implementation of the roadkill warning system is mainly integrated with existing buildings and will not damage the environment or affect the original landscape. However, Dr. Chang Chun-wei of National Chung Hsing University is concerned that malicious individuals may steal or damage the equipment, or use it to determine the possible locations of protected animals, leading to poaching and other problems. Therefore, the system facilities will be further disguised during the construction process, and the actual locations of these facilities will not be disclosed to the public. Impact – Future prospects: Artificial intelligence is expected to enable more animal conservation efforts. After the roadkill warning system was introduced on the demonstration road section, the number of leopard cat roadkill incidents has significantly decreased, with only one incident occurring within three months. The Ministry of Transportation is very satisfied with the results of this cross-sectoral cooperation project and expects to extend this application to more road sections prone to roadkill incidents in the future. Chiang Ya-yu, Assistant Professor of the Department of Mechanical Engineering at National Chung Hsing University, stated, "The test results of the leopard cat roadkill warning system are encouraging. We are actively collaborating with the government to replicate this successful experience in related animal conservation programs and promote it to other parts of Taiwan or globally, contributing to ecological conservation through AI." Animal conservation programs have become a global focus in recent years, with governments investing significant resources and many major technology companies offering more effective solutions through innovative technologies. DT42 believes that the future potential of applying artificial intelligence technology to animal conservation is considerable. Therefore, by starting from the grassroots level and collaborating with local farmers to identify root problems and address pain points in a demand-oriented manner, animal conservation can be achieved more efficiently without impacting the existing ecosystem and local development.

At GTC 2019, Foxconn showcased its solution applying computer vision technology to its production lines, making product inspection more efficient and improving overall production speed. The solution emphasizes an overall inspection accuracy of over 98% and reduces the required inspection staff from approximately 400 to just 2-3 people. Foxconn's AI-powered inspection system, currently deployed on its Wuhan production line, aims to improve upon traditional manual inspection methods. Manual inspection often suffers from decreased accuracy due to factors like visual fatigue, leading to lower product yields. Furthermore, considerations include optimizing labor costs and resource utilization, as well as improving production line efficiency. Therefore, the company hopes to enhance production efficiency through the implementation of AI technology. The solution showcased at GTC 2019 is primarily used for detailed inspection of product appearance. Foxconn stated that different technologies are also being applied in other production processes, and automation can be used to connect different production processes, allowing the entire production line to operate fully automatically with minimal human intervention. Furthermore, the production line system can automatically correct production errors based on inspection results. Regarding the inspection system design, Foxconn explained that while simple appearance checks can be performed using the Jetson system, the side panel inspection process for the main unit casing, which includes checks for appearance defects and whether the curvature of metal plate edges meets requirements, utilizes a Tesla T4 GPU system paired with multiple lenses for computer vision recognition. This achieves an accuracy rate of over 98%, surpassing the detail visible to the human eye. The system can now complete a workflow that previously required approximately 400 people, with the goal of increasing the number of items inspected at one time from 800 to 1200. In terms of learning and training, Foxconn pre-trains the system using defect example images. The learning time for recognizing new parts is generally only three days. While theoretically, with continued learning and experience accumulation, the learning time for recognizing similar parts can be reduced, the three-day training period remains the primary goal to maintain a certain level of accuracy. This recognition system is not only used on Foxconn's Wuhan production line but may also be applied to other production lines in the future, potentially even for the inspection of relatively simple and less complex parts. This situation illustrates that while AI technologies such as computer vision seem to have replaced some traditional human labor, the inspection of relatively complex parts, such as mobile phones, will likely still rely heavily on manual inspection at present. Factors such as the accuracy of current inspection systems and the overall cost compared to traditional labor costs may contribute to this. However, given Foxconn's future development goal of moving towards a fully automated production line operation model, it will undoubtedly continue to utilize AI technologies such as computer vision to further reduce production line operating costs. Besides Foxconn, more and more production lines, factories, and farms have begun to adopt artificial intelligence technologies such as computer vision. For example, Amazon started using robots to assist in picking and sorting goods a long time ago, and many modern farms have also begun to use computer vision technology to judge the yield of crops. This means that traditionally, work that was done by human labor can gradually be replaced by computer systems. However, in work that requires further judgment by the human brain and more precise operation, traditional human labor will still have its irreplaceable role. Traditional production line inspection relies heavily on manual labor and visual inspection, which can be affected by factors such as fatigue, impacting accuracy. In short, Foxconn's inspection system uses multiple lenses to capture image information and leverages Jetson AGX or Tesla T4 for accelerated computer vision learning. This allows the system to compare actual captured images with the specifications of manufactured parts. The internal structure of the inspection system is pre-programmed with example images of defects to enable deep learning. The actual operation process typically involves capturing images of a single part to check for defects in the coating, the curvature of the shape, etc., to ensure compliance with specifications. Through pre-learning, the system builds a basic judgment model, and the computer inspection system uses lenses and GPU components to accelerate the learning process.

The Magic Leap One Creator Edition, confirmed for release this August, has already garnered numerous user interface images on Twitter and Reddit, along with a growing number of gameplay videos on YouTube, giving viewers a glimpse into how this long-awaited augmented reality device works. The released images show that the Magic Leap interface will resemble virtual reality devices like Google Glass, displaying floating, flat images and allowing for gesture-based selection and content sharing via the "Cast" function. It also includes a feature to create virtual avatars. The content available for Magic Leap is broadly divided into two types: one is flat content presented in virtual windows, similar to how virtual reality devices can simulate a giant screen, even allowing multiple windows to be opened simultaneously; the other is immersive content that projects 8D virtual objects onto real-world scenes, creating interactive experiences – essentially augmented reality applications. Previously, it was officially confirmed that the Magic Leap would be powered by an NVIDIA Tegra X2 processor, but specific battery life details were not yet revealed. It was only stated that it would support gesture recognition, eye-tracking, and physical controllers. The device itself, in addition to the "Lightwear" headset, includes a "Lightpack," a computing unit similar in size to a traditional portable CD player. Through "Digital Lightfield" augmented reality technology, users can create various augmented reality virtual objects in their field of vision, presenting different virtual objects in relative positions (tracking the user's actual location through the glasses), and allowing users to hear the sounds produced by these virtual objects in a near-realistic direction. The Magic Leap One Creator Edition is expected to launch this summer, but a specific price has not yet been announced. For network connectivity, an exclusive partnership has been signed with US telecommunications provider AT&T, enabling the Magic Leap One Creator Edition to operate on AT&T's network resources.

Previously, it was rumored that virtual reality (VR) applications would be introduced into the Nintendo Switch. Nintendo had also revealed internal research projects involving VR and augmented reality applications. Furthermore, the discovery of what appears to be VR functionality in the Nintendo Switch firmware code suggests Nintendo's intention to enter the VR application race. According to Twitter user @daeken, strings resembling those for enabling and setting VR mode were found in the Nintendo Switch firmware code. It was determined that this code was not provided by NVIDIA, the processor supplier for the Nintendo Switch, and was likely added by Nintendo later. Prior to this, Shigeru Miyamoto, known as the creator of Mario, revealed that the company was already researching VR and augmented reality applications. Nintendo President Tatsumi Kimishima also expressed the company's interest in VR technology development, indicating Nintendo's interest in applying this technology to game content. Additionally, some patents suggest Nintendo will develop VR application accessories; for example, some patents previously indicated that Nintendo planned to use the Nintendo Switch tablet with a headset for VR experience modes. However, considering the current hardware specifications of the Nintendo Switch, its 6.2-inch screen with only 720p resolution seems difficult to match the usage mode of the Samsung Gear VR. But given its Tegra X1 processor's output performance, supporting more advanced virtual reality image output also seems challenging. Therefore, it's more likely that additional devices will be needed to provide higher performance. However, Nintendo's current Nintendo Switch accessories mainly focus on extending functionality or improving the experience; there are currently no accessories specifically designed to enhance performance. Therefore, whether accessories compatible with virtual reality display output will be released in the future remains uncertain. Just found this while looking at IPC interfaces on the Switch. Iiiiinteresting! #reswitched pic.twitter.com/J3QoKrJWKN — Cody Brocious ...

Update: CatFi clarified that August is currently only for trial production, and the final mass production and market launch dates will depend on the trial production results; therefore, there is no clear timetable at present. Regarding the actual market price, CatFi stated that it is still under internal discussion and has not yet been determined. The cat face recognition feeder currently uses a MediaTek processor, previously used the Rockchip RK3288 processor. The specific functions and analysis learning of the cat face recognition feeder are mainly assisted by cloud computing. Chi-Chun Technology, a company established in the British Cayman Islands two years ago, chose to dissolve and liquidate its debts due to changes in funding and organization. After clarifying its financial separation from the new funding, Chi-Chun Technology, which created the "CatFi" cat face recognition feeder, is now relaunching under the same name. During Computex 2017, they revealed that they will begin mass production of the revamped "CatFi" cat face recognition feeder in August, with shipments expected to begin as early as September. However, to address the issue of mass production, CatFi later opted to redesign the product and modularize some components to control production costs. The initial supply was prioritized only to Baker, who had previously participated in the crowdfunding campaign, before being released to the general market. The suggested retail price remained within the original range of $199 to $239, making the actual selling price more reasonable and avoiding a large discrepancy with production costs. Regarding the initial miscalculation of production costs that prevented the original design from being mass-produced, CatFi stated that it has subsequently obtained assistance through relevant resources and successfully improved the cost structure. CatFi stated that if Baker, who previously participated in the crowdfunding campaign, is not satisfied with the revised design, it will provide a refund. Regarding the new design, the biggest difference from the original cat face recognition feeder is the modular design. For example, the footrest for measuring the cat's weight is now an optional accessory, and most of the original metal and glass components have been replaced with plastic. Furthermore, the design, which previously used NVIDIA's Tegra processor for image recognition, now uses chips from other manufacturers, though it wasn't disclosed whether they partnered with MediaTek, Qualcomm, or other chipmakers. Regarding whether the design changes have altered the features emphasized in the original version, CatFi explained that while the footrest, which originally measured the cat's weight to determine its health, is now an optional accessory, facial recognition to identify weight gain or loss, combined with statistical analysis of recent feeding patterns, can still roughly estimate changes in the cat's weight and thus determine if the cat is sick. With the new modular design, the core cat face recognition feeder handles the basic functions, while the modular accessories serve as auxiliary tools to enhance feeding and analysis capabilities. Furthermore, despite the change in processor specifications, CatFi emphasizes that it can still successfully identify different cats in the home and individually record their feeding status and weight changes. Originally named Bistro and later renamed CatFi, the cat face recognition feeder design team utilized display processing power to accelerate learning, allowing the system to automatically identify which cat in the home is being fed, or to determine whether a cat is overfed, underfed, or potentially sick through weight measurement and facial recognition. This attracted considerable attention from cat lovers when it launched and opened crowdfunding, resulting in a significant amount of funding raised in a short period. However, because the design team primarily excelled in software but lacked experience in marketing and cost estimation, they misjudged the actual costs required for mass production and market launch, ultimately failing to ship the product and drawing criticism from many fundraisers.

The new NVIDIA Jetson TX2 development board, originally expected to be unveiled at CES 2017 alongside the new Shield TV, has been officially launched. It can be used in robotics, drones, and other embedded device designs, and while still integrating AI technologies such as image recognition, it further enhances on-device computing power, thereby increasing the widespread adoption of AI technologies. NVIDIA also confirmed that the Jetson TX2 uses the new Tegra processor architecture, codenamed "Parker," providing approximately twice the computing performance of the Jetson TX1 module. NVIDIA announced the new Jetson TX2 development board, emphasizing its new LTaaS (Lettuce Thinning as a Service) application mode and low power consumption of only 7.5W. The thinner and lighter design allows embedded devices such as robots and drones to achieve higher endpoint computing power, while providing more accurate image recognition, navigation, and voice recognition through AI technologies such as computer vision. According to Deepu Talla, Senior Vice President of NVIDIA's Tegra Business Unit, the Jetson TX2 will offer approximately twice the computing performance of its predecessor, the Jetson TX1. It also enhances overall computing power through integrated graphics. A key feature is bringing AI computing capabilities to embedded devices, allowing on-device computation to be performed more efficiently without relying entirely on cloud-based collaborative computing resources. This enables high-performance functions to run more easily, even when the device is offline. NVIDIA specifically explained that the processor used in the Jetson TX2 uses a system architecture codenamed "Parker," employing two special Denvor cores and four Cortex-A57 cores to form a unique six-core heterogeneous architecture. It also features 8GB of LPDDR4 memory and a 256-core Pascal GPU. Currently, Jetson TX2 partners include Cisco, Fellow Robots, Live Planet, and MIT. Applications include screen-sharing and online collaborative conferencing systems, store management systems that automatically detect inventory, 360-degree video stitching for 4K video streaming, and real-time image recognition for drones. The Jetson TX2 development kit is available for pre-order in Europe and the US starting today, with other regions to follow. In Taiwan, it is expected to be supplied via order, with a suggested retail price of US$599, and shipping expected to begin on March 14th. The Jetson TX1 will continue to be sold at US$499...

In addition to confirming the Nintendo Switch's release, Nintendo further revealed specific details about the new console, including a 6.2-inch 720p multi-touch screen, a custom NVIDIA Tegra processor (likely the "Parker" version), 32GB of built-in storage with microSD card expansion support, and a card slot design similar to SD cards for game storage, but it is not compatible with games from the Nintendo 3DS or Nintendo Wii U. The built-in battery provides approximately 2.5 to 6.5 hours of continuous use, connected via a USB Type-C port to a dedicated dock. According to Nintendo, the battery life varies depending on usage and game content, but is generally between 2.5 and 6.5 hours. The dedicated dock allows for the connection of three additional USB 2.0 ports and one HDMI output, while charging remains via USB Type-C. All computational performance is handled by the screen's built-in processing unit; the dock itself serves only as an expansion port and output device. The original box includes the Nintendo Switch itself, a dedicated dock, and gray left and right Joy-Con controllers (also available in red and blue). A separate accessory is also provided to combine the Joy-Con controllers into a single controller. The Joy-Con controllers feature analog sticks and multiple buttons, along with NFC, infrared sensors, and motion sensors to recognize forward gestures and operate in a Wii Remote-like manner. They also include vibration feedback for realistic haptic feedback depending on the game. In terms of accessories, a single Joy-Con controller is priced at 4480 yen, while purchasing both left and right Joy-Cons together costs 7480 yen. The Switch Pro Controller, designed for hardcore gamers, is priced at 6980 yen. Other accessories to be released include a charging controller that allows you to play and charge simultaneously, a dedicated steering wheel, and a Nintendo Switch-specific carrying case. In the future, Nintendo also plans to release different controller accessories for more game content and gameplay modes, allowing the Nintendo Switch to support more ways to play.

Update: According to information obtained by Laura Kate Dale, a freelance writer for LetsPlayVideoGames, from Nintendo, Ubisoft, and several independent game developers, the Nintendo Switch is confirmed to feature a touchscreen. However, the vast majority of games still rely on controllers, with touchscreen usage being relatively rare. It is expected that Nintendo's own games will incorporate some touchscreen functionality. Furthermore, the dedicated dock for the Nintendo Switch is confirmed to only provide additional output functionality, while the console itself still possesses considerable display output capabilities. However, Ubisoft has revealed that the console's battery life is only about 3 hours. Regarding the new next-generation console, the Nintendo Switch, expected to launch in March 2017, Nintendo revealed further details in an earlier interview with the Japanese magazine FAMITSU. These include the console still being positioned as a home console, the interchangeable left and right controllers being bundled together, the confirmation that the new console will not use disc storage media (i.e., games will only be played via cartridges or digital downloads), and the ability to enjoy smooth gameplay using the built-in battery. According to an interview conducted by Japan's FAMITSU magazine with Nintendo regarding the recently announced next-generation console, the Nintendo Switch, it has been confirmed that the new console will still be positioned as a home console, but it is not positioned as a successor to the Nintendo Wii U or Nintendo DS series, and it will not be compatible with existing cartridge games from the Nintendo DS series. Furthermore, since the Nintendo Switch itself and its peripherals do not include disc media access functionality, the new console will not support Nintendo Wii U or Wii game discs. Regarding the interchangeable left and right controllers mentioned in the video, as an important accessory for the console, they will be bundled with the original packaging. However, whether other components, such as the dedicated dock, will also be included in the bundle is currently uncertain; Nintendo only stated that adjustments may be made before the official release, and it cannot be confirmed at this time. As for the cartridge specifications used, Nintendo has not yet provided more specific details, but it is likely that the actual design differs from previous cartridge formats. Regarding game content, it's currently confirmed that the console will be compatible with cartridge-based games and digital downloads. Whether future digital downloads will further support games previously available on the Nintendo Wii U or Wii remains unclear, as Nintendo has not yet provided further details. Furthermore, given the Nintendo Switch's use of NVIDIA's new custom Tegra processor, coupled with partnerships with mobile game developers like DeNA and game engine developers like Unity, Nintendo has not commented on whether future mobile game releases on the platform are possible. However, regarding the Nintendo Switch's user experience on the go, Nintendo emphasizes that the new console, powered solely by its built-in battery, can maintain a comfortable and smooth gaming experience, but has still not provided details on battery life.