Premio EDGEBoost Nodes
Industry 4.0 is driving the need for more intelligence at the rugged edge. Businesses and organizations want better access to real-time insights that can enable better business results. Especially with the convergence of the latest technologies in compute, storage, and wireless connectivity working in tandem, industrial computing designers need to evaluate the best design benefits and hardware requirements for reliable, low-latency performance.
Edge computing brings businesses and organizations tremendous benefits by processing data closer to the source of data generation. For example, edge computing reduces the amount of internet bandwidth required to process data, resulting in significantly lower latency when compared to cloud computing. Also, edge computing enables new applications that require extremely low latency. Furthermore, allocating more resources to the edge creates better workload utilization for real-time intelligence and automation. For this reason, Premio designs and builds rugged edge computing solutions that can be deployed in mobile, remote, and harsh environments that are not suitable for regular desktop computers.
When designing the RCO-6100 Series of AI Edge Inference Computers, the engineers moved outside of their comfort zone, ultimately creating a computing solution that addresses the critical hardware requirements for power-efficient computing at the edge. The RCO-6100 Series AI Edge Inference Computer is designed to deliver maximum processing performance, accessible ultra-fast NVMe SSD storage, and high-speed wired and wireless connectivity at the edge.
Introducing Premio’s EDGEBoost Nodes for Rugged Edge Computing
The RCO-6100 Series incorporates advanced performance thanks to the inclusions of Intel 9th Generation Core Processors, GPU support, and the availability of hot-swappable NVMe SSD storage. Powerful CPUs combined with powerful GPUs and ultra-fast NVMe SSD storage provide organizations with low latency data processing and the ability to run complex inference analysis in challenging settings.
The new two-piece modular design of the RCO-6100 Series allows organizations to boost the performance of the industrial computer for machine learning and deep learning inference analysis. EDGEBoost Nodes easily attach to the lower portion of the RCO-6100 Series, offering hardware acceleration for edge-level workloads that require data acquisition for real-time insights.
The two-piece modular design allowed Premio to maintain the ruggedness of the industrial PC module that houses the CPU, motherboard, and sensitive electronics. The bottom module provides performance acceleration via NVMe SSDs and GPUs. Each EDGEBoost Node uses powerful high-RPM active cooling in order to ensure reliability for powerful GPUs and NVMe SSD storage drives, preventing the system from overheating. As a result, system integrators can quickly scale edge deployments with the RCO-6100 AI Edge Inference computer with the necessary compute, storage, and connectivity requirements.
EDGEBoost nodes enhance the inferencing capabilities of the RCO-6100 Series by providing additional high-capacity, high-speed NVMe SSD storage and GPU performance acceleration for edge application deployments. In addition, NVMe SSD support and PCIe Gen 3 connectivity efficiently feed integrated CPUs and GPUs with volumes of data, enabling more reflexive inference analysis at the edge.
Premio’s leadership in rugged edge system design and hardware manufacturing introduces a new realm of IoT integration and automation capabilities, powering a fundamental transformation in how industrial businesses operate and compete.
Premio’s edge computing solutions offer robust performance while maintaining ruggedness, compactness, and energy efficiency. Learn more about the different ways you can configure the all-new RCO-6100 Series of AI Edge Inference Computers below.
EDGEBoost nodes are highly configurable performance boosters designed to meet the demand for complex applications that require powerful data processing, high-speed data storage, and inference capabilities at the edge. EDGEBoost nodes attach to the lower portion of the RCO-6100 Series, providing NVMe storage and GPU acceleration for complex edge workloads. EDGEBoost Nodes bring the latest technologies into hardware reality and offer new possibilities for machine learning and intelligent automation for the edge. Here are the three different EDGEBoost Nodes Premio offers:
EDGEBoost Node #1 - RCO-6141E-4U2C-2060S AI Edge Inference Computer
The first EDGEBoost node attaches to a base RCO-6100 Series and adds a hot-swappable NVMe SSD canister, capable of being populated with up to 4x lockable and hot-swappable 2.5” U.2 NVMe SSDs in 7mm height. The EDGEBoost node also adds PCIe expansion slots, enabling organizations and system integrators to add an Nvidia 2060 Super GPU for inference acceleration.
Adding NVMe SSD storage to edge computing solutions results in more reflexive AI inferencing capabilities since NVMe SSD storage via PCIe Gen 3 offers significantly lower latency than SATA SSDs. Additionally, this EDGEBoost Node can be configured with a graphics processing unit, enabling GPU acceleration. GPUs are capable of accelerating AI workloads, such as machine learning inferencing and deep learning inferencing, because GPUs are capable of processing significantly more data in parallel than merely relying on the CPU to process data.
EDGEBoost Node #2 - RCO-6141E-4U2C-HWR AI Edge Inference Computer
The second EDGEBoost Node focuses on ultra-high-speed NVMe storage but supports NVMe SSD storage media in 2.5” U.2 15mm form factors for higher capacities. In addition, this specific EDGEBoost Node offers x4 lockable and hot-swappable NVMe SSDs configurable in hardware RAID options in 0,1,5, and 10.
The second EDGEBoost Node offers reflexive AI inferencing capabilities and adds dense NVMe SSD storage for organizations with applications requiring high capacity and ultra-high-speed SSD storage. For example, autonomous vehicle and ADAS development applications require Terabytes of SSD storage to store high-resolution camera footage and sensor data so that such data can be used at a later time to train autonomous vehicles and advanced driver assistance system algorithms.
EDGEBoost Node #3 - RCO-6141E-8U2C-SWR AI Edge Inference Computer
The third EDGEBoost node focuses on ultra-high-speed NVMe Storage. It offers users the ability to add up to 8x lockable and hot-swappable 2.5” U.2 NVMe SSDs in 7mm height via two hot-swappable NVMe SSD canister bricks. The canister design allows organizations to quickly and easily remove all SSDs from the system to offload mission-critical data onto a central computer system. This allows for an easy and efficient way to transfer data from the edge into a location with resources reserved for training machine learning and deep learning models.
For applications that require the most amount of high-speed NVMe Storage, the third EDGEBoost Node is ideal as it can be configured with up to 8 NVMe SSDs, offering high capacity, ultra-high-speed solid-state storage at the rugged edge.
EDGEBoost Node #4 - RCO-6120-2060S AI Edge Inference Computer
The GPU-only EDGEBoost node offers organizations with GPU acceleration via an Nvidia RTX2060S Graphics Processing Unit. The GPU sits comfortably in a PCIe Gen 3 x16 Slot, offering organizations with blazing-fast GPU accelerated inferencing capabilities at the rugged edge.
Configuring your system with a GPU offers inference analysis acceleration at the rugged edge. This is possible because GPUs are much more capable of performing inference analysis than CPUs. In fact, GPUs can substantially increase the speed of deep learning inference and machine learning inference because they can perform a massive number of computations in parallel, thanks to the availability of thousands of GPU cores.
EDGEBoost Node #5 - RCO-6122 Fanless Industrial Computer
The riser option EDGEBoost Node provides users with expansion slot capabilities thanks to the availability of 2x PCI/PCIe expansion slots. Expansion slots provide system integrators with the ability to configure systems with add-in cards for industrial computing workloads. The EDGEBoost node offers three expansion options:
Option 1: 1x PCIe x16 and x1 PCI Expansion Slot
Option 2: 2x PCIe x8 Expansion Slots
Option 3: 2x PCI Expansion Slots
The key benefit of this EDGEBoost Node is that it offers organizations expansion capabilities by allowing them to populate the device with PCIe or PCI add-in cards. For example, organizations can add a capture card or any other PCIe/PCI expansion cards into the available slots.
Top 10 Must-Have Requirements for Rugged Edge Deployments
Although Premio’s RCO-6100 Series AI Edge Inference Computers provides a ton of performance at the edge, Premio understands that its solutions are often deployed in volatile environments that are unfit for regular desktop computers. As such, Premio has hardened the RCO-6100 Series to provide high-performance computing power while withstanding deployment in the world’s most challenging environments.
Edge Computing Hardware Requirements
Here are the hardware requirements that Premio engineers follow when designing and building rugged edge computing hardware:
1. Edge Computers Must Be Rugged & Fanless
Premio’s edge computing hardware is ruggedized to withstand deployment in harsh environments that are unsuitable for regular desktop computers due to the presence of dust, debris, shock, vibration, extreme temperatures, and other challenging environmental factors. The most critical design element Premio incorporates in edge computing hardware, including the RCO-6100 Series, is the fanless design, which eliminates all vents and openings from the system, creating a closed system resistant to dust, debris other small particles.
Moreover, eliminating the fans creates more reliable and durable computing solutions since fans are a common failure point for electronics, including industrial edge PCs.
Additionally, edge computing hardware features a wide operating temperature, ranging from -40°C to 85°C, enabling the deployment of the solution in environments that experience extreme temperatures. For example, systems are deployable in scorching hot environments such as the Mojave Desert during the summer or in Antarctica during the freezing cold winter. This is significantly different from regular desktop computers that can only withstand deployment in temperatures that range from 5°C to 35°C.
In addition to having a wide operating temperature range, edge computing solutions must have shock and vibration resistance. For this reason, Premio incorporates shock and vibration resistance into its systems. For example, Premio’s RCO-6100 Series features 50Gs of shock resistance and 5GRMs of vibration resistance in compliance with the MIL-STD-810G. Shock and Vibration resistance has been achieved by eliminating all cables from the system, reducing the number of moving parts that can fail, and equipping systems with SSDs (solid-state drives) instead of HDDs (hard disk drives).
2. Edge Computer Must Meet Performance Requirements
The RCO-6100 offers plenty of performance via Intel Core i3, i5, and i7 socket processors. Socket processors tend to provide more performance than their SoC counterparts. Edge computing solutions must be able to perform the tasks and workloads they’re being deployed to perform.
Underpowering systems with a low-tier CPU can result in sluggish performance, and in some cases, systems overheating and thermal throttling. That said, equipping your system with too powerful of a processor is a waste as you’ll need to ensure the system has a well-designed thermal solution to keep the processors from overheating.
3. Edge Computing Solutions Must Be Compact & Mountable
Edge computing solutions are compact because they are often deployed in space-limited environments that are too small for full-size desktop computers. As a result, edge PCs are designed to have a small footprint, permitting deployment in closets, cabinets, under furniture, mounted on walls, rails, ceilings, or other small spaces.
4. Edge PCs Must Be Equipped with Rugged High-Speed Storage
Edge computing solutions and, more specifically, AI edge computers are often deployed at the edge. They are tasked with processing large amounts of data, necessitating the need for large capacity, high-speed storage capable of keeping the CPU saturated with data. The RCO-6100 Series of AI Edge Computers can be configured with high-speed NVMe SSD, offering blazing-fast data transfer speeds, which are excellent for performing machine learning inference analysis at the edge
Additionally, equipping systems with SSDs (solid-state drives) makes the entire system more rugged because SSDs store data on silicon NAND chips vs. the spinning metal platters that HDDs (hard disk drives) use to store data, making them more rugged and able to handle exposure to shock and vibration.
5. Edge Computing Solution Must Have a Rich I/O
Rugged edge computing solutions, including the RCO-6100 Series, are equipped with rich I/O ports that allow systems to connect to both new and legacy technologies. Some standard ports you’ll find on rugged edge computing solutions include USB Type-A ports, Serial COM ports, Ethernet Ports (RJ45/M12), and GPIO ports, allowing systems to accommodate an extensive range of peripherals, sensors, and devices.
6. Edge Computers Must Have Rich Wired and Wireless Connectivity Options
Among the computer hardware needs for edge computers are both wired and wireless connectivity options. As such, edge computers come equipped with wired, wireless, and cellular connectivity options. Edge PCs come equipped with two RJ45 LAN ports for blazing-fast wired data transfer, ranging from 1 GbE to even 10GbE. Moreover, if wired connectivity is not available, systems can still connect to the network thanks to the availability of Wi-Fi 6 and 4G, LTE, and 5G cellular connectivity. Edge computing solutions have Dual SIM sockets, allowing for dual SIM cards to be inserted for redundant cellular connectivity.
7. Edge Computing Wide Power Range
Edge computing hardware must be equipped with a wide power range, enabling compatibility with various power input scenarios. Additionally, edge computers come with a number of power protection features that include overvoltage protection, overcurrent protection, and reverse polarity protection.
8. Edge Computing Solutions Must Be Secure
Edge computers are often deployed in remote environments where they are unmonitored, so they must be secure from being tampered with. For this reason, edge PCs are equipped with TPM 2.0 (trusted platform module) that utilizes a cryptoprocessor that makes systems tamper-resistant by securing hardware through integrated cryptographic keys. This protects the system from brute force attacks and hardware theft.
9. Edge Computers Must Offer Support For Performance Accelerators
Edge computers are great for collecting, storing, processing, and analyzing data at the edge; however, edge computers should be equipped with performance accelerators for real-time processing and decision-making for complex industrial workloads. New computing and storage designs maximize performance as close to the data as possible. Here are some of the most popular performance accelerators used in edge computing solutions as more processing power shifts to the edge.
A. Multi-core CPUs – Multi-core processors, provide more processing power than single-core CPUs because they are akin to having multiple processors on a single chip. Multi-core sequential processing allows the processor to utilize multiple cores to process data, with each core functioning as an individual processing device, permitting multiple tasks to run simultaneously (running more tasks at the same time). The more cores you have in a CPU, the better the system’s performance, as it will be able to handle multiple processes at the same time.
B. GPUs – Graphics Processing Units are often installed in industrial computing solutions to accelerate AI workloads such as machine learning and deep learning. The role of performance accelerators continues to increase as computing power shifts to the edge. Performance accelerators deployed at the edge can process mission-critical data in real-time with low latency since edge PCs are deployed close to the source of data generation. In addition, GPUs are extremely effective with real-time processing and inference analysis since they use a large number of cores for parallelism versus a sequential CPU.
C. VPUs – Vision processing units can be added to accelerate machine vision algorithms. VPUs are great for machine vision applications because they are optimized for machine vision, machine learning, facial recognition, and high-end image processing.
D. FPGA - A field-programmable gate array (FPGA) is a performance accelerator used to optimize embedded systems for a particular workload. FPGAs are able to accelerate workloads such as inference analysis, AI, and performing analysis of large amounts of data for machine learning. In some instances, high-end FPGAs can outperform GPUs in performing some tasks while using less power and producing less heat than GPUs.
E. NVMe Computational Storage – NVMe computational storage accelerates computing by performing data storage and processing on the storage drive itself. This means that data never has to move out of the SSD for processing, resulting in low-latency data processing.
10. Edge Computing Solutions Must Be Certified to Pass Telemetry to the cloud
The final hardware requirement for edge computers is that they must be certified to pass data telemetry to the cloud. Edge computing hardware provided by Premio Inc has been certified to pass data telemetry to the cloud by AWS (Amazon Web Services) IoT Greengrass and Microsoft Azure IoT.
Applications for Edge Computing Solutions
Edge computers are often deployed in factories and manufacturing facilities for industrial automation and control purposes. Edge computing hardware is used to enable communication between sensors, factory machinery, and other devices. Additionally, edge computers consolidate workloads by grouping multiple operations onto a single system, reducing the number of systems that must be managed and maintained. Furthermore, edge PCs are used to perform metrology and defect detection. Also, edge computing solutions can gather data from multiple sensors for metrology purposes and defect detection. They are much faster and more accurate at detecting defects than humans, resulting in greater efficiency and product quality.
2. Autonomous Vehicle Data Capture & Commercial Fleet Telematics
Autonomous vehicles and vehicles equipped with advanced driver assistance systems (ADAS) are heavily reliant on edge computing hardware to collect and store the data used to train AI models that guide vehicles. Training autonomous vehicles and ADAS for vehicles requires the collection of a large amount of data to train deep learning and machine learning algorithms that assist cars with driving, avoiding obstacles, and avoiding accidents. Training such systems typically requires a powerful computer capable of capturing and storing real-world data to train algorithms later. For this reason, powerful edge computing solutions are used to connect to, gather data from sensors, and store it to train models at a later time. Additionally, rugged edge vehicle computers are often deployed in fleet vehicles to pass data telemetry to the cloud. Edge computers offer CANBus support to integrate into a vehicle’s CANBus network to pass rich vehicle data to the cloud for remote monitoring of fleet vehicles.
Rugged edge industrial PCs are often integrated into mining facilities, equipment, and vehicles to remove operators from hazardous situations. Also, edge computers are replacing humans to perform mundane tasks that can be done more efficiently by industrial computing solutions. For example, automating mining tasks keeps mining operators out of harm’s way and instead moves them to safe underground control rooms to remotely control mining vehicles and equipment.
Edge computing hardware is often deployed to manage intelligent surveillance systems in challenging environments that are not friendly to regular desktop computers. Edge computers are used to gather, process, and analyze video footage, only sending footage that sets off specific triggers to the cloud for remote monitoring and analysis. This reduces the amount of required internet bandwidth since not all video footage has to be sent to the cloud; only specific clips that set of triggers are sent to the cloud for additional analysis and inspection. This is different from the traditional model, where all video footage was sent to the cloud for remote monitoring and analysis. Deploying edge PCs to manage smart surveillance systems is especially beneficial for those on metered data plans where they pay for the data used.