The edge refers to computational processes occurring close to the data sources at these locations – be they sensors, devices, or end-users – rather than relying on centralized systems that are common in traditional AI processing. This unlocks hyper-personalized experiences such as dynamic menu recommendations in quick-service restaurants, while also contributing to significant savings in data transfer bandwidth.
Supermicro and NVIDIA deliver Edge AI solutions that streamline deployment through pre-integrated components with optimized hardware and software for real-time processing.
Real-Time
From data analysis, monitoring, predictive maintenance, or personalization, AI at the edge will empower you to meet your business needs as they happen
Uninterrupted
Offline processing and inference enables uninterrupted operations in remote or disconnected environments
Personalized
Understanding and responding to customer preferences in a more natural and engaging manner to create innovative experience, powered by generative AI
![YouTube: [Supermicro] Delivering AI to the Edge](https://img.youtube.com/vi/IQuT2PC2nJ0/maxresdefault.jpg)
The Two Halves of Edge AI
The deployment of AI at the Edge can be for two distinct yet complementary purposes: predictive and generative AI. Predictive AI is the analytical powerhouse, leveraging historical data to forecast future events.
Generative AI, conversely, is a creative catalyst that uses underlying data patterns to synthesize new information, driving innovation and personalized experience.
Retail
Leveraging predictive AI to revolutionize retail spaces
- Asset Protection
- Autonomous Shopping
- Demand Forecasting
- Recommenders & Hyper-Personalization
In the retail sector, predictive AI is revolutionizing asset protection. By analyzing patterns and identifying potential threats, AI systems provide retailers with preemptive solutions to reduce losses and enhance store security.
These calculations must happen at the edge to facilitate the real-time responses that are required to respond effectively.
Product Use Case
SYS-110D-20C-FRAN8TP*
Our medium-size solution (SYS-110D-20C-FRAN8TP*) is perfect for retail stores to bring their asset protection analytics into a new world of predictive analytics.
* Additional models include SYS-110D-4C-FRAN8TP, SYS-110D-8C-FRAN8TP,SYS-110D-14C-FRAN8TP, and SYS-110D-16C-FRAN8TP.
Quick-Service Restaurants
Enabling restaurants to stay in front of the competition
- Speech-enabled food kiosks
- Food quality and operational efficiency
- Forecasting & inventory
Quick service restaurants are embracing generative AI to transform their customer experience. Speech-enabled food kiosks, powered by generative AI, offer personalized interactions, understanding and responding to customer preferences in a natural and engaging manner.
Generative AI can offer upsell and complimentary suggestions on the fly or offer human-like customer service right within the kiosk. When combined with predictive AI that learns from a customer’s loyalty card history, each subsequent visit can provide deeper customizations and more delightful interactions.
Product Use Case
SYS-E300-13AD
For quick-service restaurants, our small system (SYS-E300-13AD) would fit into many business models. Whether you are a high volume single building or a chain, this product can bring AI within your grasp.
Manufacturing
Streamlining end-to-end manufacturing challenges
- Workforce Health and Safety
- Automated Quality Inspection
- Mitigation of Skilled Employee Attrition
The manufacturing industry is leveraging predictive AI to address one of its most important and expensive challenges: workplace safety. Companies are using Edge AI hardware to connect data feeds from CCTV, machinery, and operational software to provide real-time predictive analysis of potential hazards.
Another significant application is automated quality inspection. For example, by continuously monitoring an automotive parts production line and feeding data into a predictive AI system, it is possible to provide real-time insights for quality control and significantly reduce the risk of defects.
Product Use Case
SYS-221HE-TNR(D)
Our extra-large (SYS-221HE-TNR(D)) system can make even the largest logistical challenges in manufacturing a breeze.
Healthcare & Medical Devices
Helping those who help others provide the highest quality care
- Patient safety and monitoring
- Diagnostic imaging
- Wayfinding
In healthcare, predictive AI is increasingly being trialed with diagnostic imaging, attempting to enhance the accuracy and efficiency of medical diagnoses by augmenting trained experts.
Another example, particularly in hospital settings, is to use generative AI for wayfinding - an application that enhances the patient and visitor experience. Digital kiosks and concierge services, powered by AI, guide individuals through complex hospital layouts, offering directions and information either on wall-mounted screens that follow the visitor’s progress or on updates delivered right to a mobile device.
Product Use Case
SYS-E403-13E-FRN2T
Healthcare comes in many shapes and sizes and our offerings can match those needs. Our large system (SYS-E403-13E-FRN2T) can assist hospitals or multi-site healthcare networks meet their patients where they are.
Smart Spaces
Elevating how people interact with other people and their environment
- Crowd and parking management
- Safety enhancement
- Access control
- Traffic management
Smart spaces, such as venues and stadiums, are adopting predictive AI for crowd and parking management. By analyzing real-time data, these systems effectively manage foot traffic and parking logistics, enhancing the overall customer experience while improving safety at the same time.
Product Use Case
SYS-E300-13AD
Our small offering (SYS-E300-13AD) will revolutionize your smart spaces. Deployable in standalone applications or in multiple sites across an entire venue you will be able to meet your audience needs and enrich their experience.
Small
SYS-E300-13AD
Single A2 or T100
| Dimensions (mm) H × W × D | 43.0 × 264.8 × 225.8 |
|---|---|
| Key Features for Predictive | AI computer vision for up to 8 streams1 |
| Up to ~2000 automatic speech recognition (ASR) samples per second2 | |
| Key Features for Generative | LLM up to 8 billion parameters |
| Image and video generation with Stable Diffusion at ~1 image every 4-5 seconds |
Medium
SYS-110D-20C-FRAN8TP3
Single L4 or RTX 4000 SFF Ada
| Dimensions (mm) H × W × D | 43.0 × 437.0 × 399.0 |
|---|---|
| Key Features for Predictive | AI computer vision for up to 16 streams1 |
| Up to ~6000 automatic speech recognition (ASR) samples per second2 | |
| Key Features for Generative | LLM up to 24 billion parameters |
| Image and video generation with Stable Diffusion at ~1 image every 2 seconds |
Large
SYS-E403-13E-FRN2T
Single L40S, L40, or RTX 6000 Ada
| Dimensions (mm) H × W × D | 117.3 × 266.7 × 406.4 |
|---|---|
| Key Features for Predictive | AI computer vision for up to 32 streams1 |
| Up to ~22,000 automatic speech recognition (ASR) samples per second2 | |
| Key Features for Generative | LLM up to 48 billion parameters |
| Image and video generation with Stable Diffusion at ~1-2 images per second |
Extra large
SYS-221HE-TNR(D)
Multiple L40S, L40, or RTX 6000 Ada
| Dimensions (mm) H × W × D | 88.9 × 436.9 × 574.0 |
|---|---|
| Key Features for Predictive | AI computer vision for up to 48 streams per GPU1 |
| Up to ~32,000 automatic speech recognition (ASR) samples per second per GPU2 | |
| Key Features for Generative | LLM up to 80 billion parameters |
| Image and video generation with Stable Diffusion at ~3 images per second |
- Based on using an image classification model similar to EfficientNet-B4, dependent on video stream compression and other workloads on the system
- Based on using an ASR model like QuartzNet
- Additional models include SYS-110D-4C-FRAN8TP, SYS-110D-8C-FRAN8TP, SYS-110D-14C-FRAN8TP, and SYS-110D-16C-FRAN8TP
Supermicro + NVIDIA: A Range of Solutions
At the heart of any Edge AI implementation is a robust and cohesive architecture that underpins the solution. Supermicro and NVIDIA’s collaborative approach provides just that – a comprehensive framework that integrates CPUs, GPUs, and optimized memory, all orchestrated within the resilient infrastructure of Supermicro’s systems.
The stack is engineered to ensure maximum efficiency and performance for deploying AI models at scale. Middleware solutions ensure that the systems are not only powerful but also intelligently coordinated and easily integrated into existing workflows.
Security, storage, and networking are foundational elements of this architecture, guaranteeing that data integrity and transmission are never compromised. The result is a scalable solution architecture that empowers end-users to unlock the full potential of Edge AI and edge computing.
Accelerating Adoption of Edge AI
