Using a GPU server for AI and automation

What is a GPU server? 

A GPU server uses the processing power of a dedicated GPU (Graphics Processing Unit) to process large data sets quickly and efficiently. GPUs are designed for parallel processing and were originally created to speed up the rendering of 3D graphics for immersive games and other video functions. This parallel computing lends itself well to heavy graphics workloads such as image recognition. In addition, it accelerates the training and inference of neural networks for deep learning-algorithms. 

GPU servers are used in numerous industries to manage the ever-increasing amounts of visual data collected. If you need to handle AI, machine learning and automation workloads at the edge, you need the power of a GPU server. 

Example: Visual inspection with AI

In a manufacturing environment, visual inspection helps ensure that products meet specifications and quality standards. Companies use AI for visual inspection to intercept product deviations that can be easily missed during human inspection. 

Miniserver

Imagine a manufacturing facility that wants to implement a quality control system directly on a production line. The environment may be exposed to temperature fluctuations, dust and vibration. Therefore, a ruggedized computer, such as the Karbon 700 (K700-X2), a good choice to drive computer vision. Check out our customer case study on Artemis Vision If you want to know how they implemented the Karbon 700 as a miniserver for one of their vision inspection solutions.

Rackmount edge server 

For larger production needs with multiple lines that each need to be inspected, you can install an industrial PC at each production line. Another option is to have all lines monitored and controlled by one high-performance rackmount edge server. Edge server high-performance computing can support multiple production lines with one piece of hardware. 

GPU server

We can take that one step further. Suppose you have even more complex workloads on each line that require deep learning analysis or other advanced algorithms. You can send your data to the cloud for analysis, but that leads to additional latency and - depending on bandwidth requirements - can be expensive. 

To avoid latency and operational costs, many organizations use cloud repatriation to move their computing resources onsite. An edge server configured with a GPU has the processing power to meet the requirements for this scenario.

Edge server with GPU for automation

When it comes to automation, such as a SCADA System server, you can use an edge server with a GPU to process even more tags and more HMI clients connect than was previously possible. These high-performance edge servers give you the ability to run SCADA solutions, such as Ignition from Inductive Automation execute. In addition, you can further extend these applications thanks to improved GPU computing power. 

For example, you can take a time series of PLC tag data directly from the historian database and feed it into an AI-driven predictive maintenance program, taking advantage of the built-in GPU. That has the added benefit of lower cloud transmission costs plus the fact that confidential data is kept onsite. 

You also have the option for additional management tools, such as creating a high-availability Kubernetes cluster for mission-critical operations. 

Installation environment concerns

There are different types of edge computing workloads, and environments in which they are installed can vary widely. This is why edge computers come with different specifications and in a variety of models, from fanless or fanless hybrid computers, mini servers, rackmount edge servers and GPU servers. 

A fanless PC works reliably in demanding environments. This makes it a good choice if it must be used in harsh conditions such as dust, high temperatures, impact forces and vibration.

Most edge servers require some cooling and protection from dust. OnLogic offers high-performance edge servers designed for the edge, such as the Axial AC101 Edge Server. This flat 1U server offers a number of features for AI and automation, with an operating temperature range of 5 ~ 40 °C (ASHRAE A3). However, you still need a semi-controlled environment to take advantage of that level of computing power at the edge.

In some cases, you can implement hardware in a communications cabinet or the rack of a local server, where it is protected from the elements. But sometimes you need an edge server right next to equipment for applications with extremely low latency, such as high-speed robotics. In that case, we can provide enclosures that provide cooling and dust protection. 

Another option you can look at are the mini servers. These offer many of the same computing features as traditional rack servers, but in a smaller size specifically designed to hold up in challenging conditions. 

What features should you consider for an edge server or GPU server?

CPU performance at the edge

When people start thinking about a server, CPU performance is usually the first thing they look at. Some applications require scalable Intel® Xeon® CPUs. For other applications, modern Intel Core™ CPUs, which integrate many of the features of the previous generation of scalable Xeon processors, offer a more economical, yet extremely powerful alternative. Axial AC101 offers the latest Intel 13th generation processors with hybrid architecture and up to 24 cores and 32 threads.

Limitations of GPU power and space.

For AI workload acceleration, a PCIe GPU is usually a primary requirement for performance. The AC101 is a shallow server that supports a full-height, full-length GPU. 

• This is important because many rack enclosures in industrial environments do not support full servers. In addition, you may be limited by the total number of rack units available. A 1U may be the only option when you need a UPS, network switch and other equipment.
• There are several frameworks that can take advantage of GPU computing power for AI. Examples include Nvidia's CUDA framework that runs primarily on their GPU architecture and Intel's OpenVino platform that can use CPU, iGPU and Intel-based GPUs.

Baseboard Management Controller (BMC).

With a BMC you can control the server at remote locations. This can be essential in a distributed architecture at the edge. Effective device management on a large scale becomes essential. Unlike many other servers on the market, the AC101 does not require users to pay additional licensing fees to take advantage of the full capabilities of the BMC. 

Protection against sabotage

Edge systems cannot always be implemented in areas with access control. To solve this problem, the AC101 comes with a lockable security front and restricted front I/O. This prevents access to the 4x internal U.3 (NVMe) or SATA SSDs and 2x M.2 SSDs.

Want to know how to use a GPU server for your AI and automation? Explore our full selection computers with GPU And contact our team!

Editor's note: Inductive Automation has discontinued their Ignition onboarding program. Ignition licenses must now be purchased directly through Inductive Automation. Although IGN versions of our solutions are no longer available, our computers remain excellent for use with Ignition software. Discover our recommended hardware here.