Loading...
Loading...
Loading...
AIAugust 8, 202510 MIN READ

AI at the Edge: Unlocking New Possibilities with Smart Compute

Fireside Chat-Hunter Golden Tile

Dive into this chat with OnLogic's Senior Product Manager, Hunter Golden, that will empower you with information to optimize AI and automation deployments. He’ll offer his expert insights on balancing performance, cost, and longevity while avoiding common pitfalls and ensuring tangible ROI. Watch the video below, or continue reading for the recap. 

What is edge computing?

Imagine all the places where computers are needed and all the use cases, not those that are cloud-based but "out in the field" where the action happens. That's edge computing. Over time, especially with the growth of edge devices and AI,  the need for computer power at the edge has grown tremendously, but the physical space available for the equipment hasn't. This means companies need more power in a smaller package. That’s what we mean when we say compute density.

AI: The game-changer at the edge

AI is making a powerful impact at the edge, especially through computer vision. By processing data at the source, OnLogic is helping customers realize significant gains in the following:

  • Throughput: How much work gets done.

  • Quality Control: Ensuring products meet standards.

  • Safety: Protecting workers.

AI helps by using data to reveal insights and boost productivity. For example, in worker safety, AI used to just check if someone was wearing a hard hat in real-time. Now, with more advanced AI algorithms, it can predict who might break safety rules and suggest ways to prevent accidents. Another example is preventative and predictive maintenance which can identify maintenance needs, through the data, before something goes wrong.

Smart choices: Balancing performance, power, and cost

When using AI at the edge, it's crucial to find the right balance between: 

  • Computer power

  • Energy usage

  • Computer cost

OnLogic emphasizes two key things here:

  1. Hardware Selection: Choosing the right computer parts.

  2. Workload: What the application is actually trying to achieve.

It's easy to "over-spec" hardware – meaning, buying a super powerful, expensive computer when a simpler, more affordable one would do the job. Many AI tasks and functionality can run well on regular computer processors (CPUs), integrated graphics (iGPUs), or special AI chips (NPUs), without needing a massive, expensive graphics card (discrete GPU). Understanding your application's goals, like how many video streams it needs or its target frame rate, helps you pick the right, cost-effective hardware.

Learn more about how to choose the right accelerator for your needs here. 

Clearing up AI misconceptions

There are two big misunderstandings about AI at the edge:

Misconception 1: You always need a discrete GPU for AI. 

Not true! As mentioned, many AI applications can run on less powerful, integrated hardware.

Misconception 2: AI can't work without a huge amount of cloud compute.

While powerful cloud data centers are used for training large AI or machine learning models (like ChatGPT), most AI at the edge is focused on "inference" (using the AI model), which requires much less computing power. In fact, some AI models can even run on a regular laptop.

Relying too much on the cloud for critical edge applications can lead to problems:

  • Latency: Data has to travel farther, causing delays because of bandwidth.

  • Data sovereignty: You lose direct control over your data.

  • Network reliance: If the internet goes down, your critical application stops working.

Modernizing your infrastructure with flexibility

Upgrading technology in industrial settings can be challenging because older machines are often still in use. That is why we offer legacy connectivity on our industrial computers. When modernizing, it's important to think not just about today's goals but also what you plan to do in the next 2 to 5 years. OnLogic helps customers plan for the long term so they don't have to replace hardware later.

A common problem is that hardware used for testing might not be available when it's time to fully deploy a project. OnLogic helps by offering guaranteed 5-year lifecycles for its products, ensuring you'll have the right hardware for your multi-year projects.

The secret to ROI: Setting clear objectives

Customers sometimes struggle when they don't clearly define their project goals upfront.

  • Over-specifying hardware can make the project too expensive to roll out widely, destroying the "Return on Investment" (ROI). 

As Hunter put it, "Getting one Ferrari approved might be fine, but getting a fleet of Ferraris approved might be a challenge".

  • Under-specifying hardware can lead to an application working in a lab but failing in the harsh real-world environment.

OnLogic helps customers find that "just right" balance by understanding their specific goals and how the application will be used.

Edge and cloud: A hybrid approach

While edge ai applications are "born" at the edge due to their need for low latency or data control, many companies are also moving workloads from the cloud to the edge because cloud costs can become very high. This trend is sometimes called cloud repatriation.

This hybrid approach combines the cloud and the edge makes a lot of sense. This allows companies to use the power of the cloud while still reaping the benefits of edge AI and having compute power right where the data is created.

Where to see the biggest return

The greatest ROI for upgrading infrastructure depends on each company's specific goals and challenges. For some, it might be improving worker safety. For others, it could be boosting throughput, reducing waste, or improving product quality. With AI, almost anything can be optimized to unlock massive opportunities.

Getting started with OnLogic

OnLogic works with customers across almost every industry, from smart agriculture, to manufacturing, to healthcare. We help companies get ready for AI and automation by offering scalable industrial computers and edge servers. This means you can start with a certain setup, test it, and then optimize it. For example, that might mean removing a discrete GPU if it turns out not to be needed, which significantly lowers costs for large-scale deployment. 

Ready to explore how AI and automation can transform your operations?  Explore our industrial hardware. 

Explore Hardware

Sarah Lavoie Content Creator for OnLogic
Sarah Lavoie
Content Marketing / Marketing, OnLogic
Sarah Lavoie is a content creator for OnLogic. When not writing, she can usually be found exploring the Vermont landscape with her camera looking to photograph something amazing.

Stay in the know.

Get first access to the newest products and Industry 4.0 innovation, OnLogic Live! events and more.