VersaLogic Validates AI Capability with Hailo

Press Office, VersaLogic Corporation, 06/18/25

As artificial intelligence continues to transform embedded systems, VersaLogic is keeping a persistent effort to ensure its products stay ahead of the curve. Most recently the company conducted a comprehensive evaluation of Hailo’s AI acceleration chips on their compact and rugged embedded computer, called “Swift”. The goal? To validate that adding AI capability to this VersaLogic product is not only feasible—but powerful and efficient.

VersaLogic’s approach to benchmarking new AI hardware is rigorous and application-focused. The process begins by identifying real-world AI tasks relevant to their customer base, such as object detection, classification, and anomaly detection. They then select industry-standard test models to gauge inference speed, power consumption, and thermal performance under edge deployment conditions.

In the Hailo integration test, VersaLogic conducted a series of hands-on evaluations using different Hailo-8 AI accelerators paired with the Swift embedded computer. The tests began with the Hailo-8R mPCIe card, which delivers 13 TOPS (Tera Operations Per Second). The Swift successfully reached the full 13 TOPS in testing.

Next, they swapped in the Hailo-8L M.2 E-Key card, rated at 26 TOPS. That board achieved 24 TOPS in the testing regimen, very close to the rated performance. This was followed by the Hailo-8L M.2 B-Key card, also rated for 26 TOPS, which similarly yielded 24 TOPS in testing.

Finally, VersaLogic installed both the mPCIe and E-Key cards simultaneously—fully populating the Swift’s expansion slots. Each card delivered its previously observed performance while running lightweight inference tests concurrently.

Throughout all four test configurations, the Swift’s CPU load was monitored. The CPU usage remained low and well within expected parameters, indicating the AI workloads were being efficiently offloaded to the Hailo accelerators.

The Hailo AI expansion boards, plugged into the Swift embedded computer, demonstrated smooth real-time inference across multiple neural networks with minimal power draw (typically 60W). It confirmed that VersaLogic products could meet AI demands in medium performance applications ranging from surveillance and robotics to industrial automation—without sacrificing the rugged reliability that VersaLogic is known for. The compact Swift system, including the Hailo AI expansion boards, measures only 90 x 96 x 63 mm.

This successful testing marks another milestone in VersaLogic’s roadmap for AI integration. By proving the performance of AI coprocessors like Hailo’s in real-world conditions, VersaLogic ensures customers can confidently deploy edge AI with the same trust they’ve always placed in its embedded systems.

VersaLogic is showing that AI capability isn’t just coming to the edge—it’s already here.

Interested in learning more? Contact the VersaLogic team today: