Migrate to Open Edge Platform#
Open Edge Platform enables you to migrate your existing systems to Intel hardware without much effort, which is especially seamless with moving from NVIDIA.
Why migrate#
AI inference has changed significantly since gaining recognition as a productivity solution. It is no longer defined by hugely complex cloud frameworks powered by enormous datacenters. It is now available to a much wider user and developer base, running on all kinds of edge devices, from tiny wearables, to industry-grade local servers.
Scalability
Build AI applications once and deploy them in different environments, from constrained edge devices to powerful servers.
A successful AI system is not the one with the highest possible compute, but one with the right hardware for the job. With Open Edge Platform, your system scales seamlessly, automatically adapting to a variety of devices, without core logic rewrites. With one code base, you can run inference on a mobile device powered by NPU, or a high-performance Xeon-based server. You can scale the feature set as easily, using the platform’s modular approach.
Power Efficiency and Cost Reduction
Reduce operational costs, extend device battery life, and reach deployment scenarios that are not covered efficiently by NVIDIA-optimized architectures.
While discrete GPUs excel at raw performance, they often fall behind under strict power and thermal constraints. Intel edge solutions support various hardware options. With the selection of CPU, iGPU, and NPU devices, you can achieve just the right fit for your performance-per-watt needs.
No hardware lock-in
Avoid dependence on a single vendor ecosystem and keep your software portable across a wide range of hardware.
Relying on a tightly-knit ecosystem of tools and APIs can make maintenance and scaling costly. With Open Edge Platform you build on flexible foundations, using broadly adopted frameworks and standard development workflows. This means more freedom to adjust the hardware to fit your performance, power, and cost targets, with minimal reconfiguration effort.
Open source flexibility
Own your source code with as little dependency on proprietary software as possible.
Build an Intel-optimized AI system and still own the code base. If you have very specific needs, you do not have to wait for the vendor to provide an update, just tweak your solution freely. Open source also means the freedom to play with the software stack before deciding to switch on production.
Extending compatibility with edge devices
Support more deployment targets without redesigning your AI pipeline for each device class.
Edge environments are highly diverse, ranging from low-power endpoints to industrial systems and local servers. Open Edge Platform helps bridge these differences by combining standardized frameworks with Intel-optimized runtimes, so the same application architecture can run across multiple hardware profiles. This improves rollout speed, simplifies maintenance, and makes it easier to expand solutions to new edge locations over time.
With the Intel-optimized software stack, you can select from a range of hardware options that best fit different AI systems:
CPU
The most universal processing unit, used anywhere from mobile devices (Core) to enterprise servers (Xeon). It is ideal for latency-sensitive workloads and offers the widest model support.Integrated GPU (iGPU)
Built into most Intel processors, reducing the need for additional hardware installation. It offers excellent performance-per-watt for both edge devices and client endpoints.Discrete GPU (dGPU)
Hardware expansion enabling the highest inference throughput but due to its physical installation requirements most typically used for server-side and work-station inference.Neural Processing Unit (NPU)
An accelerator offering increased performance for low power consumption use cases. Integrated with a selection of Intel CPUs, it enables efficient and cost-conscious AI workloads on edge devices.
Importantly, the transition process may differ between cases, depending on the initial software stack. Here are some areas you should consider before you start: