Core Concepts: The BDP and AIR Indices

Core Concepts: The BDP and AIR Indices

Standard web performance metrics (like Core Web Vitals) measure how fast a page loads over the network. They do not measure if the device has the physical computing power to execute the code once it arrives.

ScaleDynamics introduces two proprietary indices to quantify client-side execution power in real-time.

What is BDP (Browser Device Performance)?

The BDP index measures the real-time execution capacity of the user's CPU and memory within the browser environment. It evaluates how quickly the device can parse, compile, and execute heavy JavaScript (like React/Vue hydration) and complex DOM manipulations.

We classifies devices into 4 deterministic tiers:

  • BDP 1: Entry-Level (The Danger Zone). Legacy devices or low-power modes (e.g., iPhone 8 on battery saver, cheap Androids). These devices will choke, overheat, or crash on heavy JavaScript or WebGL. Action: Serve static HTML, disable animations, remove 3D elements.

  • BDP 2: Basic. Older mid-tier devices. They can handle standard web applications but suffer from micro-stutters during heavy client-side routing. Action: Defer non-critical third-party scripts.

  • BDP 3: Mid-Range. Modern average consumer hardware. Capable of handling standard SPAs and complex logic without noticeable delay.

  • BDP 4: High-End. Flagship devices (e.g., iPhone 15 Pro, M3 MacBooks). Massive CPU firepower. Action: Safe to unlock rich WebGL, 3D interactive elements, and complex real-time collaborative features.

What is AIR (AI Readiness)?

The AIR index is specifically designed for Edge AI computing. It benchmarks the device's capability to run local AI workloads (using WebAssembly, WebGL, WebGPU, or WebNN) without degrading the core user experience or draining the battery.

We classifies Edge hardware into 4 tiers:

  • AIR 1: Baseline Edge. Limited to sequential CPU computing. Cannot run LLMs locally without freezing the browser. Action: Route all AI prompts to Cloud APIs (OpenAI, Anthropic).

  • AIR 2: Accelerated Edge. Basic hardware acceleration kicks in, enabling light parallel computing. Suitable for very small, quantized tasks (like basic sentiment analysis).

  • AIR 3: High-Performance Edge. Massive firepower, capable of heavy matrix computation. Action: Safe to execute medium-sized local models in the browser to save cloud costs.

  • AIR 4: Neural Edge. The absolute pinnacle. Hardware featuring dedicated NPUs (Neural Processing Units) or high-end WebGPU support. Action: Execute full local LLMs and real-time computer vision entirely on the client.

How is the score calculated?

We do not rely on inaccurate User-Agent string parsing, nor do we run heavy, real-time micro-benchmarks that slow down your users. Instead, our scripts are using a zero-overhead lookup process:

  • Signal Capture: Upon initialization, our ultra-lightweight script captures specific, privacy-first hardware and browser capabilities.

  • The SpeedPower Engine: These signals are instantly cross-referenced against our proprietary SpeedPower.run database—a massive dataset built from thousands of real-world device performance benchmarks.

  • Tier Deduction: We identify the exact raw processing, rendering, and AI SpeedPower scores for that specific hardware profile. These raw scores are immediately translated into deterministic BDP and AIR tiers (1 to 4).

What is SpeedPower benchmark?

Have a look to https://speedpower.run (opens in a new tab).

Get Startedsd-agent.js Installation Guide