AI Compute
EQ Syntropy is optional and requires GPU-accelerated hardware. Every gateway runs EQ Coherence and EQ Sight without it. When you want Syntropy’s physics-informed analytics, event detection, and AI-assisted diagnostics, choose where to run it:
| Option | Hardware | Data Location | Best For |
|---|---|---|---|
| Edge AI (Orin / Thor / DGX Spark) | NVIDIA GPU on-site | Local (NVMe SSD) | Air-gapped sites, data sovereignty, lowest latency |
| EQ-Hosted | Dedicated EQ server | Replicated to EQ server | No GPU needed on-site; cross-site analytics |
| Customer-Hosted | Customer data center GPU (RTX 4090 class or better) | Customer servers | Enterprise IT with existing GPU infrastructure |
All options deliver the same Syntropy capabilities through EQ Sight. The choice depends on your network environment, data policies, and existing infrastructure.
With edge AI or customer-hosted deployments, AI inference runs entirely on local hardware. There are no per-query fees, no usage limits, and no dependency on external services.
EQ-hosted deployments provide the same Syntropy capabilities with EQ managing the AI infrastructure on your behalf.
Edge AI platforms
Three on-site GPU platforms, from hardened industrial edge to a high-capacity developer node:
| Platform | Tier | AI Performance | Memory | Sensor Network | Operating Temp | Form Factor |
|---|---|---|---|---|---|---|
| AGX Orin (NRU-220S) | Industrial edge | 275 INT8 TOPS | 64 GB LPDDR5 | 5 direct GbE PoE+ ports | −25 to 50 °C | Fanless industrial |
| Thor (MIC-742-AT) | Industrial edge (high density) | 2,070 FP4 TFLOPs | 128 GB LPDDR5X | QSFP28 breakout to switches | −10 to 60 °C | Industrial edge server |
| DGX Spark | Developer / lab / eval | ~1 PFLOP (FP4) | 128 GB unified | 200GbE via switch | ~0–35 °C (office) | Desktop (not industrial) |
| Standard Gateway | No AI (baseline) | None (CPU only) | 4 GB | 1 dedicated port | −40 to +80 °C | DIN-rail gateway |
Orin and Thor are hardened, always-on industrial edge units. DGX Spark is a developer/evaluation-class desktop (the AI-compute analogue of the Lab Gateway) — use it for prototyping, benchmarking, and high-capacity on-prem evaluation, not hardened field deployment.
Software stack
Edge AI deployments run the same EQ stack regardless of platform:
| Layer | Component | Purpose |
|---|---|---|
| OS | JetPack (Jetson Orin/Thor) or DGX OS (DGX Spark), Ubuntu-based | NVIDIA-optimized Linux |
| Runtime | CUDA, TensorRT | GPU acceleration framework |
| Data | EQ Coherence | Collection, storage, REST API |
| AI | EQ Syntropy | Physics-informed analytics |
| Interface | EQ Sight | Web-based visualization |