Tech Support
Support for VPU-based video infrastructure—designed for real deployments, not experiments.
Contact support for deployment assistance.
support@netint.com >
Open Tickets
View and manage active support tickets.
Login now >
Software Releases
Firmware, drivers, and VPU software releases.
Download now >
Technical Documentation
Technical guides,
SDKs, and APIs.
View portal >
Meet our Team
Introducing our Field Application Engineers. Think of them as your technical co-pilots, ready to step you through your VPU deployments.
Deployment FAQs & Runbook
Common Issues, Causes, and What to Check First
Use this guide to quickly isolate issues in production, pre-production, and hybrid deployments.
1. Why isn’t the VPU being detected by the OS or application?
Most likely causes: Initialization, BIOS, or driver alignment.
Check:
- PCIe slot enabled and correctly sized
- BIOS settings (lane allocation, SR-IOV if required)
- Firmware and driver versions match
- OS-level enumeration (lspci, device manager)
Rule: If the OS doesn’t see the card, the application won’t either.
2. What causes performance to be lower than expected?
Most likely causes: System configuration, not hardware limits.
Check:
- Encoder and codec profile settings
- Host CPU capacity feeding the VPU
- NUMA alignment and PCIe placement
- Workload matches intended VPU use case
Rule: VPUs deliver predictable performance when the surrounding system is tuned correctly.
3.What causes instability at high channel density?
Most likely causes: Power, thermal, or system-level constraints.
Check:
- Server and rack power headroom
- Thermal limits under sustained load
- PCIe contention
- Memory bandwidth saturation
Rule: High density requires balanced systems—not just more cards.
4. How to safely handle firmware and software updates?
Most likely risks: Version incompatibility and uncontrolled rollout.
Check:
- Version compatibility matrix
- Validation in a staging environment
- Rollback plan in place
Rule: Treat updates like infrastructure changes, not patches.
5. Why does encoding fail or stall intermittently?
Most likely causes: Input irregularities or resource exhaustion.
Check:
- Input stream integrity
- Supported codec profiles and parameters
- Burst load behavior
- Application logging and retry logic
Rule: Intermittent failures usually originate upstream of the VPU.
6. How do we scale deployments without introducing risk?
Most likely failure mode: Scaling too fast without validation.
Check:
- One workload or region scaled at a time
- Density and power assumptions revalidated
- Monitoring expanded with scale
Rule: Scale incrementally and validate at each step.
7. In a multi-vendor stack, who owns the issue?
Ownership follows the architecture.
Check:
- Hardware, firmware, SDK → NETINT
- Application logic and orchestration → customer or partner
- Shared workflows → coordinated escalation
Rule: Escalate based on root cause, not symptoms.
8. Why does performance differ between on-prem and cloud nodes?
Most likely causes: Infrastructure asymmetry.
Check:
- CPU generation differences
- PCIe and memory bandwidth
- Network latency between environments
- Cloud instance constraints
Rule: Hybrid performance must be evaluated per environment.
9. Why does orchestration behaves inconsistently in hybrid deployments?
Most likely causes: Assumptions of uniform infrastructure.
Check:
- API rate limits
- Storage and network I/O differences
- Failure and retry behavior per environment
Rule: Hybrid orchestration requires environment-specific tuning.
10. How do we prevent version drift in hybrid deployments?
Most common hybrid failure mode: Inconsistent firmware, drivers, or SDKs.
Check:
- Version parity across environments
- Compatibility confirmed before upgrades
- Staged rollout sequence documented
Rule: Control upgrades centrally and apply incrementally.