WINK Virtual Appliance Hardware Requirements
2025 Specifications for vForge, vRouter (Media Router), and vTraffic
Last Updated: January 2025 | Version 2.0
📋 Support Scope & Guidelines
What we support: WINK provides comprehensive support for our virtual appliance software, including configuration, optimization, and troubleshooting.
What we don't support: Virtualization platform issues, hypervisor configuration, or infrastructure problems.
For platform-specific help:
- VMware vSphere → VMware Support
- Microsoft Hyper-V → Microsoft Support
- Proxmox VE → Community Forums
- Cloud Platforms → AWS/Azure/GCP Support
💡 Pro tip: Most issues can be resolved by checking our documentation first. We're here to help with WINK software questions!
Executive Summary
WINK virtual appliances are designed for modern datacenter deployment with specific CPU, memory, and storage requirements. All specifications assume:
- Physical CPU cores (not hyperthreaded/virtual cores)
- 2021 or newer CPU architecture (Intel Xeon Silver/Gold/Platinum, AMD EPYC)
- Dedicated resources (not oversubscribed)
- Low-latency storage (SSD or better)
GPU Note: GPU acceleration does NOT help with live video transcoding in WINK vForge. While GPUs have thousands of CUDA cores, they cannot efficiently handle multiple independent transcoding threads. Each video stream requires sequential processing that doesn't parallelize well on GPU architectures. For vForge, invest in more CPU cores instead of GPU hardware.
Exception: The WINK AI Appliance (not covered in this document) can leverage GPU acceleration for AI inference tasks, which is fundamentally different from video transcoding.
Other WINK Software: Most WINK streaming software applications are multi-platform and simply require a modern system with at least 4 GB of free RAM. This document specifically covers the high-performance virtual appliances (vForge, vRouter, vTraffic) that have specialized hardware requirements.
Note: WINK Archive, WINK Wall, and WINK Encoder are special use cases with their own hardware requirements. WINK Wall and WINK Encoder are reaching End of Life (EOL) - please see our EOL announcement for migration options.
Quick Reference Table
| Product |
Use Case |
CPU Cores |
RAM |
Storage |
Network |
| WINK vForge |
Per HD stream (≤720p) |
1 physical core |
2 GB |
20 GB |
1 Gbps min
10 Gbps rec |
| WINK vForge |
Per 2K stream (≤1440p) |
2 physical cores |
4 GB |
| WINK vRouter |
100 streams / 1K viewers |
3 physical cores |
16 GB |
20 GB base |
1 Gbps min
10 Gbps rec |
| WINK vRouter |
100 streams / 10K viewers |
5 physical cores |
16 GB |
| WINK vRouter |
300 streams / 20K viewers |
7 physical cores |
16 GB |
| WINK vRouter |
500 streams / 100K viewers |
16 physical cores |
48 GB |
| WINK vTraffic |
Per HD stream (≤720p) |
2 physical cores |
8 GB |
100 GB base
+2 GB/stream |
1 Gbps min
10 Gbps rec |
| WINK vTraffic |
Per 2K stream (≤1440p) |
4 physical cores |
16 GB |
Detailed Specifications
WINK vForge (Virtual Transcoder)
Live Transcoding Requirements
Formula: Number of concurrent streams × cores per stream = Total physical cores needed
- HD Streams (720p and below): 1 physical core + 2 GB RAM per stream
- 2K Streams (1080p-1440p): 2 physical cores + 4 GB RAM per stream
- 4K Streams (2160p): 4 physical cores + 8 GB RAM per stream
Example: 50 concurrent HD streams = 50 physical cores + 100 GB RAM minimum
Network Requirements: vForge requires 2 internal IP addresses for proper operation (management and media interfaces).
Hyperthreading Recommendation: For optimal vForge performance, we recommend disabling hyperthreading in your BIOS/hypervisor settings. This ensures predictable performance and prevents thread contention during intensive transcoding operations.
Scaling Guidelines
| Deployment Size |
HD Streams (720p) |
2K Streams (1080p) |
CPU Cores |
RAM |
| Small |
10-25 |
5-12 |
25 |
50 GB |
| Medium |
50 |
25 |
50 |
100 GB |
| Large |
100 |
50 |
100 |
200 GB |
Scaling Note: For deployments beyond 120 streams, it is essential to deploy multiple vForge instances behind a load balancer. Each instance should not exceed 120 concurrent streams for optimal performance and reliability.
Supported CPU Architectures (2021+)
- Intel Xeon: Silver 4300 Gold 5300/6300 Platinum 8300
- AMD EPYC: 7003 Series 9004 Series
- Minimum clock speed: 2.4 GHz base frequency
WINK vRouter (Media Router)
Scaling Guidelines
| Deployment Size |
Streams |
Concurrent Viewers |
CPU Cores |
RAM |
| Small |
Up to 100 |
1,000 |
3 |
16 GB |
| Medium |
Up to 100 |
10,000 |
5 |
16 GB |
| Large |
Up to 300 |
20,000 |
7 |
16 GB |
| Enterprise |
Up to 500 |
100,000 |
16 |
48 GB |
Scaling Note: For deployments beyond 500 streams or 100,000 viewers, deploy multiple vRouter instances behind a load balancer.
Network Requirements: vRouter requires 2 internal IP addresses for proper operation (management and media interfaces).
WINK vTraffic (AI Analytics)
Enhanced Requirements
vTraffic requires approximately 2× the resources of vForge plus 4× the RAM due to AI processing overhead:
- HD Analytics (720p): 2 physical cores + 8 GB RAM per stream
- 2K Analytics (1080p-1440p): 4 physical cores + 16 GB RAM per stream
- Storage: 100 GB base + 2 GB per stream for model caching
Important: vTraffic performs real-time AI inference. Underprovisioning will result in dropped frames and missed detections.
Supported Virtualization Platforms
Network Requirements
Bandwidth Calculations
- Inbound: (Number of streams × Average bitrate) + 20% overhead
- Outbound: (Number of viewers × Average bitrate) + 20% overhead
- Minimum NIC: 1 Gbps (we support whatever your virtual hardware supports)
- Recommended NIC: 10 Gbps for production deployments
- IP Requirements: 2 internal IP addresses required (vForge & vRouter)
- Latency: < 5ms to source cameras for live transcoding
| Stream Quality |
Typical Bitrate |
50 Streams Inbound |
1000 Viewers Outbound |
| 480p |
300 Kbps |
18 Mbps |
360 Mbps |
| 720p |
750 Kbps |
45 Mbps |
900 Mbps |
| 1080p |
1.5 Mbps |
90 Mbps |
1.8 Gbps |
NIC Driver Note: When using non-standard network cards or 10 Gbps adapters, NIC driver compatibility may need to be discussed with WINK support. Standard virtual network adapters (vmxnet3, virtio, etc.) are fully supported.
Storage Requirements
Performance Specifications
- Type: SSD required (NVMe preferred for vTraffic)
- IOPS: Minimum 3000 IOPS per instance
- Throughput: 200 MB/s sequential read/write minimum
- Latency: < 1ms average
Capacity Planning
- vForge: 11-20 GB OS/App (depending on release and bundled features) - no data storage
- vRouter: 11-20 GB OS/App (depending on release and bundled features) - no data storage
- vTraffic: 100 GB OS/App + 2 GB per stream for AI models
Note: vForge and vRouter do not store video data. They are pure processing appliances that transcode and route streams in real-time.
Cloud Deployment Guidelines
AWS EC2 Instance Types
Recommended Instances
- vForge: C6i.xlarge (per 4 HD streams) or C6i.8xlarge (per 32 HD streams)
- vRouter Small: C6i.large (3 cores equivalent)
- vRouter Enterprise: C6i.4xlarge (16 cores equivalent)
- vTraffic: C6i.2xlarge (per 4 HD streams)
Azure VM Sizes
Recommended Instances
- vForge: F4s_v2 (per 4 HD streams) or F32s_v2 (per 32 HD streams)
- vRouter Small: F2s_v2
- vRouter Enterprise: F16s_v2
- vTraffic: F8s_v2 (per 4 HD streams)
Performance Optimization Tips
- CPU Affinity: Pin vCPUs to physical cores for consistent performance
- NUMA Awareness: Keep VM within single NUMA node when possible
- Memory Reservation: Always reserve full memory allocation
- Network Optimization: Use SR-IOV or DPDK where available
- Storage Tiering: Use fastest storage tier for vTraffic AI models
Common Mistakes to Avoid
- ❌ Using hyperthreaded cores as "physical" cores
- ❌ Leaving hyperthreading enabled for vForge (disable it!)
- ❌ Oversubscribing CPU resources
- ❌ Using mechanical hard drives
- ❌ Deploying on CPUs older than 2021
- ❌ Expecting GPU to help with transcoding
- ❌ Ignoring network bandwidth requirements
Deployment Examples
Example 1: Small City Traffic System
Requirement: 50 HD traffic cameras with 500 concurrent public viewers
- vForge: 50 cores, 100 GB RAM, 20 GB storage
- vRouter: 3 cores, 16 GB RAM, 20 GB storage
- Total: 53 physical cores, 116 GB RAM, 10 Gbps network
Example 2: Large Enterprise Security
Requirement: 200 HD cameras with 20,000 employee access
- vForge: 200 cores, 400 GB RAM, 20 GB storage
- vRouter: 7 cores, 16 GB RAM, 20 GB storage
- Total: 207 physical cores, 416 GB RAM, 10 Gbps network
Licensing and Sales
Important: Virtual appliance licensing is based on your deployment size and use case.
Contact our sales team to discuss:
- Licensing options for your specific requirements
- Volume discounts for large deployments
- Support packages and SLAs
- Professional services for deployment assistance
Sales Contact: sales@wink.co | +1-312-281-5433
Support and Contact
Remember: WINK supports our software, not your infrastructure.
For WINK software issues: support@wink.co
For virtualization platform issues: Contact your platform vendor
© 2025 WINK Streaming | Hardware Requirements v2.0 | wink.co