On-Demand Webinar: Is Your Infrastructure Ready for the AI Era?
Michael Ko, Managing Director of Vitex, breaks down the essentials of 800G deployment — drawing on over two decades of experience to deliver a practical guide to the hardware, cabling, and architecture driving the next generation of connectivity.
📋 Session Overview
5 key topics covered — navigate directly to what you need
▶️ Watch the Webinar
800G Deployment Strategies with Michael Ko
🤖 1. The AI & AGI Infrastructure Driver
As Artificial Intelligence and Artificial General Intelligence workloads surge, the demand for bandwidth is skyrocketing. The transition to 800G is no longer just a roadmap item — it is an immediate necessity for modern data centers striving to keep pace with high-performance compute requirements.
AI and AGI workloads are pushing current infrastructure to its absolute limits
NVIDIA-based GPU clusters are reshaping market demands and network architecture requirements
No longer a future consideration — 800G is an immediate operational necessity for competitive AI infrastructure
🏗️ 2. Architecture & Form Factors: OSFP vs. QSFP-DD
A detailed comparison of the two dominant 800G form factors — and why thermal management is the deciding factor when choosing between them for your deployment.
OSFP
- Larger, high-thermal-capacity form factor
- Perfect for 800G deployments where cooling is challenging
- Optimized for dense, high-power AI rack environments
- Best path toward 1.6T next-generation upgrades
QSFP-DD
- Widely adopted for flexibility and backward compatibility
- QSFP-DD ports run both 400G and 800G modules
- Enables incremental, phased upgrades on existing infrastructure
- Ideal for brownfield environments migrating from 400G
Next-Gen Connectors
The session also explores emerging connector standards including MDC, SN, and Dual MPO-12 — the next generation of high-density physical interfaces enabling greater port counts in the same panel space as legacy LC and MPO-16 formats.
🔬 3. Transceiver Deep Dive
An inside look at what is actually inside an 800G module — from DSPs and drivers to lasers — and how to select the right variant for every reach requirement in your cluster.
| Reach Requirement | Typical Module Type | Use Case |
|---|---|---|
| Up to 50m | SR8 (OSFP or QSFP-DD) | Intra-rack and top-of-rack GPU-to-switch connections |
| Up to 100m | DR8 / SR8 extended | Adjacent rack and short inter-row connections over OM4/OM5 |
| Up to 2km | 2FR4 (OSFP or QSFP-DD) | Inter-hall and inter-building AI cluster links on singlemode |
| Long range (2km+) | 2LR4 / LR variants | Campus DCI and long-reach AI pod interconnects |
🔌 4. The Cabling Dilemma: DAC, AEC & AOC — Solved
Practical criteria for choosing the right short-reach cabling solution for every scenario inside and between racks — one of the most frequently misunderstood decisions in AI cluster deployment.
DAC — Direct Attach Copper
- Passive copper — zero power consumption
- Cost-effective for very short runs up to 3–5m
- Best for dense top-of-rack wiring on a tight power budget
- No signal regeneration — distance-limited by copper attenuation
AEC — Active Electrical Cable
- Active copper with signal conditioning at each end
- Extends copper reach beyond passive DAC limitations
- Lower cost than AOC for medium short-reach scenarios
- Good balance of cost and reach for 5–15m connections
AOC — Active Optical Cable
- Fiber-based, hot-swappable, plug-and-play
- Reach up to 100m — covers most inter-rack scenarios
- EMI-immune — ideal for high-interference GPU rack environments
- Best for GPU-to-switch and switch-to-switch across rows
Quick Selection Guide
- Under 5m, cost-sensitive: DAC
- 5–15m, cost-balanced: AEC
- 15–100m, or EMI concern: AOC
- Beyond 100m: discrete transceiver + structured fiber
⚙️ 5. DSP Technology: Standard, LRO & LPO Explained
Demystifying the architectural differences between the three DSP approaches used in modern 800G transceivers — and their real-world impact on power consumption and deployment complexity.
Full digital signal processing at both ends — maximum reach and signal integrity, higher power draw
DSP on transmit only; linear analog receive — reduces power consumption while maintaining good reach performance
No DSP in the module — processing offloaded to the host ASIC — lowest module power, lowest latency for short-reach AI cluster links
⏱️ Key Segments & Timestamp Guide
Navigate directly to the technical details you need most in the recorded session.
| Timestamp | Topic Covered |
|---|---|
| 01:12 | Market Drivers: The impact of AI & NVIDIA on bandwidth requirements |
| 08:24 | Form Factor Battle: The critical differences between OSFP and QSFP-DD |
| 10:14 | Next-Gen Connectors: Exploring MDC, SN, and Dual MPO-12 |
| 16:15 | Evolution of DSPs: Standard vs. LRO vs. LPO performance |
| 20:55 | Cabling Guide: Practical criteria for choosing DAC, AEC, or AOC |
🎤 About the Speaker
🚀 Ready to Upgrade Your Network?
Don't let outdated infrastructure bottleneck your AI potential. Vitex provides end-to-end solutions backed by rigorous US-based testing facilities to ensure your transition to 800G is seamless and reliable.
Call Us
- Speak directly with a Vitex engineer about your specific architecture needs
- +1 201-296-0145
- US-based technical support — available for pre-sales specification review
Email Us
- Request a quote, compliance documentation, or custom assembly specifications
- sales@vitextech.com
- Rapid response for volume orders, TAA/NDAA compliance, and custom configurations
