2PDLC
for Server/Switch
Product
Beyond the Limits of Traditional Liquid Cooling
Redefining Thermal Efficiency for High-Power AI Computing
As AI chip power consumption continues to surge, traditional single-phase direct liquid cooling (SPDLC) is rapidly reaching its limits. Based on sensible heat transfer (convection), SPDLC is only suitable for moderate thermal loads. Under extreme workloads, coolant temperatures can rise significantly—impacting performance and system stability.
Moreover, attempts to enhance SPDLC performance often result in high pressure drops and increased pumping power, which can even lead to structural deformation challenges and added complexity in manufacturing and validation.
Next-Generation Cooling with 2PDLC
Built for High-Density, High-Power AI Workloads
Arivor Tech’s 2PDLC (Two-Phase Direct Liquid Cooling) is engineered to overcome these limitations, delivering breakthrough performance for next-generation AI infrastructure.
SPDLC vs. 2PDLC
Category | SPDLC (Single-Phase DLC) | 2PDLC (Two-Phase DLC) |
|---|---|---|
Scalability | Limited for high-density systems | Built for AI-scale deployment |
Pumping Power | High | Significantly reduced |
Energy Efficiency (PUE) | ~1.5 – 1.2 | ~1.1 – 1.05 |
Temperature Control | Large temperature variation | Near-isothermal stability |
Cooling Mechanism | Sensible heat (convection) | Latent heat (phase change) |
Thermal Capacity | Limited for high-power AI workloads | Up to 4000W per die |
Flexible Deployment & Intelligent Management
Designed for Scalability and Sustainability
■ Versatile Infrastructure Support
-
Flow-boiling cold plate design ensures uniform heat distribution
-
No orientation limitations; supports vertical installation
-
Compatible with EIA / OCP racks
-
Equipped with 4U In-Rack CDU, delivering up to 200kW cooling capacity
■ Smart & Sustainable Operation
-
Integrated AI-driven power and thermal management
-
Built-in pressure and leakage detection
-
Supports coolant vapor-liquid recovery
-
Enables a more efficient and sustainable cooling ecosystem