Specifications
Product
Iceotope KUL Extreme
Cooling method
Precision liquid cooling - single-phase dielectric, direct-to-everything
Proven chip-level capacity
Up to and beyond 1,500 W per device (laboratory thermal test vehicle)
KUL SINK thermal resistance
0.039 K/W at 7 L/min and a 1,000 W heat load
Benchmark advantage
11.4% improvement over best-in-class tank immersion (like-for-like)
Energy saving
Up to 40% lower power usage per kW of ITE power vs air cooling
Water saving
Up to 96-100% lower water consumption vs air cooling
Cooling cost
Up to 83% lower cooling cost
Rack density
Up to 6x more compute per square metre
Acoustics
Fan-less operation, below 40 dB
Heat rejection
Dry coolers - no evaporative water use
Heat reuse
Captured heat can be rejected or recovered for reuse
Target workloads
AI training and inference, HPC, edge AI and vision
Overview
Iceotope KUL Extreme is a precision liquid cooling system that immerses the electronics in a sealed single-phase dielectric loop and transfers heat out through a wall-mounted heat exchanger to a warm-water circuit. Every heat-generating component - processors, memory, storage and power - is cooled without fans, and the loop is proven in laboratory testing to remove more than 1,500 W from a single device, well past the accepted single-phase ceiling.
The measured thermal resistance of the copper-pinned KUL SINK is 0.039 K/W at a 7 L/min flow rate with a 1,000 W load, which Iceotope reports as an 11.4% improvement over best-in-class tank immersion in like-for-like tests. At system level the company quotes up to 40% lower energy use, up to 96-100% lower water consumption, up to 83% lower cooling cost and up to six times the compute density per square metre versus air cooling, with fan-less operation under 40 dB and dry-cooler heat rejection.
Key Benefits
Proven beyond 1,500 W per chip - ready for the coming generation of AI accelerators. 0.039 K/W KUL SINK delivers best-in-class thermal resistance. Fan-less, sub-40 dB operation suits offices and edge sites. Up to 100% water saving with dry-cooler heat rejection.
Applications
High-density AI and HPC clusters, edge AI and vision systems, telco and RAN sites, and any deployment where water use, noise or cooling cost is constrained.
Request a Quote — ICEOTOPE KUL EXTREME — PRECISION LIQUID COOLING FOR 1,500 W AI ACCELERATORS
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →