
Date published
28 Jan 2025
Author
Iceotope
Key highlights:
-
Stranded capacity occurs when there is an imbalance of power, cooling and space in data centers. Servers that are undercooled won’t be able to achieve full usage.
-
Data centers handling HPC and the likes of AI and ML face issues with increased heat of GPUs. This results in higher power consumption across data center workloads.
-
Precision Liquid Cooling reduces data center energy consumption by up to 40% and water usage by up to 96%. It reduces stranded capacity and enables data centers to make better use of the compute that already exists.
As more and more data centers are increasingly handling High-Performance Computing (HPC), they continue to be pushed to their limits. Data center managers are continually faced with the challenge of delivering the output demanded whilst balancing the economic viability of data centers with sustainability requirements.
This increased pressure when handling HPC has led to many data centers dealing with stranded capacity, whereby they can’t access capacity required to support critical load. Here we look at the problem of stranded capacity and how Precision Liquid Cooling can free data center capacity.
What is stranded capacity?
Power, cooling and space are all interlined components in data centers. Stranded capacity occurs when there is an imbalance in any of the three components. Servers that are undercooled for example won’t be able to achieve full usage. If there isn’t enough power, capacity will be down.
Take the example of power capacity. This becomes stranded capacity when cooling equipment isn’t being used to its full potential, resulting in consumption of non-useful power. If a data center only uses half of its potential power during operation, 50% of the power is stranded power.
Stranded power, cooling, and space capacity in data centers have a knock-on effect on both efficiency and scalability. And it’s a widespread problem across the world. In fact, most servers run at a fraction of the capacity due to the constraints of air cooling technology.
Increased power consumption
As more and more organizations are utilizing the capabilities of HPC, specifically around AI and machine learning, this continues to put more pressure on data center capacity. Graphics processing units (GPUs) are the cornerstone of AI processing and play a crucial role in data centers – they are used for intensive computing applications. GPUs are particularly effective at performing calculations on large data sets and are increasingly being relied on today.
Yes, this brings issues with data center energy consumption. The high-power, high-density GPUs and CPUs powering today’s artificial intelligence applications result in much higher power consumption across data center workloads.
The more power being used, the higher the costs and the higher the potential for stranded capacity if these components aren’t cooled properly.
The pressure on data center managers
Data center managers are facing pressure from all angles when it comes to data center capacity planning and management. With data centers required to deal with HPC and the likes of AI processing, energy usage is on the rise. Not only does this cause issues with increasing power costs but there is also the matter of impending government regulations.
Evaporative cooling techniques for heat rejection are sometimes popular. They bring both high efficiency and large cooling capacity. But they also consume substantial amounts of water, which can increase operational costs. It also puts data center managers under increased scrutiny to reduce water usage – especially the case in regions without enough water resources.
Precision Liquid Cooling and freeing stranded capacity
Precision Liquid Cooling is a significant advancement in data center cooling technology. It easily accommodates the increasing heat loads from the latest processor roadmaps and manages them effectively.
Crucially, it offers enhanced cooling efficiency to underpin data center performance. Precision Liquid Cooling provides targeted cooling to efficiently remove heat from critical components. This enhances both performance and reliability of the servers.
Alongside this, Precision Liquid Cooling significantly reduces water usage from the cooling system design by up to 96%. Crucially, it also reduces energy usage by 40%. Precision Liquid Cooling targets the hottest components first with a small amount of dielectric fluid, delivering exceptional thermal stability through the precise delivery of dielectric fluid across the entire IT stack. It also prevents performance-degrading hotspots from occurring too.
This type of cooling system uses a closed loop of dialectic coolant that circulates within each server’s sealed chassis and is then delivered to the heat-generating components. The coolant absorbs the heat and transfers it via heat exchange to a warm water-based circuit. It also uses dry coolers to remove heat, not fans or evaporation.
Precision Liquid Cooling reduces stranded capacity and enables data centers to make better use of the compute that already exists.
When HPC pushes data center capacity and raises the issue of stranded capacity, Iceotope's Precision Liquid Cooling optimizes these workloads by boosting total system performance by 7% and reducing data center energy consumption by up to 40%. Through Precision Liquid Cooling solutions, data center managers can simply maintenance, increase density and lower the total cost of ownership.