Is Liquid Cooling Ready to Go Mainstream?

By Steve Campbell

February 13, 2017

 

Editor’s Note: It’s no secret that heat is a killer of electronics – performance, density, reliability, and energy efficiency all suffer. In this commentary Steve Campbell, co-founder and managing partner at OrionX, a strategy and research firm, contends liquid cooling may be closer than we think to widespread use in HPC and traditional datacenters and no longer just the purview of supercomputers. Many of the benefits are familiar. Campbell argues that improved cooling technology, expanding offerings from vendors, and potentially major energy savings are the drivers of the liquid cooling adoption trend. SC16, he says, may have represented an early tipping point. See if you agree.

Lost in the frenzy of SC16 was a substantial rise in the number of vendors showing server oriented liquid cooling technologies. Three decades ago liquid cooling was pretty much the exclusive realm of the Cray-2 and IBM mainframe class products. That’s changing. We are now seeing an emergence of x86 class server products with exotic plumbing technology ranging from Direct-to-Chip to servers and storage completely immersed in a dielectric fluid.

Most people know that liquid cooling is far more efficient than air-cooling in terms of heat transfer. It is also more economical, reducing the cost of power by as much as 40 percent depending on the installation. Being more efficient with electricity can also reduce carbon footprint and contribute positively to the goals of “greenness” in the data centers, but there are other compelling benefits as well, more on that later.

Most HPC users are familiar with the Top500 but may not be as familiar with the Green500, which ranks the Top500 supercomputers in the world by energy efficiency. The focus of performance-at-any-cost computer operations has led to the emergence of supercomputers that consume vast amounts of electrical power and produce so much heat that large cooling facilities must be constructed to ensure proper performance.

To address this trend, the Green500 list puts a premium on energy-efficient performance for sustainable supercomputing. The most recent Green500, released during SC16, has several systems in the top 10 using liquid cooling.

CoolIT CHx650

US data centers consumed about 70 billion kilowatt-hours of electricity in 2014, about two percent of the country’s total energy consumption, according to a 2014 study conducted by the US Department of Energy in collaboration with researchers from Stanford University, Northwestern University, and Carnegie Mellon University. Liquid cooling can reduce electrical usage by as much as 40 percent, which would take a huge bite out of datacenter energy consumption.

Liquid cooling can also increase server density. The heat generated from the HPC servers rises up, and in a full rack of servers those at the top will experience temperature increases and ultimately shut down. Consequently, you can’t completely populate the rack with servers all the way to the top; instead you need additional racks and extra floor space to get to processing power wanted. Liquid cooling eliminates the need for additional racks, creating higher data center server density using less floor space. This need for processing power and capacity has only been increasing in a race to the top.

Another liquid cooling benefit is higher-speed processing – CPUs and other components can run at higher speed as they are cooled more efficiently. Also, servers require no fans making them operationally silent. The more servers in a datacenter the more fans are required and noise levels increase – until it hits a painful point, sometimes literally. Liquid cooling eliminates fans and thus reduces acoustic noise levels.

Reliability can also be improved as mechanical and thermal fatigue have been reduced in liquid cooling systems as there are no moving parts, no vibrations from fans for example, and the systems are being cooled more efficiently. The elimination of hot spots and thermal stresses will also lead to improved overall reliability, performance and life.

Liquid Cooling Round-up

Following is round up of vendors demonstrating liquid cooled servers at SC16:

Aquila developed Aquarius water cooled server system offered in an Open Compute Platform (OCP) rack, in partnership with Clustered Systems utilizing their cold plate cooling technology. Aquila has also partnered with Houston based TAS Energy to co-develop an edge data center around the Aquarius platform.

Asetek, a provider of hot water, direct-to-chip liquid cooling technology showcased solutions in use worldwide at HPC users with their OEM partners such as Cray, Fujitsu, Format, and Penguin. Liquid cooling solutions for HPE, NVIDIA, Intel, and others were also on display.

Asetek’s direct-to-chip cooling technology is deployed nine installations in the November 2016 Green500 list. The highest ranked is #5 on the list; The University of Regensburg QPACE3 is a joint research project with The University of Wuppertal and Jülich Supercomputing Center. Featuring Asetek liquid cooled Fujitsu PRIMERGY servers, it is one of the first Intel Xeon Phi KNL based HPC clusters in Europe. Ranked #6 on the Green500, Oakforest-PACS is the highest performance supercomputer system in Japan and ranked #6 on the Top500. Fujitsu also deployed HPC clusters with PRIMERGY server nodes at the Joint Center for Advanced High-Performance Computing (JCAHPC) in conjunction with the University of Tokyo and Tsukuba University.

Asetek also announced liquid cooling technology is cooling eight installations in the November 2016 edition of the TOP500 list of the fastest supercomputers in the world.

CoolIT Systems is a leader in direct Contact Liquid Cooling energy efficient liquid cooling solutions for the HPC, Cloud and Enterprise markets. CoolIT’s solutions target racks of high-density servers. The technology can be deployed with any server in any rack, according to CoolIT.

CoolIT has several OEMs including:

  • Hewlett Packard Enterprise Apollo 2000 System
  • NEC Blue Marlin
  • Dell PowerEdge C6320
  • Lenovo NeXtScale product offering

CoolIT Systems have also partnered with STULZ and showcased their Chip-to-Atmosphere concept within a micro datacenter. CoolIT was recently selected by the University of Toronto to provide custom liquid cooling for its new signal processing backend, which will support Canada’s largest radio telescope, the Canadian Hydrogen Intensity Mapping Experiment (CHIME), a joint project between the National Research Council of Canada (NRC) and three major Universities (McGill, Toronto, UBC).

Ebullient has developed a two-phase cooling system for data center servers. Low-pressure fluid, 3M Novec 7000, is pumped through flexible tubing to sealed modules mounted on the processors in each server. The fluid captures heat from the processors and transports it back to a central unit, where it is either rejected outside the facility or reused elsewhere in the facility or in neighboring facilities. Ebullient’s direct-to-chip systems can cool any server, regardless of make or model.

Ebullient is an early stage company founded in 2013 based on technology developed University of Wisconsin. The company raised $2.3 million in January 2016

Green Revolution Cooling’s CarnotJet System is a liquid immersion cooling solution for data center servers. Rack-mounted servers from any OEM vendor can be installed in special racks filled with a dielectric mineral oil. On show at their SC16 booth was the Minimus server, their own design to further cost reduce the server component of the overall system.

In December Green Revolution announced a strategic partnership with Heat Transfer Solutions (HTS), an independent HVAC manufacturers’ representative in North America. As part of the partnership, HTS is making a financial investment in GRC, which will provide growth capital as the company continues to expand its presence in the data center market. In addition, a new CEO was appointed to help grow the company.

LiquidCool Solutions is a technology development firm specializing in cooling electronics by total immersion in their own proprietary dielectric fluid. LiquidCool Solutions was originally founded in 2006 as Hardcore Computing with a focus on workstations, rebranding in 2012 to LiquidCool Solutions and its focus on servers. The company has demonstrated two new liquid submerged servers based on the Clamshell design. The Submerged Cloud Server, a 2U 4-node server designed for Cloud-computing applications and Submerged GPU Server, is a 2U dual node server designed for HPC applications that can be equipped with four GPU cards or four Xeon Phi boards.

LiquidMips showcased a server-cooling concept, a single processor chip immersed in 3M Fluorinert. It’s a long way from being a commercially viable product but represents another company entering the immersive cooling market.

Inspur Systems Inc., part of Inspur Group, showed two types of cooling solutions at SC16, a phase changing cooling solution with ultra-high thermal capacity, and a direct contact liquid cooling solution which allows users to maximize performance and lower operating expenses.

Allied Control specializes in 2-phase immersion cooling solutions for HPC applications. Having built the world’s largest 40MW immersion cooled data center with 252kW per single rack resulting in 34.7kW/sqm or 3.2kW/sqft incl. white space, Allied Control offers performance centric solutions for ultra-high density HPC applications. Allied Control utilizes the 3M Novec dielectric fluid.

The BitFury Group (Bitcoin mining giant) acquired Allied Control in 2015. In January 2017 BitFury Group announced a deal Credit China Fintech Holdings to set up a joint venture that will focus on promoting the technology in China. As part of the deal, Credit China Fintech will invest $30 million in BitFury and the setting up of the joint venture that will sell BitFury’s bitcoin mining equipment.

ExaScaler Inc. is specialized in submersion liquid cooling technology. ExaScaler, and its sister company PEZY Computing, unveiled ZettaScaler-1.8, the first Super Computer with a performance density of 1.5 PetaFLOPS/m. The ZettaScaler-1.8 is an advanced prototype of the ZettaScaler-2.0 due to be released in 2017 with a performance density three times higher than the ZettaScaler-1.8. ExaScaler immersion liquid cooling using 3M Fluorinert cools ZettaScaler-1.8 Super Computer.

Fujitsu demonstrated a new form of data center, which included cloud-based servers, storage, network switch and center facilities, by combining the liquid immersion cooling technology for supercomputers developed by ExaScaler Inc. with Fujitsu’s know-how on general-purpose computers. Fujitsu is able to capitalize on three decades of liquid cooling expertise with mainframes, to supercomputers to Intel x86.

This new style of data center uses liquid immersion cooling technology that completely immerses IT systems such as servers, storage, and networking equipment in liquid coolant in order to cool the devices.

The liquid immersion cooling technology uses 3M’s Fluorinert, an inert fluid that provides high heat-transfer efficiency and insulation as a coolant. IT devices, including servers and storage, are totally submerged in a dedicated reservoir tank filled with liquid Fluorinert, and the heat generated from the devices is processed by circulating the cooled liquid through the devices. This improves the efficiency of the entire cooling system, thereby significantly reducing power consumption. A further benefit of the immersed cooling is that it provides protection from harsh environmental elements, such as corrosion, contamination, and pollution.

3M’s HPC offers solutions using 3M Engineered Fluids such as Novec or Fluorinert. Perhaps the winner at SC16 for immersed cooling is 3M as most of the vendors mentioned here use 3M Engineering Fluids. 3M fluids also featured in some of the networking products at the event. Fully immersed systems can improve energy efficiency, allow for significantly greater computing density, and help minimize thermal limitations during design.

Huawei announced a next-generation FusionServer X6000 HPC server that uses a liquid cooling solution featuring a skive fin micro-channel heat sink for CPU heat dissipation and processing technology where water flows through memory modules. This modular board design and 50ºC warm water cooling offers high energy-efficiency and reduces total cost of ownership (TCO).

Other vendors

HPE Apollo

HPE and Dell both introduced liquid cooling server products in 2016. Though they do not have the lineage of Fujitsu they nevertheless recognize the values liquid cooling delivers to the datacenter.

HPE’s entrance is the Apollo family of high-density servers. These rack-based solutions include compute, storage, networking, power and cooling. Target users are high-performance computing workloads and big data analytics. The top of the server lineup is the Apollo 8000 uses a warm water-cooling system whereas other members of the Apollo family of servers integrate the CoolIT Systems Closed-Loop DCLC (Direct Contact Liquid Cooling).

Dell, like HPE, does not have the decades of liquid cooling expertise of Fujitsu. Dell took the covers of the Dell Triton water cooling system in mid 2016. Dell’s Extreme Scale Infrastructure team built Triton as a proof of concept for eBay, leveraging Dell’s rack-scale infrastructure. The liquid-cooled cold plates directly contact the CPUs and incorporates liquid to air heat exchanges to cool the airborne heat generated by the large number of densely packed processor nodes.

Can we add liquid cooling to existing servers?

Good question, and the answer is no you really cannot. Adopting liquid cooling only makes sense on new server deployments. That is not to say it is impossible, but there are lots of modifications needed to make water cooling, like direct-to-chip, or fully immersed work, big maybe and not really recommended. An existing server has cooling fans that need to be disabled and CPU cooling towers removed and so on. You also need to add plumbing to your existing rack, which can be a pain.

There is no question that a prospective user needs to consider the impact and requirements on existing datacenter infrastructure, the physical bricks, mortar, plumbing, etc. For users considering water-cooled solutions you will need to plumb water to the server rack. If you are in a new datacenter that is one level of effort but if your datacenter is a large closet in an older building, like 43 percent of North American datacenter/server rooms, it may be a lot more difficult and expensive.

If you are considering a fully immersed solution, such as Fujitsu, no plumbing is required; all you need to do is hook up to chiller. It may be easier and less expensive than water-cooling. As a completely sealed unit it is conceivable that liquid immersed cooling solutions can be deployed almost anywhere, no datacenter required.

Most vendors covered in this market are small emerging technology companies. Asetek’s data center revenue was $1.8 million in the third quarter and $3.6 million in the first nine months of 2016, compared with $0.5 million and $1.0 million in the third quarter and first nine months of 2015, respectively. Asetek is forecasting significant data center revenue growth in 2016 from $1.9M in 2015.

CoolIT reported 2014 revenue of $27 million for all product categories. It is worth noting that Asetek and CoolIT data center revenues are less than 10% of total company revenue. The remaining 90% is workstation and PC liquid cooling solutions. Ebullient, Liquid MIPS, LiquidCooling, Green Revolution and Aquila have very few customers and probably below $10M annual revenues.

The obvious question is since most of the vendors are small and very early stage – is there truly a market for liquid cooled servers? Industry analysts believe there is and forecast the market to grow from about $110 million in 2015 to almost $960 million in 2020, an additional $850 million of incremental revenue in just five years.

With healthy future growth prospects, we’ve started to see larger players enter this market such as Fujitsu. In addition, the HPC system vendors are all OEMing liquid cooling technology solutions to solve the big system cooling issues in the datacenters. With the huge increase in data being generated, artificial intelligence and other applications need to mine this data. Consequently, more and more server power is required and new innovative cooling solutions needed making liquid cooling a practical and feasible solution.

As a side note, more and more government RFPs are asking for liquid cooling solutions. Solutions such as the one from Fujitsu can make the crossover from HPC to commercial datacenter a reality.

Could 2017 the breakout year for liquid cooling, move from innovator to early adopter?

The Supercomputing Conference is frequently a window into the future. At SC16, there were over a dozen companies demonstrating server liquid cooling solutions, with technologies ranging from Direct-to-Chip to Liquid Immersive Cooling where servers and storage are fully immersed in dielectric fluid.

Today the majority of providers are early stage or startup companies with a notable exception. Fujitsu, a global IT powerhouse brought over thirty years of liquid cooling experience and demonstrated an immersive cooling solution that had Intel-based servers, storage and network switches fully immersed in Fluorinert.

We will see cooling technology move from the confines of high-end supercomputers to a nice niche in the enterprise datacenter for such workloads as big data analytics, AI and high frequency trading.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from its predecessors, including the red-hot H100 and A100 GPUs. Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. While Nvidia may not spring to mind when thinking of the quant Read more…

2024 Winter Classic: Meet the HPE Mentors

March 18, 2024

The latest installment of the 2024 Winter Classic Studio Update Show features our interview with the HPE mentor team who introduced our student teams to the joys (and potential sorrows) of the HPL (LINPACK) and accompany Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the field was normalized for boys in 1969 when the Apollo 11 missi Read more…

Apple Buys DarwinAI Deepening its AI Push According to Report

March 14, 2024

Apple has purchased Canadian AI startup DarwinAI according to a Bloomberg report today. Apparently the deal was done early this year but still hasn’t been publicly announced according to the report. Apple is preparing Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimization algorithms to iteratively refine their parameters until Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the fi Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimizat Read more…

PASQAL Issues Roadmap to 10,000 Qubits in 2026 and Fault Tolerance in 2028

March 13, 2024

Paris-based PASQAL, a developer of neutral atom-based quantum computers, yesterday issued a roadmap for delivering systems with 10,000 physical qubits in 2026 a Read more…

India Is an AI Powerhouse Waiting to Happen, but Challenges Await

March 12, 2024

The Indian government is pushing full speed ahead to make the country an attractive technology base, especially in the hot fields of AI and semiconductors, but Read more…

Charles Tahan Exits National Quantum Coordination Office

March 12, 2024

(March 1, 2024) My first official day at the White House Office of Science and Technology Policy (OSTP) was June 15, 2020, during the depths of the COVID-19 loc Read more…

AI Bias In the Spotlight On International Women’s Day

March 11, 2024

What impact does AI bias have on women and girls? What can people do to increase female participation in the AI field? These are some of the questions the tech Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Analyst Panel Says Take the Quantum Computing Plunge Now…

November 27, 2023

Should you start exploring quantum computing? Yes, said a panel of analysts convened at Tabor Communications HPC and AI on Wall Street conference earlier this y Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Training of 1-Trillion Parameter Scientific AI Begins

November 13, 2023

A US national lab has started training a massive AI brain that could ultimately become the must-have computing resource for scientific researchers. Argonne N Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire