Cutting-edge server chips are getting a lot of attention for their potential to boost performance, but next-generation processors also run hotter than older designs, and data center operators will have a hard time figuring out what to do with limited guidance from chipmakers.
At the same time, the role IT equipment can play in energy efficiency efforts will come under increasing scrutiny. These interconnected trends are one of the top predictions for the data center this year from the Uptime Institute.Edge computing gateway
“Operators will be working hard to use new, hotter server chips,” Jacqueline Davis, a research analyst at Uptime, said during a webcast on the institute’s 2023 Data Center Forecasts. At the same time, “the focus on energy efficiency will expand to include the IT equipment itself, which we believe is long overdue.
Server heat rises
The data centers being built today need to remain economically competitive and technologically capable in 10 to 15 years, yet new chip technologies are causing operators to question traditional data center design guidelines.
0 seconds 0 seconds 13 minutes 48 seconds Trading volume 0%
This video file cannot be played. (Error code: 232001)
“Data center design must respond to server power and cooling requirements, which have remained consistent over the years. Designers can plan for four to six kilowatts per rack,” said Daniel Bizo, research director at Uptime. “Continuous IT updates do not require upgrades to power or cooling infrastructure.”
That’s changing now. Power density per rack and per server box is rising. For example, Intel’s 4th generation Xeon Scalable processor, code-named Sapphire Rapids, has a thermal design power (TDP) of up to 350 watts, and AMD’s 4th-generation Epyc processor, code-named Genoa, has a TDP of up to 360 watts.
“Future product roadmaps call for mainstream server processors with 500 to 600 watt TDP within the next few years,” Bizo said. “So this trend will soon start to destabilize facility design assumptions as we see mainstream Servers are approaching or exceeding one kilowatt.
Dedicated GPU-based high-performance computing (HPC) systems can require hundreds of watts per chip at peak power. In addition to high thermal power, they also have low temperature limits.
“They effectively put a double bind on the cooling system because they generate more heat energy and many of them also require lower operating temperatures,” Bizo said. Removing large amounts of heat to achieve cryogenic temperatures is technically challenging, he said, which will force operators to approach cooling differently. For example, some data center operators will consider supporting direct liquid cooling.
The design dilemmas posed by niche HPC applications can be considered an early warning of the power consumption and cooling challenges that high TDP processors will bring to the mainstream enterprise server market. “Now this requires some guesswork,” Bizo said. “How powerful is a typical IT rack? How powerful will high-density racks become? What cooling models will data centers need to support by the end of this decade?”
A conservative approach may continue with low-density rack designs, but this increases the risk of the data center becoming too limited or even obsolete before its time. However, Bizo warns that more aggressive design approaches require highly dense racks, increasing the risk of underutilized capacity and capacity overruns.
“Operators will be faced with various choices when dealing with new generations of IT technology. They can limit air temperatures and accept efficiency losses. Or as [U.S. industry body] ASHRAE recommends Class H1 [thermal standards], they can limit air temperatures for these IT systems create dedicated environments to minimize the hit to facility efficiency. Or they can facilitate conversion to direct liquid cooling,” Bizo said. “While semiconductor physics is the fundamental physical driver here, infrastructure economics will impact the outcomes of data center design and operations. With little guidance from chipmakers, adaptability will be key.
Energy efficiency shifts focus to IT
Uptime predicts that enterprises and data center operators will continue to face pressure for greater IT efficiency.
In the past, the responsibility has been on the facilities side. Data center operators have borne the brunt of sustainability and efficiency requirements and have made huge strides in power usage effectiveness (PUE) over the past 15 years or so, Bizo said. (PUE measures the ratio of energy used by IT equipment to the energy used by the entire data center. Through initiatives such as air separation, more flexible temperature allowances, tighter cooling, fan and power distribution controls, and the use of renewable energy, facilities are improving on their resources. Utilization is more efficient.
“Data center operators are doing a lot on the sustainability front. At the same time, by and large, IT has been absent from these discussions,” Bizo said. “This is becoming increasingly untenable. Why? Well, simply because, what’s left in terms of efficiency improvements is hidden in IT energy performance.
According to Uptime, pressure is increasing in four key areas:
Municipal resistance to new large-scale data centers: Concerns about power and land availability have led to greater restrictions on new data center construction since 2019 and are likely to intensify. In Frankfurt, Germany, for example, new cloud and hosting construction is restricted to specific areas because of concerns about massive land use and changes to the city’s skyline.
Limited grid power availability to support growing data center capacity: Uptime cited two regions – Dublin, Ireland and Northern Virginia – where the grid is strained and power companies are suspending or limiting new connections due to insufficient generation or transmission capacity.
Increasing regulation on sustainability and carbon reduction, as well as stricter reporting requirements: in the EU, for example, data centers are facing new, more detailed reporting requirements on energy consumption and must make their energy performance indicators public. According to Uptime, similar initiatives are now happening in the United States.
High Energy Costs: Before Russia invaded Ukraine, energy prices and electricity prices were on an upward trajectory. Wholesale electricity forward prices have surged in 2021 in European and U.S. markets. The trajectory remains towards more expensive electricity, which is creating an environment conducive to volatility.
Dealing with high energy prices, carbon reporting and grid capacity shortages has always been the domain of facility operators. But according to Uptime, there are diminishing returns on the facility side – more interventions bring fewer and fewer benefits. In contrast, IT has matured in improving data center energy performance, especially in the area of server hardware.
Underutilized and inefficient server hardware is a key area where IT can improve energy performance. For example, having fewer servers with higher performance metrics can lead to energy gains. Bizo noted that IT can make better hardware configuration choices and use dynamic workload consolidation. Power management features such as energy saving states and power throttling capabilities can also provide significant energy efficiency gains.
Better build IT infrastructure
Geopolitics, cloud and data center costs
To recap, two predictions from Uptime’s 2023 forecast will make IT focus more on server hardware: Power demand is rising as higher-power server processors enter the market; energy efficiency expectations now fall on IT teams, and It’s not just the facilities team.
Uptime’s five data center predictions for 2023 are:
Geopolitics will continue to deepen supply chain concerns: Uptime cited political tensions between the U.S.-led Western alliance, China and Russia as increasing the risk of supply chain disruptions. The company predicts that semiconductor supply chains and undersea cable systems are particularly vulnerable to potential economic and military confrontations.
Cloud migrations will face greater scrutiny: Uptime predicts that in a time of economic stress and uncertainty, the threat of spiraling migration costs and cloud costs will slow or prevent some mission-critical migrations.
Data center costs will rise: IT and data center facility costs have been on a downward trend in recent years. Uptime predicts that this trend has now ended and prices will rise. Supply chain issues, rising energy prices and higher labor costs have all contributed to rising costs. That doesn’t mean higher prices will dampen demand for data centers, but they could drive new efforts to improve efficiency.