Ever since the first modern computers were designed and built, how to keep IT equipment suitably cool – to ensure safety, longevity and reliability – has been high on the list of design considerations. Heat sinks, vents, fans, and air-conditioning are all either options or necessities and are common design features. But a step-change in improvements to cooling is much-needed and well overdue, and here’s why.
Current data centre energy efficiency
First, let’s look at ‘what already is’. The average efficiency of data centres in 2021 – i.e., data centres with ‘traditional’ servers cooled by ‘traditional’ means – was 1.57 PUE (63.5% DCIE). In layman’s terms: of the total electricity used by an average data centre, 63.5% is consumed by the IT load. The remaining, substantial 36.5% goes to the cooling and general running of the facilities.
Let’s put both these percentages into context. Data centres already account for about 1% of the world’s energy demand, proportion forecast to rise significantly. A calculator is not required to realise that even small improvements to PUE/DCIE metrics have the potential to put a significant dent into global energy demands.
Coming to the attention of authorities
Neither the current efficiency levels of data centres nor their absolute levels of energy consumption have gone unnoticed by various authorities.
For example, the European Commission has declared that “Data centres and telecommunications will need to become more energy efficient, reuse waste energy, and use more renewable energy sources. They can and should become climate neutral by 2030.” In the context of the lifespan of a server, this is not far away.
Meanwhile, the absolute levels of energy consumption by data centres bring into play the prospect of regulation. In Ireland, for example, data centres were responsible for 14% of all electrical consumption. It is predicted that this proportion could rise to 25% by 2030. This is causing energy concerns.
New processors add to the pressure for new approaches
Second, consider what’s about to be, because it’s going to exacerbate the situation. The Thermal Design Power (TDP) of new processors released by Intel and AMD has increased from 270W for Intel and 280W for AMD to 350W and 400W, respectively. This is one of the largest increases in yearsu and will push air-cooling towards its limits.
Taken together, these factors alone are strong reasons for manufacturers to reexamine their approaches to cooling. It’s also a reason for buyers of servers to think carefully about their ability to meet future energy efficiency targets, whether optional or, as seems increasingly likely, mandatory.
Server-level innovations
Necessity is – still, and often – the mother of invention. Of necessity, new approaches to cooling have been developed and explored, and are finally starting to impact server and data centre design.
The two most market-ready new cooling technologies are direct-to-chip and liquid immersion. Of these two, immersion cooling currently shows more potential in improving data centre efficiency. Immersion cooling enables a greater density of servers in the data centre. Such centres have already achieved a PUE of 1.03 (97% DCIE).
An additional benefit is that most immersion cooling tanks are closed-loop systems, meaning zero water waste. It is so efficient that manufacturers can continue building high-density servers, with their ever-increasing power consumption, without worrying about cooling. Netted-off, immersion cooling is expected to reduce the total cost of ownership of servers by 25-40%.
Data centre-level innovations
Of course, improvements to server design will generate consequences for, and drive innovations in, data centre design. There are already experiments using natural cooling environments such as underwater data centres. Another nice example is a small fuel cell-powered data centre in Luleå in northern Sweden. It uses the “waste” heat from the data centre to supply the local district with heating.
Such scenarios are interesting, but the efficacy of such approaches is ultimately likely to depend on applications. This is because location is also very important to the end user. Finding cooling innovations that can be built anywhere, and which are effective anywhere, will ultimately be more advantageous.
So, who should be responsible?
Vendors and manufacturers must invent and implement new technologies that increase efficiency, lower power consumption and minimise the environmental impact of servers. And, indeed, they are – with the result that data centre efficiency is likely to improve naturally as a result of standard equipment replacement cycles.
But ultimately, responsibility is shared. Vendors and manufacturers may guide data centre operators on how to meet these new challenges. But customer prioritisation of energy efficiency must also change significantly.
New research commissioned by Asus shows that over 70% of organisations with fewer servers – 10 or less – say that energy efficiency is a factor in their server purchasing decision. But for organisations with 11 or more servers – those with the most to gain or lose from increased server energy efficiency – this figure drops to just 34%.
If we are, collectively, to increase the sustainability of data centres more quickly, then more active consideration of new cooling technologies by customers is likely to be needed.
ASUS employs over 5,000 R&D professionals globally, in more than 1,000 service centres covering 98 countries. The company has built high-quality servers and workstations for 25 years, designed for data-centre environments of all sizes and for diverse purposes such as content creation, cloud gaming, AI and high-performance computing. Its server range includes rack, GPU and high-density servers, while its workstation range comprises ‘performance’, ‘mainstream’ and ‘essentials’ systems. Its products are backed by the industry’s only five-year warranty and an outstanding channel that make ASUS the most trusted partner in the data centre.
