It’s not just households feeling the pinch as the cost of living continues to increase. Businesses are also finding their costs going up, in no short amount due to the rising cost of energy. Even as prices start to level out or even drop off a little, the cost of electricity is still eye-wateringly high.
For most organizations, one of the biggest contributors to their energy costs will be their IT estate, with the data center (or data centers) taking the lion’s share of that if present.
From the temperature the air conditioning unit is set to, to whether you use HDDs or SSDs, everything can make a difference – right down to the type of chips your servers are using.
Keeping it cool
When thinking about making your data center more efficient, there are certain clear areas to address. It’s likely that cooling is the second most power-hungry activity after actually powering the servers. There are a few things you can do to address this: If you’re building a new data center or overhauling one already in situ, consider moving from wall-mounted air conditioning to floor-based cooling, adiabatic cooling towers, or rear door cooling.
It’s also worth checking what the manufacturer’s recommended running temperature is for your servers – nowadays it’s possible it could be closer to 25ºC than the roughly 20ºC of the past. Raising the temperature of the data center by a few degrees may seem like a small intervention, but it can have a significant impact on cost.
Take a look at the servers themselves, too. Options like variable fan speed or liquid cooling – which is far more efficient than air cooling – can reduce the total amount of energy required to cool the data center.
Processor efficiency
Given their diminutive size, it’s easy to overlook the importance of processors in reducing energy consumption in the data center. However, they can be a significant contributor to overall energy costs.
Migrating to modern servers within data centers can decrease service workloads and enable a smaller physical footprint.
What’s more, as Moore’s Law continues to hold with the numbers of transistors on a silicon chip roughly doubling each year, they have become increasingly power hungry.
With this in mind, it’s worth considering installing more energy efficient CPUs and GPUs in your servers in order to reduce overall energy consumption even as power and performance increase.
AMD’s 4th Gen EPYC CPUs, for example, are by design some of the most efficient on the market, without sacrificing on performance. For example, according to the company’s analysis running 2,000 virtual machines on 11 2P AMD EPYC 9654 processor-powered servers will use up to 29% less power each year than a similarly-specced competitor’s hardware while still delivering the same performance.
The company even has a roadmap for energy efficiency in two or the most demanding settings for any processor: High Performance Computing (HPC) and artificial intelligence (AI) training.
Its stated goal is to deliver a 30x increase in energy efficiency over five years for AMD processors and accelerators powering servers for HPC and AI, starting from 2020. This, the company claims, would save “billions” of kilowatt hours of electricity and reduce the power required to complete a single calculation by 97% by 2025.
While this goal may be targeted at HPC and AI-training environments, it’s an ethos that filters down into all its data center chips and is worth bearing in mind when considering any hardware upgrade.
Beyond the bottom line
While money can be a major motivator in the quest for energy efficiency, it’s not the only one. Energy consumption can considerably impact an organization’s environmental profile, determining whether potential customers and partners see it as desirable or not. In the near future, a lack of care and attention when it comes to contributions to global carbon emissions caused through high energy use may also lead to investors and even governments penalizing companies that fail to change their ways.
IT finds itself once again at the center of this equation.
According to 451 Research’s Digital Pulse user survey, carried out in July 2022, 254 out of 576 global IT decision makers said that IT operations account for all (19% of respondents) or most (25%) of their organization’s environmental impact. A similar number (41%) said modernizing their core infrastructure is one of their top tech strategies for reducing this impact.
In this way, moving to more energy efficient processors in the data center serves a dual purpose, especially as more demanding workloads such as analytical and generative AI become the norm. Organizations can take an approach that reduces capital expenditure (CAPEX) by reducing energy bills, which in turn ultimately reduces total cost of ownership (TCO) of any given server, and also reduces their environmental impact.
A strengthened balance sheet thanks to lower energy costs than would otherwise be the case, plus a positive environmental profile can make an organization a more attractive commercial partner. This is especially true for businesses with an environmental strategy that includes how they choose their suppliers.
Who would have thought the humble CPU and GPU could make such a difference to environmental strategy and business partnerships, as well as their regular roles?
