More than half of respondents were motivated by the idea of wanting to do more with less: Expand computing capabilities with lower power. As the number of AI/ML/HPDA workloads performed on HPC systems continues to increase, the drive to expand computing capabilities comes up against the required energy costs. 33% say they encounter prohibitive energy costs, pointing to a more acute issue
Nearly half (49%) of data centers in this survey say they can currently measure their carbon footprint, energy use or environmental impact. Within the next 18 months, the number of sites implementing smart power management will increase to 40% from 27% using it today. We asked respondents to help build a metric that encompasses HPC sustainability in all its facets, based on factors that they view to be important. The most selected factors were energy use for the full data center, energy use efficiency ratios, and energy source (renewable energy, oil, gas). The most used metrics for HPC sustainability are energy use efficiency ratios, so adding energy use of the full data center and energy sources to an algorithm suggests a need for a new approach
60% of respondents say sustainability initiatives will impact their on-premises HPC budgets. Interestingly, half of those respondents said sustainability initiatives will reduce on-premises purchases and the other half indicated that sustainability initiatives will increase them.Plans to reduce on-premises HPC budgets and save money for sustainability initiatives indicate that HPC sites are either devoting resources to optimization efforts, changing where HPC workloads are run, or purchasing new energy efficient hardware or cooling systems.
All priorities that are implemented today are expected to continue 18 months from now, with smart power management/ telemetry (+13%), renewable energy sourcing (+6%), extending equipment life (+3%), and effective equipment retirement (+3%) gaining popularity amongst priorities.
*Chart notes- Efficient hardware configuration (Higher performance per watt) - Optimized workloads (identifying efficient workloads on- premises and in the cloud)- Smart power management (telemetry)- Efficient code (e.g., language, compilers, tools, libraries)- Effective equipment retirement (e.g., take back, recycling, proper disposal)
While performance was selected as the most important factor in on-premises procurements, as is typical of most HPC procurements across the HPC market, the second most important factors selected by respondents suggest that environmental sustainability/ energy efficiency are gaining importance in the procurement process.
87% of respondents say new equipment procurement will play a role in their sustainability plans, but equipmentprocurement timelines vary across sites. Procuring more sustainable equipment is expected to be a part of regularly scheduled procurement plans for 48% of sites, while 20% of sites will expedite procurement plans and 18% will delay procurement plans for sustainable solutions
Add the products you would like to compare, and quickly determine which is best for your needs.