Every June, the publication of the Top500 list provides a snapshot of progress in High-Performance Computing (HPC).
This year, alongside the latest rankings, another announcement attracted attention. The long-term stewardship of Top500 is beginning a transition to ACM SIGHPC, helping ensure that one of the community’s most recognised projects continues to serve future generations of researchers, operators, and technology providers.
It is a timely reminder that HPC is constantly evolving. Systems change. Workloads change. Communities change. Over time, our definitions of success often change as well.
Performance Still Matters
For more than thirty years, Top500 has helped answer an important question: how much computational performance can a system deliver?
Benchmark rankings have played a valuable role in helping the HPC community track technological progress, compare architectures, and understand how computing capability continues to advance across research, industry, and government organisations.
Performance remains important. The demand for computational capability continues to grow across simulation, modelling, data analysis, and AI. Faster systems enable new discoveries, larger workloads, and increasingly ambitious research programmes.
Yet benchmarks do more than measure technology. They also influence how we think about progress.
The metrics a community chooses to publish, celebrate, and discuss help shape collective priorities. They provide a shared language for comparison and a way to understand how the field is advancing. As HPC continues to mature, many organisations are beginning to ask whether a single measure can capture the full picture of what modern computing environments are expected to deliver.
This is not a challenge to the value of Top500. Rather, it reflects the reality that today’s systems operate within increasingly complex technical, operational, and organisational environments.
The Story Behind the Podium
One conversation at ISC26 led to an interesting comparison.
While discussing benchmarks, productivity, and system outcomes, the topic shifted to marathon running.
The winner of a marathon is easy to identify. Finishing times provide a clear measure of performance, and the podium tells us who came first.
What the podium does not show is everything that made that result possible.
Months of training. Coaching. Nutrition. Recovery. Support networks. Consistency. Preparation. The countless decisions that allow an athlete to arrive at the starting line ready to compete. The finishing time remains important, but it is only part of the story.
Computing systems are increasingly similar. Benchmark results tell us something valuable about a platform’s capabilities. They do not necessarily tell us how easy it is to use, how effectively it supports users, how sustainable it is to operate, or how successfully it enables research and innovation.
Those outcomes are influenced by many factors beyond the hardware itself. Software environments, documentation, operational processes, training, community knowledge, and long-term stewardship all contribute to the value a system ultimately delivers.
In many ways, benchmark results are the podium. What happens behind the scenes is what allows organisations to reach it.
The Metrics We Choose
One of the more interesting aspects of any metric is that it is never entirely neutral.
Every measure reflects a decision about what is important.
Top500 elevated performance because computational capability was a defining challenge for the field. Measuring performance provided a way to track progress and celebrate achievement.
Later, the Green500 emerged as energy efficiency became an increasingly important consideration. The introduction of a second benchmark did not diminish the value of the first. Instead, it acknowledged that another dimension of success deserved attention. In both cases, the metric helped focus discussion around an area the community considered valuable. The measures we create tell us something about our technology, but also about our priorities.
As HPC and AI continue to evolve, those priorities are broadening. Conversations at ISC26 frequently touched on software portability, operational maturity, workforce development, sustainability, governance, accessibility, and long-term service delivery.
Many of these topics are difficult to capture in a single benchmark. Yet they increasingly influence whether a system is considered successful by the people who build, operate, fund, and use it.
This should not be interpreted as a search for a new universal metric. If anything, it suggests that success is becoming more multidimensional. Different measures reveal different forms of value. Performance remains essential. Efficiency remains essential. But organisations are increasingly recognising that outcomes matter too.
Expanding the Definition of Success
Perhaps this is one of the more encouraging developments in the field.
As computing environments become more capable, our understanding of success is becoming richer. Rather than replacing one metric with another, the HPC community is gradually expanding the conversation about what progress looks like. The ability to compare systems and track technological advancement remains valuable; yet the future of HPC is likely to be shaped not only by what we measure, but also by what we choose to value.
The challenge is no longer simply building powerful systems. It is creating environments where technology, software, operations, and people work together to produce meaningful outcomes.
In that sense, the definition of success in HPC may be expanding in much the same way the field itself continues to expand.
—
This article grew out of a conversation at ISC26. Like many of the questions facing HPC, it’s a discussion that is still evolving.
We regularly share similar reflections, field notes, and observations through the Alces Flight newsletter.
You can subscribe here if you’d like to follow along.