World's Fastest Supercomputers

Data

Source: TOP500.org — November 2024 ranking. Performance measured in LINPACK benchmark petaFLOPS (Rmax).

Note: What You Should Know About Supercomputers

A supercomputer isn't just a "really fast PC" — it's a precision-engineered machine built from thousands of processors working together to solve problems that a normal laptop would take years, or even centuries, to crack. The machines ranked on this page are the giants of the field, and understanding how they work makes the numbers above a lot more meaningful. Here's the friendly rundown.

What exactly is a supercomputer?

• A supercomputer is a massively parallel machine that bundles tens of thousands of processors — and often custom accelerators — into a single system.
• Instead of one processor doing a task alone, work is split across millions of cores that crunch away simultaneously, which is why "cores" (see the table) matter so much.
• Systems like Frontier at Oak Ridge pack more than 8.7 million cores, while Fugaku pushes past 7.6 million — numbers that would have been unimaginable a generation ago.
• The top machines aren't your average desktop: they fill entire data-center halls and need their own cooling plants and dedicated power substations.

How is speed measured?

• Performance is measured in FLOPS — floating-point operations per second, the arithmetic steps a machine can do each second.
• A petaFLOPS is a quadrillion (10¹⁵) operations per second. The machines here range from dozens to over 1,700 petaFLOPS.
• Beyond that comes exascale — 10¹⁸ FLOPS, or a billion billion operations every second. That's roughly the speed of a million laptops working together.
• The standard benchmark is LINPACK (the "Rmax" figure), which runs a dense system of linear equations; the TOP500 list ranks machines by this result.

The leading systems

• El Capitan currently tops the list at Lawrence Livermore National Lab with over 1,742 petaFLOPS.
• Frontier at Oak Ridge was the world's first exascale machine in the traditional "first past the goalpost" sense, breaking the exaflops-class barrier and holding the #1 crown before El Capitan; on this page it ranks #2 at about 1,206 petaFLOPS.
• Aurora at Argonne rounds out the American trio, exceeding 1,000 petaFLOPS — one of the largest Intel-based exascale systems.
• Fugaku in Japan, built with ARM-based Fujitsu chips, led the world for years and is famous for its all-around balance, not just peak speed.
• LUMI in Finland and Leonardo in Italy are Europe's heavy hitters, both installed at national supercomputing centers and both powered by AMD accelerators.
• Eagle at Microsoft Azure shows that supercomputing is moving to the cloud, with a massive machine rented on demand rather than owned by one lab.

What are they actually used for?

• Weather and climate — forecasting storms days ahead and modeling climate change over decades.
• Science — simulating proteins to design medicines, modeling fusion reactors, and studying how galaxies form.
• Artificial intelligence — training the huge neural networks behind modern AI, which need enormous compute and memory.
• Engineering and energy — testing aircraft or cars in virtual wind tunnels and helping energy companies explore new resources.
• They're also handy for cryptography, materials science, and the kind of clean-energy battery research that touches everyday devices.

What is the TOP500?

• The TOP500 is the unofficial scoreboard of supercomputing, ranking the world's fastest 500 systems twice a year (June and November).
• It uses a standardized LINPACK benchmark so machines from different vendors and countries can be compared fairly.
• The ranking is volatile — new systems debut every edition, and a machine's position can slip fast as competition heats up.
• Governing bodies, research councils, and chipmakers all watch the list closely because it previews where computing is heading.

Quick tips

• Peak "theoretical" numbers always look higher than real-world LINPACK results — the Rmax figure is what actually counts.
• Higher core counts don't guarantee faster speeds; the interconnect, memory, and software all matter just as much.
• A supercomputer's electricity bill can rival a small town's — energy efficiency is now a big part of next-gen design.

Fun facts

• Frontier's ~8.7 million cores would let every person on Earth have their own dedicated core and still leave spares.
• At exascale, in a single second a machine performs more operations than there are seconds since the Big Bang.
• Fugaku also broke a "supremacy" milestone by simulating the human brain's neural network in near real time.
• Some of these systems draw more power than a typical NASA launch, and their cooling towers can look like something from a sci-fi movie set.