
Sequoia
IBM Blue Gene/Q supercomputer built for US nuclear weapons simulation at Livermore
Sequoia was a petascale Blue Gene/Q supercomputer built by IBM for the National Nuclear Security Administration under the Advanced Simulation and Computing Program. Delivered to Lawrence Livermore National Laboratory in 2011 and fully deployed in June 2012, it topped the TOP500 that month before being dismantled in 2020.
- Named the world's fastest supercomputer on 14 June 2012
- 1,572,864 processor cores across 96 racks
- 17.17 petaflops LINPACK, 63% ahead of the K computer
- 2068 Mflops per watt, about six times the Dawn prototype
- First single application to run on over one million cores, January 2013
What it is
Sequoia was a petascale supercomputer of the Blue Gene/Q family, assembled by IBM for the National Nuclear Security Administration within the Advanced Simulation and Computing Program. It sat at Lawrence Livermore National Laboratory in Livermore, California. The machine packed 98,304 compute nodes across 96 racks, giving 1,572,864 processor cores and 1.5 PiB of memory in roughly 3,000 square feet. Its main job was nuclear weapons simulation, replacing the older Blue Gene/L and ASC Purple systems, though it also ran work in astronomy, energy, lattice QCD, the human genome and climate change. In June 2012 it was named the fastest computer on Earth.
How it came about
IBM first built a smaller Blue Gene/P prototype called Dawn, rated at 500 teraflops, to test ideas before the full machine. Dawn arrived in April 2009 and reached ninth place on the TOP500 that June. Sequoia itself was delivered to Livermore in 2011 and finished deployment in June 2012. On 14 June 2012 the TOP500 Project Committee placed it first, ahead of the K computer. It held third place in June 2013 behind Tianhe-2 and Titan, had fallen to sixth by November 2017, and stood at 22nd in the November 2019 listing. The system was taken apart in 2020 to make room for El Capitan.
The compute
Each compute node held a 16-core A2 processor chip with 16 GB of DDR3 memory, 1,024 nodes per rack. Nodes were linked by a 5-dimensional torus network. The whole system ran Linux: the CNK operating system on more than 98,000 compute nodes, and Red Hat Enterprise Linux on 768 I/O nodes attached to a Lustre parallel filesystem. LLNL used the SLURM job scheduler, the same tool used by Dawn and China's Tianhe-IA. ZFS was ported to Linux as the Lustre object storage device. In September 2011 NetApp said the Department of Energy had chosen it to supply 55 PB of storage.
Where the GPUs are
The machine occupied about 3,000 square feet at Lawrence Livermore National Laboratory in Livermore, California. Power draw for the complete system was around 7.8 MW, with the infobox figure given as 7.9 MW. That was 37% below the 12.6 MW used by the K computer it displaced at the top of the TOP500. Efficiency reached 2068 Mflops per watt, roughly six times the Dawn prototype and more than twice the June 2011 TOP500 leader.
Customers and deals
Sequoia was built for the National Nuclear Security Administration and operated at Lawrence Livermore National Laboratory. Its primary workload was nuclear weapons simulation. Time was also given to other science: astronomy, energy research, lattice QCD, human genome study and climate change. In January 2013 Stanford Engineering's Center for Turbulence Research used it to predict noise from a supersonic jet engine, the first single application to use more than a million cores at once.
Money and ownership
The cost figure given is $250 million, a sum IBM did not disclose. NetApp was selected in September 2011 to provide 55 PB of storage for the Department of Energy.
Its place in AI
In June 2012 Sequoia took the top spot on the TOP500 from the K computer, with LINPACK performance of 17.17 petaflops against 10.51 for its rival, and 123% more cores. It was the first system to sustain more than 10 petaflops on science codes. The cosmology framework HACC reached almost 14 petaflops on a 3.6 trillion particle run, and Cardioid, which models the electrical activity of the human heart, came close to 12 petaflops in near real time.
Questions and criticism
The machine drew about 7.8 MW, though this was well below the power used by the system it replaced at number one. Its cost of $250 million was not confirmed by IBM. The system was dismantled in 2020 to free space for El Capitan.


Questions people ask about Sequoia
What was Sequoia built for?
IBM built it for the National Nuclear Security Administration under the Advanced Simulation and Computing Program. Its main task was simulating nuclear weapons, replacing Blue Gene/L and ASC Purple at Livermore. It also ran astronomy, energy, lattice QCD, genome and climate work.
How fast was it?
It reached 17.17 petaflops on the LINPACK benchmark when it topped the TOP500 in June 2012, 63% faster than the K computer. The infobox gives a speed of 20.13 PFLOPS. Science codes saw almost 14 petaflops from HACC and close to 12 from Cardioid.
What hardware did it use?
It was a Blue Gene/Q design with 96 racks and 98,304 compute nodes. Each node carried a 16-core A2 chip and 16 GB of DDR3 memory, giving 1,572,864 cores and 1.5 PiB in total. Nodes were joined by a 5-dimensional torus network.
What software ran on it?
The whole machine ran Linux. CNK, the Compute Node Kernel, ran on more than 98,000 compute nodes, while Red Hat Enterprise Linux ran on 768 I/O nodes linked to a Lustre filesystem. LLNL scheduled jobs with SLURM and ported ZFS to Linux for storage.
What happened to it?
Sequoia was delivered in 2011, fully deployed in June 2012, and taken apart in 2020 to make room for El Capitan. Its last TOP500 appearance was 22nd place in the November 2019 list.
Good reading and sources
- Sequoia (supercomputer) on WikipediaEncyclopedia
- More photos on Wikimedia CommonsPhotos
- Wikidata recordOpen data
The facts on this page come from the Wikipedia article "Sequoia (supercomputer)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryNeocloud rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like Sequoia
All AI supercomputers
🇨🇳 ChinaTianhe-1
Chinese petascale supercomputer at the National Supercomputing Center in TianjinSupercomputer · Other Nvidia GPUs · 2009