Supercomputers/Systems/BlueGene/L
USA · decommissioned · classified
BlueGene/L
Also known as BG/L, Blue Gene/L
Operated by Lawrence Livermore National Laboratory at Livermore Computing Center .
Researched with AI assistance, checked by scripts. These rows were researched by AI agents reading public operator, laboratory and vendor sources, appended by scripts and checked by automated validation; no subject-matter expert has reviewed them (how this site is made). A mechanical second-reader pass has since fetched every cited source: 1350 of 1581 systems have a readable citation that names them, and 64 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.
Measured performance
- Rmax
- 478.20 TFlop/s
- Rpeak
- 596.38 TFlop/s
- Rmax ÷ Rpeak
- 80.2%
- Cores
- 212,992
- Power
- 1.43 MW
- Per watt
- 0.3 GF/W
Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.
How it compares
Ranked against the other machines in this dataset that report the same figure: those that first came into service within two years of BlueGene/L, and those in USA. Computed from our tables; a metric is ranked only when at least eight peers report it.
| Metric | Value | Against its era | Against its country |
|---|---|---|---|
| Rmax | 478.20 TFlop/s | #1 of 67 · highest | #205 of 332 · bottom 39% |
| Energy efficiency | 0.3 GF/W | #1 of 11 · highest | #139 of 162 · bottom 15% |
| Rmax ÷ Rpeak | 80.2 % | #19 of 67 · top 27% | #110 of 331 · top 33% |
| Power draw | 1,433 kW | #3 of 12 · top 18% | #45 of 171 · top 26% |
Every metric is ranked largest-first, so a high power-draw rank means a big power draw, not a good one. Rmax ÷ Rpeak uses HPL-measured results only. Rankings move as machines are added; they describe this dataset, not the world.
What this machine is made of
One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.
| Role | Part | Supplier at build | Quantity | Confidence | Source |
|---|---|---|---|---|---|
| Integrator | - IBM eServer Blue Gene Solution | IBM | - | Reported | ibm.comhpc.llnl.gov |
| CPU | IBM PowerPC 440 700MHz CPUs PowerPC · 2C · 0.7 GHz · 130 nm · 13 W PowerPC 440 700MHz, 106,496 nodes | IBM | 106,496 reported | Reported | ibm.comhpc.llnl.gov |
| Interconnect | IBM BlueGene 3D torus Interconnect 3D torus, plus separate collective and barrier networks · 175 MB/s per link per direction integrated 3D torus | IBM | - | Reported | ibm.comhpc.llnl.gov |
| Storage | Lustre Storage Parallel filesystem, separate metadata and object servers connected to a 900 TeraByte Lustre file system | IBM | - | Reported | osti.gov |
| Cooling | Air cooling (CRAC/CRAH) Cooling Room-level air handling · Practical ceiling around 15-20 kW per rack IBM BlueGene/L: air cooled | IBM | - | Reported | ibm.comhpc.llnl.gov |
| Operating system | IBM Compute Node Kernel Operating systems Microkernel, no dynamic scheduling · Proprietary Compute Node Kernel | IBM | - | Reported | ibm.comhpc.llnl.gov |
Engineering, funding and access
What BlueGene/L runs on, who paid for it and how people get time on it. Each line carries who said it and when.
Funding
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Funded by | US Department of Energy ASCI Advanced Architecture Research Program, in partnership with IBM | 2020-11 | Operator stated | llnl.gov |
| Procurement | Partnership between IBM and LLNL; sited at Livermore in 2004; targeted 360 TFlop/s peak and predicted at least 15 times faster, 15 times more power efficient and about 50 times smaller per computation than contemporaries | 2020-11 | Operator stated | llnl.gov |
Results
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Notable result | 2007 Gordon Bell award for a Kelvin-Helmholtz instability simulation in molten metals, a joint LLNL and IBM team (2007) | 2007 | Operator stated | llnl.gov |
| Notable result | 2006 Gordon Bell Prize for a large-scale electronic structure simulation of molybdenum (2006) | 2006 | Operator stated | llnl.gov |
| Notable result | 2005 Gordon Bell Prize for peak performance: solidification of tantalum and uranium, the first scientific application above 100 TFlop/s sustained (2005) | 2005 | Operator stated | llnl.gov |
| Notable result | First place on the November 2004 TOP500 with 70.72 TFlop/s HPL; peaked at 596 TFlop/s after a series of expansions (2004)LLNL says it stayed fastest until June 2008; a DOE page says No. 1 from November 2004 to 2007. | 2004-11 | Operator stated | llnl.gov |
The read
IBM's argument that the right way to scale was many deliberately slow, low-power embedded cores rather than fewer fast ones. Held the top position for three and a half years, longer than any system since. Its bill of materials is unusually short: IBM supplied the processor, the integrated 3D torus interconnect, the compute node kernel and the integration. There is no third-party fabric vendor and no accelerator vendor: a vertical integration pattern that has since reappeared, in different hands, at Fujitsu and at NVIDIA. Retirement year is not recorded here because we have not found a primary source for the exact date.
Timeline
- 2001 Announced source
- 2004-11 List debut at #1 (November 2004) source
- 2004-11 First operational source
- 2004-11 Took first place on the TOP500 with 70.72 TFlop/s on HPL, the start of its run as the world's fastest computer source
- 2005-11 LLNL won the 2005 Gordon Bell Prize on Blue Gene/L, the first scientific application to exceed 100 TFlop/s sustained source
Public list appearances
Pointers only: rank and edition, linking to the canonical entry. We do not reproduce list tables. See sourcing policy.
Site & facility
Livermore Computing Center
- Location
- Livermore, CA, USA
- Cooling
- Direct liquid cooling (warm water)
- Commissioned
- 1996
Where sources disagree
We record conflicts instead of picking a winner quietly. All open questions.
- End of its time at number one: Fastest until June 2008, peaking at 596 TFlop/s against Number one from November 2004 to 2007. LLNL and the NNSA page give different end dates for the same run at the top of the list.
Nearby in scale
The machines closest to BlueGene/L by Rmax, smaller to larger.
- Shepard · USA, operational · 0.48 PFlop/s
- Mesabi · USA, operational · 0.47 PFlop/s
- iREMB · South Korea, operational · 0.47 PFlop/s
- Vestas1 · Denmark, operational · 0.48 PFlop/s
- Amazon EC2 C3 Instance Cluster · USA, operational · 0.48 PFlop/s
- Unayzah (2017) · Saudi Arabia, operational · 0.48 PFlop/s
Related systems
Same site
- Thunder · USA, decommissioned, 2004
- ASC Purple · USA, decommissioned, 2005
- MCR · USA, decommissioned, 2002
- Atlas (LLNL) · USA, decommissioned, 2007
- ASCI White · USA, decommissioned, 2000
- Hera (LLNL) · USA, operational, 2009
Same operator
- ASCI Blue-Pacific · USA, decommissioned, 1998
- Edge (LLNL) · USA, operational, 2010
- Zin · USA, decommissioned, 2011
- Cab · USA, decommissioned, 2012
- Sequoia · USA, decommissioned, 2012
- Vulcan · USA, decommissioned, 2012
Source check
We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.
| Cited source | Result | Found on the page |
|---|---|---|
| hpc.llnl.gov | does not name this system | IBM Compute Node Kernel |
| ibm.com | names system and part | IBM Compute Node Kernel |
Checked 2026-10-18 by pnpm verify. Re-run it and the table changes with the web.
Further reading
Power and cooling
Why air ran out at 20 kW a rack, how warm-water cooling removes the chillers, and what PUE conceals.
Inside a node
Latency-optimised versus throughput-optimised processors, NUMA, and the end of the host-to-device copy.
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.