Altix

[4] Later, SGI Altix 512-processor systems were officially supported[5] using an unmodified, standard Linux distribution with the launch of SLES 9 SP1.

Besides full support of SGI Altix on SUSE Linux Enterprise Server, a standard and unmodified Red Hat Enterprise Linux was also fully supported starting with SGI Altix 3700 Bx2 with RHEL 4[6] and RHEL 5[7] with system processor limits defined by Red Hat for those releases.

SGI later officially supported 1024-processor systems on an unmodified, standard Linux distribution with the launch of SLES 10[9] in July 2006.

NASA's Columbia supercomputer, installed in 2004 and decommissioned in 2013, was a 10240-microprocessor cluster of twenty Altix 3000 systems, each with 512 microprocessors, interconnected with InfiniBand.

It features DDR SDRAM and PCI-X expansion ports, and can support SCSI or SATA internal hard drives.

The 3700 is based on the third generation NUMAflex distributed shared memory architecture and it uses the NUMAlink 4 interconnection fabric.

The basic building block is called a C-brick, which contains two nodes in a 4U high rackmount unit.

Each node contains two Intel Itanium 2 processors that connect to the Super-Bedrock application-specific integrated circuit through a single front-side bus.

The Altix 3700 Bx2 is a high-end model supporting 16 to 2,048 Itanium 2 processors and 12 GB to 24 TB of memory.

Each node is contained within a blade that plugs into an enclosure, the individual rack unit (IRU).

Compute blades contain a processor node and consist of two PAC611 sockets for Itanium 2 and Itanium microprocessors, a Super-Hub (SHub) application-specific integrated circuit (ASIC) (chipset) and eight dual in-line memory module (DIMM) slots for memory.

Multiple servers can be combined on the same Numalink fabric up to the theoretical maximum of 8,192 nodes (16,384 OS CPUs).

While the earlier Itanium-based Altix systems run a single-system image (SSI) Linux kernel on 1024 processors or more using a standard SuSE Linux Enterprise Server (SLES) distribution, the Altix ICE's clustering capabilities use standard SLES or Red Hat Enterprise Linux distributions and scale to over 51,200 cores on NASA's Pleiades supercomputer.

Both configurations support either hypercube or fat tree topology, and 16 compute blades within an IRU.

Altix UV systems run either SuSE Linux Enterprise Server or Red Hat Enterprise Linux, and scale from 32 to 2,048 cores with support for up to 16 terabytes (TB) of shared memory in a single system image.