SGI® UV™ 1000/100/10

SGI UV SGI® UV is the leading scalable shared memory architecture in the industry today. SGI UV provides compelling advantages in performance and economy by leveraging SGI technology to new levels of scalability and price performance. SGI does this by using off-the-shelf hardware, including the standard Intel® Xeon® processor E7 family, and standard off-the-shelf software. The SGI UV can operate unmodified versions of Linux such as SUSE® Linux Enterprise Server and Red Hat® Enterprise Linux®. The SGI UV scalable shared memory system addresses a broad range of the most demanding data-intensive applications.

Solving the World's Most Data Intensive Problems
SGI UV scales to extraordinary levels — up to 256 sockets (2560 cores, 4096 threads) with architectural support to 32,768 sockets (262,144 cores). Support for up to 16TB of global shared memory in a single system image enables SGI UV to remain highly efficient at scale for applications ranging from in-memory databases to a diverse set of data and compute-intensive HPC applications.

Scalable System Size
The SGI UV family consists of three models that enable users to optimally size their systems, achieving the correct balance of compute, memory and storage capability. SGI UV 10 is one of the most capable quad-socket rackmount servers available, whereas SGI UV 100 and SGI UV 1000 can scale a single system image to a maximum of 2560 cores. This is possible because of their innovative NUMAflex®, blade-based architecture.

An Open Platform
The SGI UV architecture was developed to use industry standard technology. This results in the ability to use performance leading Intel® Xeon® processor E7 family CPUs and standard unmodified versions of SUSE Linux Enterprise Server and Red Hat Enterprise Linux. By using industry standard components, the application ecosystem is unrivaled, allowing for any scalable applications to take advantage of SGI UV.

Flexible System Expansion and Storage Capabilities
Industry-standard PCI Express expansion slots open up countless possibilities for system configurations with seamless support for industry-standard networking, storage and graphics/GPU cards. This enables deployment of NVIDIA® Tesla™ GPUs to supercharge performance for many workloads. It also means full support for the entire line of SGI InfiniteStorage hardware and software products.


Feature Benefit
Modular Blade Design
Multiple options in compute blades with respect to number of cores, processor speed and memory
  • Unlimited flexibility and control over system configuration
  • Improved performance density
  • Optimize for any workload (large compute, memory or I/O)
Compatible with Future Technology
Socket compatible with future Intel® Xeon® processors
  • Investment protection
  • Expand on demand
  • Easily upgrade or service individual components
Scalable System Size
Scales to 256 sockets (2,048 cores, 4,096 threads) system size and as much as 64TB memory
  • Superior sustained performance for any programming model or workload
  • Breakthrough performance for data-intensive applications
  • Simplified programming with single system image (SSI)
  • Easy, low-cost administration
NUMAlink® 6 (NL6) Interconnect, MPI Support
High bandwidth, very low latency, MPI Offload Engine
  • Industry-leading cluster application efficiency to hundreds of processors or more
  • Scale up to thousands of CPU cores while retaining coherent memory addressing to petascale
SGI Reliability, Availability and Serviceability (RAS)
Fault prevention, detection, recovery and memory RAS

  • Maximum uptime, higher utilization
Standards-Based Design
Industry standard CPUs, memory and I/O
  • Linux® performance and choice without compromised capability
  • Leading price-performance versus RISC/Unix servers
  • SGI-level quality at an affordable cost
Certified on Multiple Operating Systems
Industry-standard Linux (SUSE® or RedHat®)
  • Runs standard x86 software
  • One-stop support options

SGI UV 1000

The ultimate for compute and large data requirements.

  • 42U high rack
  • 16 blades maximum per enclosure
  • 2 enclosures maximum per rack
  • 64 sockets (640 cores, 1280 threads) maximum per rack
  • 4 rack maximum SSI size
  • Maximum 256 sockets (2560 cores, 4096 threads) per 4 rack system
  • Maximum 8 terabytes per rack (16TB total per SSI)
  • Architectural provisioning for up to 32,000 blades
  • NUMAlink 5 Interconnect

SGI UV 100

For mid-scale environments. Typically referred to as Department or Divisional servers.

  • 3U enclosure with 2 blades installed
  • 12 enclosures can fit into one 40U high rack
  • Maximum system is 2 racks (24 enclosures, 48 blades)
  • 96 sockets / 960 cores (1920 threads) / 12TB max per system (2 racks)
  • NUMAlink 5 Interconnect

SGI UV 10

Ideal for smaller compute and memory requirements. Typically referred to as Workgroup or Department servers.

  • Rack mount - 4U in height
  • Max per rack - 10 per rack
  • 4 sockets / 40 cores (80 threads) / 1TB max per enclosure.
  • Intel® Quickpath Interconnect