Cirrascale GX8 Series/ Multi-Device Peering Rackmount Platform

Cirrascale GX8 Series Multi-Device Peering Rackmount Platform
Share This Page

Cirrascale GX8 Series Rackmount

The Cirrascale GX8 Series 4U rackmount enables up to 10 PCIe Gen 3.0 x16 devices to peer on a single PCIe root complex with extreme scalability and performance for HPC and deep learning applications.

Product Features and Benefits

  • Supports the Cirrascale SR3615 PCIe switch riser enabling the peering of up to 10 PCIe Gen 3.0 x16 devices on a single root complex.
  • Supports dual Intel® Xeon® Processors E5-2600 v3 family.
  • Supports the NVIDIA® Tesla® M40 GPU Accelerator designed for deep learning, or the Tesla K80 Dual-GPU Accelerators allowing the GX8 Series to handle 16 GPUs per server.
  • A perfect solution for highly parallel applications like those used for deep learning, data analytics and molecular dynamics such as Torch 7, Theano, Caffe, and TensorFlow

Cirrascale GX8 Series Rackmount Server Overview

The new Cirrascale GX8 Rackmount Series supports the Intel® Xeon® processor E5-2600 v3 product families and up to 10 PCIe Gen 3.0 x16 compatible devices enabling multidevice peering on a single PCIe root complex, such as the NVIDIA® Tesla® M40 GPU cards. This multi-device peering on a single PCIe root complex makes it a perfect solution for GPU accelerated applications and libraries like those used for deep learning, data analytics and molecular dynamics such as Torch 7, Theano, Caffe, and TensorFlow.

The GX8 Series rackmount servers are different from other GPU supporting hardware implementations. Most of the hardware configurations available today only provide maximum performance between specific pairs of GPUs; and since GPUs are paired up, jobs requiring communication between arbitrary GPUs experience a performance impact. Additionally, there can be a significant performance impacts with trying to scale more than four GPUs on multi-socket systems. These have been persistent problems for customers who are pushing the limits of GPUs with large, complex data-sets and calculations, or where data must be streamed between GPUs. Cirrascale has been able to overcome these issues, and achieve near linear performance scaling with its design.

By utilizing the Cirrascale SR3615 PCIe 96-lane switch riser, the Cirrascale GX8 supports up to eight NVIDIA Tesla M40 GPU cards and provides room for additional InfiniBand® or NVMe storage devices while enabling increased bandwidth and lower latencies between PCIe Gen3 devices than are possible in traditional systems. By enabling up to 8 discrete GPU accelerators to communicate directly with each other on the PCI bus, free of the need for host CPU intervention, they can create a "micro-cluster", sharing a single memory address space.

Maximize PCIe Bandwidth

Cirrascale is a strong believer in utilizing a technology to its fullest potential whenever possible and GPUs and GPU Accelerators are no exception. If the GPU has a PCIe Gen3 x16 link, then it should use it when communicating with other GPUs — any other GPUs. Our switch riser technology allows us to scale and peer multiple PCIe x16 Gen3 cards on a single root hub ensuring that the maximum PCIe bandwidth is available utilized for inter-card communication.

Minimize Intercard Latency and Obtain Consistent Performance Between GPUs

Our switch riser allows GPUs to communicate as if they are all on the same bus... because they are. Gone are the days of needing a bounce-buffer in host memory, or leaving GPU DMA engines unused because they couldn't address other devices in the system. This reduces intercard latency while helping to maintain a consistent performance level between GPUs.

Enable GPU-Centric Development and Usage

Since most all of the GPU traffic is passed between the GPUs directly via the Cirrascale SR3615 switch riser, a very negligible amount of host resources are needed to perform GPU work. Additionally, with a single address space and simultaneous inter-card communication at full PCIe x16 Gen3 speeds, software can spend more time doing work than thinking about when to schedule data copies.

Supports the Largest Number of GPU Offerings

We work closely with our technology partners to ensure you're given the broadest offerings for your application. The Cirrascale GX8 Series supports both professional and consumer cards from the leading manufacturers including ground-breaking GPU accelerators, such as the NVIDIA® Tesla® K80 Dual-GPU Accelerators and the new Tesla M40 GPU accelerators designed specifically for deep learning applications.

With the consistent advances in technology, Cirrascale engineering and development teams are constantly testing and deploying the latest technical specifications being offered by our technology partners. We make every effort to provide the below specifications error-free and up-to-date. However, we always encourage our customers to contact us to discuss Cirrascale's latest improvements to its products. This product's specifications are as follows:

GX8 Series Multi-GPU Server Specifications

Processor
  • Dual Intel® Xeon® processor E5-2600 v3 product family
Chipset
  • Intel C612 PCH
Memory
  • Supports up to 1TB DDR4 Memory
  • 16 RDIMM Slots - 2400/2133/1866/1600MHz
Storage
  • Up to (8) Removable 2.5" SSD / SATA drive bays
  • (2x Removable standard / Options available for up to 8x)

  • Up to (4) Internal 2.5" or (2) Internal 3.5" SSD / SATA
  • All storage slots may not be available due to configuration options.

Networking
  • 2 x GbE LAN Ports (Intel® I210-AT)
  • 1 x Management LAN
Available Peer-to-Peer PCIe Gen3 Slots
  • 8 x PCIe x16 Gen3.0 (x16 throughput – DWFL)
  • 2 x PCIe x16 Gen3.0 (x16 throughput – SWFL)
  • All slots may not be availalbe due to configuration options

GPU / Add-In Support
  • Supports 8 GPU Graphics Cards (plus 2 SWDL PCIe cards) or up to 10 other PCIe Gen 3 compatible devices in dedicated PCIe x16 Gen 3 slots utilizing Cirrascale's SR3615 PCIe Riser Switch
  • Can support both consumer and professional graphics cards.
Motherboard Expansion Slots
  • 4 x PCIe x16 Gen3.0 (x16 throughput)
  • 2 x PCIe x16 Gen3.0 (x8 throughput)
  • All slots may / may not be available due to configuration options

Integrated Graphics
  • VGA Integrated in ASPEED 2400
  • 1920x1200@60Hz 32bpp
Power Infrastructure
  • Up to (4) 1600W or 2000W Power Supplies
  • Non-redundant and 1+1 redundant configurations available.
  • Input Voltage Required: 208 - 240VAC

Management and Monitoring
  • Complete remote server management with IPMI 2.0-compliant ASMB8-iKVM module and ASWM Enterprise
  • WfM 2.0, DMI 2.0, WOL by PME, PXE
Weight and Dimensions
  • Form Factor: 4U Rackmount
  • Height: 175mm (6.89 in.)
  • Width: 482.6mm (19 in.)
  • Depth: 762mm (30 in.)
  • Weight: Approx. 27kg (60 lbs.)
  • Weight varies due to components used in configurations

Supported Operating Systems
  • Windows® Server 2003 and later
  • RedHat Enterprise Linux 6 and later
  • CentOS 6 and later
  • Ubuntu 12.04 and later
  • Please Call for Other Support

The below links provide access to the detailed data sheets and other resources related to this and potentially several other Cirrascale products and solutions. If you can not find the information you are trying to locate, please contact us directly at (888) 942-3800 or email us at info@cirrascale.com:

Product Literature

PDF Cirrascale GX8 Series Product Data Sheet - 663kB

Request a Quote

Cirrascale supports multiple configurations of its products and prefers to work closely with its customers and partners to determine the best fit for their company's needs. We work hard to listen and understand and can tailor any of our products to your specific requirements. If you have questions, please contact us today at (888) 942-3800 and ask to speak with a Cirrascale Account Manager or use the below button to request a detailed quote.

Request Quote

Deep Learning Case Study

Read about how NYU researchers take on bigger challenges and create deep learning models that let computers do human-like perceptual tasks for research projects and educational programs at the NYU Center for Data Science.

Download Our Case Study on
NYU Center for Data Sciences
Cirrascale's peer to peer technology allows us to unlock the true power of NVIDIA GPUs for scientific computation. This is the first platform of its kind and provides the unprecedented bandwidth needed to scale an individual molecular dynamics simulation with AMBER across all the GPUs within a node. This previously unattainable performance will have impact on many fields.
Ross Walker
Associate Research Professor, SDSC and
Adjunct Associate Professor, UCSD

Share This Page:  

©2016. All Rights Reserved. Cirrascale Corporation..