By Michael O'Neill (NVIDIA), Brandon Harrell (IBM), and Deana Coble (IBM)
If you use or manage scientific, engineering, or enterprise applications, you probably know how NVIDIA Graphics Processing Unit (GPU) cards can provide a significant performance boost. GPU-accelerated computing offers this boost by offloading compute-intensive portions of the application to the GPU, while the remainder of the code still runs on the main CPU. From a user's perspective, applications simply run significantly faster.
NVIDIA Tesla and GRID GPUs significant performance boost by offloading
compute-intensive portions of the application off the main system CPU
With this month’s announcement from IBM NeXtScale System, NVIDIA brings the advantages of High Performance Computing to the virtualized datacenter.
As we described in a recent blog post, IBM refreshed the NeXtScale System product line and a noticeable addition was the availability of the PCIe Native Expansion tray. Along with the PCIe Tray comes support for 5 high-performance NVIDIA GPUs, a mix of Tesla and GRID cards. The PCIe tray enables the use of two GPUs per compute node. In a 6U chassis, this provides the ability to support up to 12 GPUs.
The NeXtScale n1200 chassis can house up to 6 Compute Nodes, each with a two-slot
PCIe Native Expansion Tray attached, allowing up to 12 GPU cards in just 6U of rack space.
With the NeXtScale PCIe Tray comes support for 5 NVIDIA GPUs that optimize processing power:
- NVIDIA GRID K1
- NVIDIA GRID K2
- NVIDIA Tesla K10
- NVIDIA Tesla K20X
- NVIDIA Tesla K40
Two of these double-wide full-length cards are supported in each PCIe Tray which means a significant amount of processing power fits in an equivalent of just 1U of rack space.
Two Tesla GPUs, on a single node, installed in an IBM NeXtScale nx360 M4 Server
NVIDIA Tesla GPUs debuted in High Performance Computing in 2008 with an initial listing on the prestigious Top 500 Supercomputing Sites list and raced to #1 just two years later on the 2010 list. Tesla continues to increase in importance in computational science with more sites being added every subsequent year.
This same processing power, that has revved up supercomputing, is now available to virtualized enterprise and cloud environments through the use of NVIDIA GRID K1 and K2 boards with the NeXtScale System. GPU accelerated data centers are gaining traction for enterprise workloads due to the numerous performance benefits. One of those benefits is for VDI (Virtual Desktop Infrastructure). VDI is known to increase data replication, mobility, security, and so on. In the past, however, some users (like designers and engineers who demand processing power) found their high end requirements too complex and costly to provide within a VDI environment. Now, there is reason to get excited!
NVIDIA GRID K1 and K2 boards with the IBM NeXtScale System offers a compelling solution. The K1/K2 boards are specifically designed to enable rich graphics in virtualized environments. So, whether the requirements call for a true virtual desktop environment or an application virtualization environment, NVIDIA GRID with the NeXtScale System makes for a speedy resolution and a winning combination for enterprises and all their users.
But wait – there’s more! We are going to cover other benefits and what is in it for you when using the NeXtScale System and NVIDIA options for enabling a GPU accelerated data center:
- GPU Accelerated Applications for Graphics (for Engineering/Design, Geospatial, Oil & Gas)
- GPU Accelerated Compute for Analytics (for Financial, Rendering, and so on)
- GPU Accelerated Big Data
- GPU Accelerated Virtualization (IBM SmartCloud, Hybrid Clouds, and Virtual Desktops)
We will be posting a blog a week for the next 5 weeks – sorry you have to wait…but hey, time flies pretty fast when you’re NVIDIA GPU accelerated!
For specific product and part information for purchasing the products mentioned in this blog, see the IBM NeXtScale nx360 M4 Product Guide from IBM Redbooks.
Michael O’Neill is an established leader for NVIDIA and provides specialized technical guidance to customers on NVIDIA GRID and Tesla GPUs in virtualized environments. He works closely with IBM to develop innovative solutions for graphical and compute heavy workloads.
Brandon Harrell is an IBM Technical Solutions Leader. He specializes in x86 Desktop and Application Virtualization as part of the Advanced Technical Sales team. Follow Brandon on Twitter at @BrandonEHarrell.
Deana Coble is an IBM Redbooks Technical Writer. She has worked over 20 years in the field of technology and collaborates with IBM subject matter experts to create IBM Redbooks publications on a variety of systems. Follow Deana on Twitter at @DeanaCoble.