GPU Supercomputing Arrives on Amazon’s Cloud

By Nicole Hemsoth

November 15, 2010

This week marks a meaningful stage in the ongoing GPU coming-out party in the realm of high-performance computing–both in the cloud and beyond.

While there have been steady conversations about the power of GPU computing for certain HPC applications for some time already, the word “GPU” on the lips of thousands more this week following the official announcement of China’s number one-ranked GPU-powered supercomputer the Tianhe-1a, which runs on 14,336 Xeon X5670 processors and 7,168 NVIDIA Tesla M2050 GPGPUs.

Another major announcement, this time from cloud infrastructure giant Amazon Web Services, brings the mighty GPU to the forefront of conversations again with its unveiling of Cluster GPU Instances, a spin-off of the still fresh Cluster Compute Instances.

Those whose work resides in visualization, rendering, oil and gas, financial and climate modeling and a wide range of other fields are being granted yet another alternative to investing in their own GPU clusters with this addition to Amazon’s short list of HPC-oriented offering. While there are other providers with GPU clusters for rent, including Penguin Computing, Peer1 Hosting and a handful of others who do have some of the support framework (and oftentimes performance benchmarking comparisons) in place, the concept of GPU supercomputing on an affordable public cloud resource is big news in the HPC “cloudosphere” indeed.

Under the Cluster GPU Hood

Outside of the addition of a pair of NVIDIA Tesla M2050 Fermi GPUs, the Cluster GPU instances don’t look too much different than Cluster Compute Instances, which were announced late this summer. This instance type hosts a pair of quad-core Intel Nehalem X5570 processors and 10 Gbps Ethernet as well as 22 GB of RAM and 16900 GB for storehousing and is designed for (and has been quite extensively tested with) a number of high-performance computing workloads.

As the GPU Cluster Instance team stated today on its blog, “Each of the Tesla M2050s contains 448 cores and 3GB of ECC RAM, designed to deliver up to 515 gigaflops of double-precision performance when pushed to the limit. Since each instance contains a pair of these processors, you can get slightly more than a trillion FLOPS per Cluster GPU instance.”

When you tie all of this together with 10Gbps Ethernet, there is an  opportunity for highly data parallel HPC and rendering capabilities that are definitely worth paying attention to since they’re arriving at $2.10 per hour standard. Just as with the Cluster Compute Instances, customers are given a base outlay of resources but have to manually request more from the AWS team.

This new instance type comes with some added complexity, at least compared to its predecessor on the HPC end, the Cluster Compute Instances where porting code was not such a tricky issue. This instance type is aimed at those who understand (or have a reliable group to help with) taking advantage of those Fermis.

The biggest hurdle for wider GPU computing adoption, whether in the cloud or not, lies simply in application porting. Since specialized code is required, namely understanding the CUDA architecture, getting started with these instance types early on will be an exercise only for those likely familiar with EC2 and also familiar with the complex coding involved with GPU computing. However, as interest in GPU computing mounts in the coming years, it is likely that some of these code-related challenges will be worked out to ensure wider access.

According to NVIDIA, “Hundreds of applications have already been ported to the NVIDIA CUDA massively parallel architecture upon which all its current GPUs are based…Some of the applications ready to be used on Amazon Cluster GPU Instances include mental images’ RealityServer and Mathworks’ MATLAB.

NVIDIA has granted a significant amount of time to reaching out to potential customers who could make use of GPGPUs, thus such a partnership with Amazon serves them very well as it opens access to their architecture via greater affordability. It’s not hard to see how this is a win-win for both AWS and NVIDIA respectively.

While the number of applications ported will grow in coming years, especially as GPU computing is being stretched out to meet a number of new users who were barred from entry financially, it’s difficult to tell how  much of a boost this will give Amazon. Their CCI offering has been overall well-received, despite some of its limitations for security and latency-aware applications, and one can guess this too will take hold among a small minority and grow in numbers as the porting issue becomes less problematic.

The Bigger GPU Cloud Picture

Not long ago, developers of advanced scientific applications found a way to use GPUs to handle a number of general purpose problems, hence leading to the growing General Purpose on Graphics Processing Units (GPGPU) movement. This is a quickly evolving space as more developers seek to extend applications beyond CPU processing power, and clearly someone at Amazon saw a value-add in granting customers access to GPUs in the cloud. This week during SC10 we hope to talk to some potential users about how they would (or wouldn’t) use such a resource and what they would perceive as the biggest barriers or benefits.

There have been several hoping to get Amazon’s attention to request GPU computing capabilities over the last couple of years, but until the official announcement, this still seemed on the horizon. Some projected that Amazon would wait to make an investment in GPU until it had sufficiently monitored and approved of Cluster Compute Instance success and furthermore, until they saw innovations in ease of use of GPUs from a code standpoint.

In addition to assuring us during an interview that Cluster Compute Instance had, indeed, been very well-received, Peter De Santis, General Manager of Amazon’s EC2 noted in the company’s release that with this new offering, “we are increasing the options available to our HPC customers by allowing them to choose between using high-performance clusters with high-performance CPUs or taking advantage of the unique processing abilities of GPU processors for applications that can benefit from the massive parallel processing power they provide.”

Mental Images, a subsidiary of NVIDIA that provides rendering and 3D web services component software targeted at CAE, visualization and entertainment is a perfect candidate to benefit from an instance type such as the new Cluster GPU provisioning. As Rolf Herken, CEO and CTO of mental images noted, “The availability of NVIDIA Tesla GPUs in the AWS cloud in the form of Amazon Cluster GPU Instances running the RealityServer platform with the iray renderer will provide architects, product designers, engineers, scientists and others with extraordinarily powerful tools that they can remotely access on mobile devices, PCs and other devices.” According to internal tests he has run with his team, it has been shown that they can reach “more than 90 percent scaling efficiency on clusters of up to 128 GPUs each.”

Amazon’s offering of GPU power to support a host of high-performance computing applications, particularly in the oil and gas, financial services and rendering arenas comes as a bit of surprise as many anticipated they wouldn’t see such an offering until there was a more swelling appeal to AWS on the part of seasoned GPU computing users who were also familiar with EC2. While it was a more of a surprise that it happened so soon, what better time to push out an announcement than during the biggest week for HPC professionals around the world…

Earlier this month we announced the winners of our 2010 Editors’ Choice Awards, of which Amazon had the distinction of being the Cloud Platform Innovator. This choice was due, in part, to their announcement of the HPC-geared Cluster Compute Instance type (CCI). Our argument in favor of Amazon’s position as an ever-evolving innovator in cloud computing now gains further credence as new doors have opened for those who could make great use of GPU computing but were barred from entry due to high up-front investment.

Cluster Compute Instances are being used, according to Tera Randall from Amazon, for a number of projects, including a few related to machine learning for infectious disease research, next-gen sequencing of genetic data, energy trading and financial modeling and other design and engineering-related tasks. Randall noted that there were “a large number of users running a variety of CFD applications.

With this adoption of Cluster Compute Instances for applications that might be a good fit for the processing power addition of GPU computing, one has to wonder if these existing customers will be up for the task of retooling their code to capture the GPU benefit or if they’ll settle for standard CCIs, which is $1.60 per instance hour versus $2.10 per instance hour for Cluster GPU.
 

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Data Vortex Users Contemplate the Future of Supercomputing

October 19, 2017

Last month (Sept. 11-12), HPC networking company Data Vortex held its inaugural users group at Pacific Northwest National Laboratory (PNNL) bringing together about 30 participants from industry, government and academia t Read more…

By Tiffany Trader

AI Self-Training Goes Forward at Google DeepMind

October 19, 2017

DeepMind, Google’s AI research organization, announced today in a blog that AlphaGo Zero, the latest evolution of AlphaGo (the first computer program to defeat a Go world champion) trained itself within three days to play Go at a superhuman level (i.e., better than any human) – and to beat the old version of AlphaGo – without leveraging human expertise, data or training. Read more…

By Doug Black

Researchers Scale COSMO Climate Code to 4888 GPUs on Piz Daint

October 17, 2017

Effective global climate simulation, sorely needed to anticipate and cope with global warming, has long been computationally challenging. Two of the major obstacles are the needed resolution and prolonged time to compute Read more…

By John Russell

HPE Extreme Performance Solutions

Transforming Genomic Analytics with HPC-Accelerated Insights

Advancements in the field of genomics are revolutionizing our understanding of human biology, rapidly accelerating the discovery and treatment of genetic diseases, and dramatically improving human health. Read more…

Student Cluster Competition Coverage New Home

October 16, 2017

Hello computer sports fans! This is the first of many (many!) articles covering the world-wide phenomenon of Student Cluster Competitions. Finally, the Student Cluster Competition coverage has come to its natural home: H Read more…

By Dan Olds

Data Vortex Users Contemplate the Future of Supercomputing

October 19, 2017

Last month (Sept. 11-12), HPC networking company Data Vortex held its inaugural users group at Pacific Northwest National Laboratory (PNNL) bringing together ab Read more…

By Tiffany Trader

AI Self-Training Goes Forward at Google DeepMind

October 19, 2017

DeepMind, Google’s AI research organization, announced today in a blog that AlphaGo Zero, the latest evolution of AlphaGo (the first computer program to defeat a Go world champion) trained itself within three days to play Go at a superhuman level (i.e., better than any human) – and to beat the old version of AlphaGo – without leveraging human expertise, data or training. Read more…

By Doug Black

Student Cluster Competition Coverage New Home

October 16, 2017

Hello computer sports fans! This is the first of many (many!) articles covering the world-wide phenomenon of Student Cluster Competitions. Finally, the Student Read more…

By Dan Olds

Intel Delivers 17-Qubit Quantum Chip to European Research Partner

October 10, 2017

On Tuesday, Intel delivered a 17-qubit superconducting test chip to research partner QuTech, the quantum research institute of Delft University of Technology (TU Delft) in the Netherlands. The announcement marks a major milestone in the 10-year, $50-million collaborative relationship with TU Delft and TNO, the Dutch Organization for Applied Research, to accelerate advancements in quantum computing. Read more…

By Tiffany Trader

Fujitsu Tapped to Build 37-Petaflops ABCI System for AIST

October 10, 2017

Fujitsu announced today it will build the long-planned AI Bridging Cloud Infrastructure (ABCI) which is set to become the fastest supercomputer system in Japan Read more…

By John Russell

HPC Chips – A Veritable Smorgasbord?

October 10, 2017

For the first time since AMD's ill-fated launch of Bulldozer the answer to the question, 'Which CPU will be in my next HPC system?' doesn't have to be 'Whichever variety of Intel Xeon E5 they are selling when we procure'. Read more…

By Dairsie Latimer

Delays, Smoke, Records & Markets – A Candid Conversation with Cray CEO Peter Ungaro

October 5, 2017

Earlier this month, Tom Tabor, publisher of HPCwire and I had a very personal conversation with Cray CEO Peter Ungaro. Cray has been on something of a Cinderell Read more…

By Tiffany Trader & Tom Tabor

Intel Debuts Programmable Acceleration Card

October 5, 2017

With a view toward supporting complex, data-intensive applications, such as AI inference, video streaming analytics, database acceleration and genomics, Intel i Read more…

By Doug Black

Reinders: “AVX-512 May Be a Hidden Gem” in Intel Xeon Scalable Processors

June 29, 2017

Imagine if we could use vector processing on something other than just floating point problems.  Today, GPUs and CPUs work tirelessly to accelerate algorithms Read more…

By James Reinders

NERSC Scales Scientific Deep Learning to 15 Petaflops

August 28, 2017

A collaborative effort between Intel, NERSC and Stanford has delivered the first 15-petaflops deep learning software running on HPC platforms and is, according Read more…

By Rob Farber

Oracle Layoffs Reportedly Hit SPARC and Solaris Hard

September 7, 2017

Oracle’s latest layoffs have many wondering if this is the end of the line for the SPARC processor and Solaris OS development. As reported by multiple sources Read more…

By John Russell

US Coalesces Plans for First Exascale Supercomputer: Aurora in 2021

September 27, 2017

At the Advanced Scientific Computing Advisory Committee (ASCAC) meeting, in Arlington, Va., yesterday (Sept. 26), it was revealed that the "Aurora" supercompute Read more…

By Tiffany Trader

How ‘Knights Mill’ Gets Its Deep Learning Flops

June 22, 2017

Intel, the subject of much speculation regarding the delayed, rewritten or potentially canceled “Aurora” contract (the Argonne Lab part of the CORAL “ Read more…

By Tiffany Trader

Google Releases Deeplearn.js to Further Democratize Machine Learning

August 17, 2017

Spreading the use of machine learning tools is one of the goals of Google’s PAIR (People + AI Research) initiative, which was introduced in early July. Last w Read more…

By John Russell

Nvidia Responds to Google TPU Benchmarking

April 10, 2017

Nvidia highlights strengths of its newest GPU silicon in response to Google's report on the performance and energy advantages of its custom tensor processor. Read more…

By Tiffany Trader

GlobalFoundries Puts Wind in AMD’s Sails with 12nm FinFET

September 24, 2017

From its annual tech conference last week (Sept. 20), where GlobalFoundries welcomed more than 600 semiconductor professionals (reaching the Santa Clara venue Read more…

By Tiffany Trader

Leading Solution Providers

Graphcore Readies Launch of 16nm Colossus-IPU Chip

July 20, 2017

A second $30 million funding round for U.K. AI chip developer Graphcore sets up the company to go to market with its “intelligent processing unit” (IPU) in Read more…

By Tiffany Trader

Amazon Debuts New AMD-based GPU Instances for Graphics Acceleration

September 12, 2017

Last week Amazon Web Services (AWS) streaming service, AppStream 2.0, introduced a new GPU instance called Graphics Design intended to accelerate graphics. The Read more…

By John Russell

EU Funds 20 Million Euro ARM+FPGA Exascale Project

September 7, 2017

At the Barcelona Supercomputer Centre on Wednesday (Sept. 6), 16 partners gathered to launch the EuroEXA project, which invests €20 million over three-and-a-half years into exascale-focused research and development. Led by the Horizon 2020 program, EuroEXA picks up the banner of a triad of partner projects — ExaNeSt, EcoScale and ExaNoDe — building on their work... Read more…

By Tiffany Trader

Delays, Smoke, Records & Markets – A Candid Conversation with Cray CEO Peter Ungaro

October 5, 2017

Earlier this month, Tom Tabor, publisher of HPCwire and I had a very personal conversation with Cray CEO Peter Ungaro. Cray has been on something of a Cinderell Read more…

By Tiffany Trader & Tom Tabor

Cray Moves to Acquire the Seagate ClusterStor Line

July 28, 2017

This week Cray announced that it is picking up Seagate's ClusterStor HPC storage array business for an undisclosed sum. "In short we're effectively transitioning the bulk of the ClusterStor product line to Cray," said CEO Peter Ungaro. Read more…

By Tiffany Trader

Intel Launches Software Tools to Ease FPGA Programming

September 5, 2017

Field Programmable Gate Arrays (FPGAs) have a reputation for being difficult to program, requiring expertise in specialty languages, like Verilog or VHDL. Easin Read more…

By Tiffany Trader

IBM Advances Web-based Quantum Programming

September 5, 2017

IBM Research is pairing its Jupyter-based Data Science Experience notebook environment with its cloud-based quantum computer, IBM Q, in hopes of encouraging a new class of entrepreneurial user to solve intractable problems that even exceed the capabilities of the best AI systems. Read more…

By Alex Woodie

HPC Chips – A Veritable Smorgasbord?

October 10, 2017

For the first time since AMD's ill-fated launch of Bulldozer the answer to the question, 'Which CPU will be in my next HPC system?' doesn't have to be 'Whichever variety of Intel Xeon E5 they are selling when we procure'. Read more…

By Dairsie Latimer

  • arrow
  • Click Here for More Headlines
  • arrow
Share This