Convey Cranks Up Performance with Latest FPGA-Accelerated Computer

By Michael Feldman

April 24, 2012

Convey Computer has launched its newest x86-FPGA “hybrid-core” server. Dubbed HC-2, it represents the first major upgrade of the system since the company introduced the HC-1 product back in 2008. The new offering promises much better performance, but with a similar price range as the original system.

The new HC-2 keeps the basic architecture Convey established with the HC-1 of an x86 host glued to a custom-wrapped FPGA board that acts as a coprocessor. Like its predecessor, the HC-2 is not a flops machine. The FPGA-based coprocessor is designed to accelerate data-intensive workloads like genome sequence alignment and other types of data mining workloads. The coprocessor memory subsystem is built like that of a vector supercomputer, able to deliver reads and write at much higher bandwidth than that of a standard commodity server.

Although there are some important hardware tweaks with the HC-2, binary compatibility has been retained so that software developed for the original HC-1 platform can execute as is on the new platform. As before, the coprocessor subsystem is tapped into via Convey-supplied libraries and tools, which allows the user to build (or reuse) customized application-specific instruction for accelerating codes. And since the platform is essentially an x86 Linux server, standard HPC software, like MPI or a workload manager, runs on the platform transparently.

Hardware-wise, most of the HC-2 upgrades are on the host side, where Convey has swapped out the older Intel Xeon X5400 (“Harpertown) used on the HC-1 with CPUs of more recent vintage, namely the Xeon X5600 (“Westmere”) and Xeon E5-2600 (“Sandy Bridge”) processors. The Westmere-based hardware will employ the 6-core X5690 CPU running at 2.93 GHz, while the Sandy Bridge-based servers will come in two basic flavors, 4-core and 8-core, but offered at clock speeds ranging from 2.4 to 3.3 GHz. The coprocessor will use the same Xilinx FPGAs present in the HC-1: Virtex-5 for the baseline server and Virtex-6 for the more performant “ex” variant.

Memory capacity is quite a bit higher on the new servers too. The Westmere-based system can reach up to 192 GB, which is a nice bump from the 128 GB limit on the HC-1 gear. Thanks to the new Sandy Bridge design, though, servers equipped with those CPUs can be outfitted with as much as 768 GB of memory. The memory capacity on the FPGA coprocessor board memory will top out at 64 GB, the same as it was on the HC-1.

I/O has been kicked up as well. Instead of just a single Gen 2 PCIe port, a SATA interface, 1 to 3 hot swap SATA drives, and an IDE optical drive, the newer servers sport up to 8 SATA drives, 2 Intel I/O modules slots, and 5 PCIe Gen 3 ports (although not all these options are available in all configurations).  Since these are essentially data-crunching machines, the extra I/O support should be much appreciated.

The biggest design change was a result of the Westmere/Sandy Bridge CPU upgrade. This forced the Convey engineers to make a decision about the host-coprocessor interconnect, which on the Harpertown-equipped HC-1 was based on Intel’s Front-Side Bus (FSB). Since Harpertown was the last FSB-based Xeon chip, with all subsequent design using the new QuickPath Interconnect (QPI), Convey had to go either stay native and build a QPI-based system or use PCIe. They opted for the latter, which, given that PCIe is an industry standard, gave them the most flexibility going forward.

Moving to PCIe also freed up a CPU socket that was taken up by the coprocessor interface under the FSB scheme. Thus the HC-2 servers can all be dual-socket servers instead of the single-socket systems of the HC-1. That’s a fortuitous development since, according to Convey CEO Bruce Toal, there was somewhat of an imbalance in the original design.

In particular, with only a single dual- or quad-core Xeon in the HC-1, in some cases there wasn’t enough x86 performance to keep up with a fully outfitted four-FPGA coprocessor, whilst simultaneously handling I/O. Now with the extra CPU on the freed socket (not to mention the more numerous cores of Westmere and Sandy Bridge), the host has a lot more cycles available to it to feed the coprocessor and I/O devices.

The company is claiming a 2- to 3-fold improvement in application performance for the HC-2 compared to the original HC-1 and a 10- to 50-fold performance boost compared to a vanilla x86 server. For example, a 12-core x86 server running a genome sequence application using the Burrows-Wheeler Aligner (BWA) algorithm can align a little over 7 thousand sequences per second. Even the older generation HC-1ex managed 27 thousand alignments per second, but with the new HC-2ex gear, that goes up to over 66 thousand alignments. A similar performance boost has been demonstrated with BLAST, the Basic Local Alignment Search Tool.

As impressive as that is, the energy efficiency of the new system is equally noteworthy. While the highest performing Sandy Bridge CPU-based servers will require an 1800 watt power supply, the majority of the HC-2 configurations will fit into a 1570 watt box. That’s only a marginally higher than the 1520 watt power supply on the original top-of-the line HC-1ex, but since the new version is promising to be at least twice as fast, performance-per-watt has improved substantially.

That plays into one of Convey’s main pitches for their hybrid-core solution: reducing the total cost of ownership (TCO). And if these HC-2 performance and power numbers hold up, that story just got better. The company says for a typical bioinformatics application you only need about 1/15 the number of HC-2ex nodes compared running the same job on a standard x86 (dual-socket Westmere) cluster. Although each Convey node is going to draw more power than a commodity server, it still works out to 83 percent less power overall for a given amount of application throughput. Likewise, datacenter floor space is just a fraction of what would be needed for an all-x86 setup. Overall, Convey estimates a three-year TCO savings of 75 percent.

Toal says it’s a bit easier to convince commercial users of the TCO advantages, since they are more sensitive to datacenter operations costs than their academic counterparts. In research environments, the electricity bill is typically paid by the institution, so these users tend to be less concerned about energy efficiency. “You have to find the guy that owns the power budget,” says Toal.

Nevertheless, a good chunk of Convey’s business is done with non-commercial customers across all their four principle application domains: bioinformatics, government, telco, and research. Not surprisingly for an FPGA-based solution, bioinformatics is the biggest vertical, which, according to Toal, represents 36 percent of their shipments. Government, which includes a lot of classified apps (think national security-type data mining) and defense work, is second at 21 percent. The research space proper, which is centered on deployments at DOE national labs like Oak Ridge and Lawrence Berkeley National Labs, is 18 percent of the business. Lastly is the telco market, which represents 17 percent of Convey’s shipments.

System pricing on the new HC-2 offerings are in line with the HC-1 products. A Westmere-based HC-2 server with a minimal configuration begins at $40,000; the corresponding Sandy Bridge box starts at $41K. Adding faster processors and more memory will kick up the price accordingly. The Westmere-based servers are already available and have been shipping for some time, including one deployment at Jackson Laboratory (JAX), in Maine. The Sandy Bridge gear won’t be available until July.

Related Articles

Convey Bends to Inflection Point

Convey Debuts Second-Generation Hybrid-Core Platform

Startup Provides a New Twist on Reconfigurable Supercomputing

Subscribe to HPCwire's Weekly Update!

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

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire