March 17, 2009
Supercomputer maker Cray has launched the XT5m, a midrange supercomputer based on its top-of-the-line XT5. The new offering inherits the basic architectural elements of its petascale-capable parent along with the associated software compatibility.
Except for the system interconnect, which we'll get to in a moment, the XT5m is almost identical to the XT5. Both products are based on dual-socket blade nodes and are currently shipping with the latest quad-core AMD "Shanghai" Opteron processors. When AMD releases the six-core "Istanbul" Opterons later this year, all systems will be field upgradeable to the new chips. The cabinets themselves are also designed for compatibility with future blades using AMD's G34 socket technology, so a system installed today could be continuously upgraded through 2012.
The rest of the components, including the software libraries and tools, the cooling system (air or liquid cooled ECOPhlex), and cabinet enclosures, are also identical across both products. An XT5m can be upgraded to an XT5 for an additional 15-20 percent, which is in line with the cost differential between the two products. In fact, one of Cray's claims for the XT5m is that it is the only midrange supercomputer that maintains upward compatibility with a petascale architecture (although Sun Microsystems would probably point to their own Constellation system as an example of another architecture than can span the entire range of supercomputing).
Where the XT5 and its lesser sibling part company is scalability. The XT5m tops out at six cabinets, which translates to a maximum of 1,120 sockets. Even with that upper limit, a six-cabinet XT5m equipped with the upcoming Istanbul chips will offer peak performance of around 60 teraflops and a maximum memory capacity of 17.2 TB.
The primary architectural difference between the XT5 and the XT5m is the implementation of the SeaStar system network. The XT5 uses a 3D Torus interconnect, while the XT5m uses a 2D Torus. According the Cray, the less performant 2D implementation is fine for this midrange system, since internal bandwidth demand is capped by the size of the system. The lesser Torus is the major reason for the 15-20 percent cost savings, when compared to an equivalent XT5.
In the low six-figure price range, the new midrange super is positioned to go head-to-head with x86/InfiniBand-style commodity clusters from the likes of IBM, HP, Dell, Sun, SGI, and others who sell into the sub-petascale HPC segment. According to Barry Bolding, Cray's VP of Scalable Systems, the XT5m competes very well on price-performance with every player in that segment. "On a per core or per socket basis, if you price any one of our competitors, we will be right in line with their pricing," he claims.
The XT5m is aimed at the lower end of IDC's supercomputer category, which starts at $500,000 and climbs to nine figures for the most elite systems. Most of the XT5 systems are in the $5-10 million range on up, while Cray expects to sell most of its XT5m systems for $1-3 million. This gives the supercomputer maker access to market segments it hasn't played in for awhile, namely lower-tier customers adjacent to its traditional government and higher education markets as well as commercial users in aerospace, automotive, manufacturing and pharmaceuticals.
Bolding says the company is especially looking to universities who are interested in high-end HPC, but who don't have the budget of a supercomputing center. Smaller government agencies and smaller groups within larger agencies may also find systems in this price range attractive, especially as the US federal stimulus money starts to be dispersed.
The first XT5m customer happens to be in Germany at the High Performance Computing Center Stuttgart (HLRS). Scientists at HLRS intend to use their new two-cabinet system for a variety of research programs, including applied research in partnership with the German automotive industry. One of the applications that will be running on the machine is PAM-CRASH, an automobile crash simulation code that ESI recently ported to the XT5/XT5m platform and was able to scale to 1024 cores. Cray is hoping the presence of a midrange supercomputing offering will encourage more ISVs to port their codes onto the XT platform.
But selling directly into the commercial space is bound to be especially tough right now. Cray came up with the midrange supercomputer concept last year before the economy headed south. At that point they thought the XT5m could get significant traction with automotive and aerospace customers. Now with many of the big manufacturers struggling, Cray's expectations for commercial XT5m sales have been scaled back.
But Bolding says the company is willing to be patient. According to him, because they were able to re-use a lot of the existing technology in the XT5, the amount of R&D they put into the XT5m was minimal. Testing the 2D Torus network and validating all the libraries, tools, and applications for the platform encompassed most of the effort to get the XT5m productized. "So if our volumes are relatively modest, that's not going to particularly effect us," explains Bolding. In any case, Cray is not looking at the XT5m as a big revenue driver in 2009.
The longer term goal of the new product -- expanding their reach in their traditional government and education customers and gaining a bigger footprint in commercial HPC -- is in line with Cray's overall strategy of relying less on their elite XT and XMT supers to carry the company. Last year's introduction of the CX-1 deskside systems for workgroup HPC was the first part of that strategy. Using the XT5m to cover the lower end of the supercomputing segment was their next step.
But it's not likely that Cray will become a soup-to-nuts HPC system vendor anytime soon. Specifically, the company has made no move to venture into the $250,000 to $500,000 segment, corresponding to IDC's divisional HPC category. Cray doesn't have a particular technology that it can leverage to compete against the rack-optimized servers that make up that highly cost-sensitive segment. As Bolding notes, "That's a vicious marketplace."
May 16, 2013 |
When it comes to cloud, long distances mean unacceptably high latencies. Researchers from the University of Bonn in Germany examined those latency issues of doing CFD modeling in the cloud by utilizing a common CFD and its utilization in HPC instance types including both CPU and GPU cores of Amazon EC2.
Read more...
May 15, 2013 |
Supercomputers at the Department of Energy’s National Energy Research Scientific Computing Center (NERSC) have worked on important computational problems such as collapse of the atomic state, the optimization of chemical catalysts, and now modeling popping bubbles.
Read more...
May 10, 2013 |
Program provides cash awards up to $10,000 for the best open-source end-user applications deployed on 100G network.
Read more...
May 09, 2013 |
The Japanese government has revealed its plans to best its previous K Computer efforts with what they hope will be the first exascale system...
Read more...
May 08, 2013 |
For engineers looking to leverage high-performance computing, the accessibility of a cloud-based approach is a powerful draw, but there are costs that may not be readily apparent.
Read more...
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/15/2013 | Bull | “50% of HPC users say their largest jobs scale to 120 cores or less.” How about yours? Are your codes ready to take advantage of today’s and tomorrow’s ultra-parallel HPC systems? Download this White Paper by Analysts Intersect360 Research to see what Bull and Intel’s Center for Excellence in Parallel Programming can do for your codes.
In this demonstration of SGI DMF ZeroWatt disk solution, Dr. Eng Lim Goh, SGI CTO, discusses a function of SGI DMF software to reduce costs and power consumption in an exascale (Big Data) storage datacenter.
The Cray CS300-AC cluster supercomputer offers energy efficient, air-cooled design based on modular, industry-standard platforms featuring the latest processor and network technologies and a wide range of datacenter cooling requirements.