May 26, 2010
The 35th running of the TOP500 sweepstakes will take place at the International Supercomputing Conference in Germany next week, and while I haven't heard about any challengers to the reigning champ (Oak Ridge National Lab's "Jaguar" supercomputer), I expect to see a few petaflop or near-petaflop machines fill in the top 10.
Overall though, I anticipate that this will be a time of consolidation for elite supercomputers. We're sort of in a lull in major technology deployments, at least on the processor front. The spring crop of x86 server silicon from Intel and AMD is just making its way into actual systems and the next-generation NVIDIA Fermi GPU is about to do so. So wake me up in November.
That said, there are a couple of systems slated to move into the top ranks of the list, assuming, of course, that their masters turned in their Linpack homework on time. For example, there's a new GPU-accelerated super in China that is using NVIDIA gear and Mellanox' GPUDirect technology to power a petaflop system. The machine, known as the Mole-8.5, is deployed at the Institute of Process Engineering (IPE) at the Chinese Academy of Sciences and is being billed as "the first petaflop GPGPU supercomputer in China." The other system is NASA's Pleiades, which recently received of 32-rack upgrade of SGI's latest ICE 8400 hardware, boosting its aggregate performance to 973 peak teraflops. And I wouldn't be surprised if some other top systems lurking out there have been upgraded surreptitiously and are set to jump a few spots in the list.
By the way, if you're wondering where your own super -- real or imagined -- might fit in Linpack-wise, Dell has come up with a nifty little HPC performance calculator that computes theoretical (peak) and Linpack results for a given machine. The computation is based on six inputs: operations per clock, clock speed, cores per socket, sockets per node, number of nodes and Linpack efficiency. The first five inputs are straightforward enough, but the Linpack efficiency is a bit trickier. According to Dell's Dr. Jeff Layton, who developed the calculator, the efficiency estimation is usually based on a variety of factors, including the system interconnect, the number of nodes, and even the memory capacity per node.
Besides furnishing the FLOPS numbers, the Dell calculator also tells you where your machine fits into the TOP500 rankings. If the specified system is 23 Linpack teraflops or lower (that is, below the 500th spot on the latest list), the calculator spits out:
"Sorry but xxx GFLOPS is too low to rank in the current Top500 list (Nov. 2009). Get a life."
It doesn't actually say "Get a life." That's just implied.
If you want to give the performance calculator a whirl, go to Dell's Web page here and have at it.
Posted by Michael Feldman - May 26, 2010 @ 9:47 AM, Pacific Daylight Time
Michael Feldman is the editor of HPCwire.
No Recent Blog Comments
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?