PSC Powers up World's Largest Shared Memory Systems

Print Friendly, PDF & Email

Thanks to a $2.8M award from the National Science Foundation, the Pittsburgh Supercomputing Center has deployed Blacklight, their newest supercomputer with 512 eight-core Intel Xeon processors with 32 terabytes of memory. Accessible via the TeraGrid, the SGI Altix UV1000 system allows all processors to share a single view of the memory space, which should make it much easier to program than clusters that employ message passing.

Blacklight is actually partitioned into two separate 16-terabyte shared memory systems, which are reportedly the two largest coherent shared-memory systems in the world. Both “systems” will be used for molecular biology simulations, extreme-scale performance engineering, chemistry, fluid dynamics, and genomics.

Researchers are extremely excited about what they will be able to achieve with 16 terabytes of coherent shared memory, or even accessing all 32 terabytes, as can be accomplished with a bit more programming,” says Nick Nystrom, PSC’s director of strategic applications. “Also revolutionary is that Blacklight offers unparalleled ease of use for rapidly testing new ideas and for expressing complex algorithms, dramatically increasing users’ productivity.”

For more details on how the SGI Altix UV works to provide a single system image, check out my interview with Eng Lim Goh, SGI’s Chief Technology Officer.

Comments

  1. Xray Penguin says

    Hmmm… this article is a little misleading.
    The naive reader might think that two SMP machines of 2048 cores each (with 16TB RAM each) are a total of $2.8M.
    Let’s do the math for two main components: processors and RAM:
    Assuming 8GB DIMMs at a fair price of $250 a unit, a 16TB (16.384) of RAM would cost $512,000. So the memory for two machines in a little over $1M.
    Each machine has 1048 cores. Assuming 8-cores/processor, and not taking the most expensive part number, we can assume $2,500 per processor. That’s $640,000 for processors, or about $1.3M for two machines.
    So, memory and processors only, for the configurations mentioned, are $2.3M; this leaves on $0.5M for the rest – very unlikely to fit the bill.

    Can insideHPC find how much money was paid (tax payer’s money… this should be public info), and educate the readers?

  2. Xray Penguin says

    that’s a typo… 2048 cores, not 1048 🙂

  3. The award from the NSF was $2.8M

    That does not mean that SGI received that amount from PSC, which has other sources of funding.