Posted: June 13, 2016, 11:35pm
When I came to USC, I had no idea that I’d be able to work with a supercomputer. Actually, I had no idea really what a supercomputer was. I would say that it sounds cooler than it actually is, but the truth is that it’s super cool. Pun intended.
USC has a very heterogeneous computing cluster. And that’s exactly what a supercomputer is: a cluster of relatively normal computers strung together to work for a common simulation. On their website, the USC Center for High Performance Computing (HPC) lists the following resources:
- 264 Hewlett-Packard SL250, dual Sandy Bridge Xeon 8-core 2.6 gigahertz, dual-processor, dual NVIDIA K20 GPUs containing 2,496 cores each, with 64 gigabytes of memory
• 208 Hewlett-Packard SL230, dual Sandy Bridge Xeon 8-core 2.6 gigahertz dual-processor CPUs with 64 gigabytes of memory
• 18 Lenovo nx360m5 2.6 gigahertz dual-processor, dual NVIDIA K40 GPUs containing 2,880 cores each with 64 gigabytes of memory
• 5 Lenovo nx360m5 2.6 gigahertz dual-processor, dual NVIDIA K80 GPUs containing 2 x 2,496 cores each with 64 gigabytes of memory
So, for example, if your laptop was made by HP and it has a dual-core processor with some NVIDIA graphics processor, imagine taking several hundred of those and putting them in a single building and letting power-hungry grad students like me gain access to them through the internet. All told, the USC HPC has about 10,000 cores and they give access to researchers across campus in a variety of disciplines.
Our group uses the HPC for our cancer models. We build predictive models of breast and lung cancer, and use the supercomputer to test the efficacy of various therapeutic strategies on tumor growth and metastasis. Pretty cool! In fact, I’m usually running simulations, even overnight. So if my advisor ever asks if I’m working, I can confidently reply yes. Well, sort of, anyway!
Published on July 25th, 2017
Last updated on August 10th, 2017