This is the first article in a series taken from the insideHPC Guide to Flexible HPC.
Organizations that implement high-performance computing (HPC) technologies have a wide range of requirements. From small manufacturing suppliers to national research institutions, using significant computing technologies is critical to creating innovative products and leading-edge research. No two HPC installations are the same.
For maximum return, budget, software requirements, performance and customization all must be considered before installing and operating a successful environment.
Challenges in HPC Environments
While all users of HPC technology want the fastest performance available, price and power consumption always seem to come into play, whether in the initial planning or at a later time. Standard performance measures exist that may or may not relate to an end user’s application mix, but it is important to understand the various benchmark results that go into determining the performance of a CPU, a server or an overall cluster.
The acknowledged leader in providing enterprise-grade, high-performance CPUs is Intel®. Although other companies do provide instruction-compatible CPUs for data center computing, the leader in both market share and technology is Intel. When looking at CPU performance, a number of factors must be considered. A measure of CPU performance, although not the most important one, is the clock rate. Basically, the faster the clock rate of the chip, the faster an application will run. The increase in clock rates on a specific line of CPUs marched forward at a predictable and amazing rate until about 2005, when clock rates leveled out at about 3GHz. As the clock rates on popular processors began to level off, the increase in the number of independent cores on a single chip began to increase. At first, two cores per chip appeared, quickly followed by quad-core designs. The technology has advanced at such a rapid rate that today’s most modern CPUs contain more than 20 cores per chip. The absolute performance of a system must account for the clock rate, instructions executed per clock tick, the number cores and the number of sockets in a single server.
In coming weeks, this series will consist of articles that explore:
- Challenges in HPC Environments (this article)
- Requirements in HPC Environments
- TYAN® Solutions For Flexible HPC
- Successful Implementations